aiLL 7/15/23 - Inside the Wild West World of AI

Description

Chapters

Transcript

0:19 welcome welcome
0:20 welcome welcome to the AI learning lab
0:24 we are learning about AI
0:26 well I don't know if we're learning
0:28 about AI we're chatting about Ai and
0:32 it's like a lab but it's more like a
0:33 home office so uh welcome welcome
0:38 welcome I'm here in Denver Colorado
0:40 Taylor Swift is just about to go on
0:43 stage at the giant Stadium full of
0:46 screaming
0:47 people and uh happy Saturday night
0:51 Saturday Saturday Saturday come to the
0:54 buck tractor pools
0:58 yeah anyone grow up in pencil Tucky oh
1:01 look at the nice heart already
1:04 fantastic thank you Tori appreciate
1:08 that um let's see so if you're new here
1:13 this is AI learning lab my name is Kyle
1:14 Shannon uh I'm an entrepreneur started a
1:17 bunch of
1:18 companies uh I've been doing this for a
1:21 long long time I started one of the
1:22 first digital agencies in the mid99s so
1:25 I got to be there for the beginning of
1:26 the worldwide web and now I'm seeing
1:29 this generative AI stuff and it's
1:30 feeling mighty familiar it's feeling a
1:33 lot like huh this is one of those one of
1:37 those transformative things although I
1:39 got to tell you this stuff feels bigger
1:42 bigger more profound all that sort of
1:44 stuff um let's see what
1:48 else I played with uh there's a company
1:51 called inflection they've got a new
1:53 thing called piie which is a it's a chat
1:56 bot like chat GPT but their app their
2:00 iPhone app um you can basically put it
2:02 in voice mode and so I was driving
2:05 around in the car just talking to piie
2:08 and I gave it like you know this this
2:10 lovely Australian kind of kind of voice
2:14 and I was just talking with Chach and
2:16 it's connected to the internet so it
2:17 looked up that Taylor Swift was doing
2:20 her thing tonight so it was connected to
2:21 the internet so that was kind of cool we
2:23 were going to CCO shopping and I said I
2:27 asked it what specials were at Costco at
2:30 at the location we were going to and it
2:32 was able to pull it up so it was Prett
2:33 pretty interesting it was Pi Pi pi. a or
2:37 hey pi.com um and then there's they've
2:41 got a an IOS app I don't think they have
2:43 an Android app they might but I don't
2:45 think so um but the company's in
2:48 flection and
2:50 um yeah just go get the IOS app if you
2:53 have IOS and and it's pi and then you um
2:57 you just launch it and then in the lower
2:59 right hand corner you pull up a a little
3:00 menu and you hit the the phone phone
3:02 icon and you just talk to it and then it
3:04 talks back it's pretty cool it was
3:07 pretty cool like it starts to it starts
3:09 to feel
3:12 like I you know imagine if if you know
3:16 Siri or Alexa or anything like that you
3:18 could ask it anything like I asked it I
3:22 asked it you know what were the big
3:23 concerts happening in Denver tonight I
3:25 asked it what were the specials at um
3:28 Costco and then I asked it to explain
3:30 quantum mechanics to me and then I asked
3:33 it to explain how some disease uh worked
3:37 in the body if you're new here my name's
3:38 Kyle Shannon this is the AI learning lab
3:41 uh if you have any questions about AI
3:42 pop them in the questions below happy to
3:44 talk about them this little window right
3:47 here let me let me show you what this is
3:49 all so if you have not played with
3:51 generative AI if you've not played with
3:53 chat GPT that's what all these all these
3:55 URLs this used to be a list of three now
3:57 it's five and it's there's two little
3:59 resour sources things below um you have
4:03 a lot of options now things are heating
4:05 up in the uh in the chat GPT space so
4:10 chat. open.com that top one that's
4:12 that's the uh the official chat GPT
4:15 website uh code interpreter is back it
4:17 was down yesterday I guess they found
4:20 some bugs in their beta test when they
4:21 rolled it out apparently they fixed it
4:23 cuz it's back you got bing.com which is
4:26 also chat GPT it's gp4 thank you for the
4:29 row I appreciate that buzz um one of the
4:33 nice things you can do at Bing is you
4:36 can make images so if you just go to
4:37 bing.com and type make me an image of I
4:40 don't know Taylor Swift in an arena
4:43 it'll do its best uh it's based on DOL
4:46 so it's not it's not very good but it's
4:48 it's free and it and it works um so
4:51 that's kind of fun and then claw.
4:55 AI has a lot going for it so this is
4:58 from a company called anthropic
5:00 the way they do their fine-tuning I'm a
5:02 real fan of these guys use human
5:04 reinforcement learning these guys use a
5:07 thing called constitutional learning
5:09 which it's good it's a good thing I can
5:11 talk more about that if anyone wants to
5:13 geek out some other things about claw.
5:17 you can upload files to it so similar to
5:19 code interpreter you can upload up to
5:21 five files here I think 10 megabytes
5:23 each and then start interacting with
5:25 them so take some PDFs of whatever you
5:28 want to learn or or take documentation
5:31 for some code you want to work on or
5:34 whatever throw a GitHub repository up
5:36 there I don't know whatever the hell you
5:37 want to do uh and what else it's also
5:40 got a 100,000 token context window so
5:43 that's somewhere between 50 and 75,000
5:46 words you can put in your prompt copy
5:48 and paste your great American novel and
5:51 start editing it so that's that one
5:54 po.com gives you access to uh three
5:58 different companies models it's got claw
6:01 it's got open Ai and they've got their
6:02 own and then Bard is Google's entry into
6:07 the space and
6:08 Bard had been absent from my list for
6:12 the past two months or so because it
6:15 sucks well it still sucks but it's it
6:18 sucks it sucks less enough now like it
6:21 it is close enough to usable now that I
6:23 put it on the list one of the things
6:25 they did was they copied the interface
6:27 of Po and open AI kind of merg them um
6:32 so it it's got a very familiar interface
6:34 now uh and it's actually usable it's
6:36 actually not bad the the writing isn't
6:38 as bad as it used to be so I don't know
6:39 what they did over there but but
6:41 Google's waking up the dragon is waking
6:43 up and then prompt. chat down below
6:46 that'll teach you how to do prompting
6:47 and how to teach these things to act
6:49 like certain personas uh yeah by the way
6:51 if you would just go ahead and tap the
6:53 the like screen uh let people know we're
6:55 here um share the live if you're willing
6:57 to do that and then future cia. that's a
7:01 directory of probably 4,000 at this
7:04 point um AI
7:08 tools and what else uh I mean there's
7:12 just Lots there I can't keep up with
7:14 this [ __ ] anymore but futurepedia doio
7:16 does so so I I have delegated
7:28 that hey
7:30 good to see you retr punk yo I got
7:33 barred to code a simple website from an
7:35 image of a sketch oh that's super cool
7:38 retro that was the thing that they
7:39 showed that that they showed in the uh
7:42 the
7:43 gp4 launch demo um the CEO from from
7:48 open AI showed that so um I'm curious if
7:52 uh if Bing does that I haven't tried it
7:54 I was going to try that today and didn't
7:56 get around to it but but you can you can
7:59 upload images to Bing and to Bard and do
8:02 analysis on them so it looks like you
8:04 got it to code a website based on a
8:06 napkin sketch that's impressive very
8:10 cool all right if you have questions
8:12 about uh anything AI pop it in the com
8:15 comments below if you want to troll me
8:17 you got to be creative don't just get
8:19 you can't just be a nasty troll it's got
8:21 to be creative or we just they it it you
8:24 know what it is if you write a shitty
8:26 troll it just Ricochet off me it doesn't
8:29 hurt at all
8:30 if you do one that's somewhere between a
8:32 compliment and an insult that that digs
8:35 deep and as a troll isn't that what you
8:37 want you want to hurt people you don't
8:39 want it to just
8:41 ping like some spaghetti western no you
8:45 want your bullets to
8:48 land all
8:50 right what else oh I swear a little if
8:52 you don't like swearing I wouldn't stay
8:54 it's Saturday night why are you here but
8:56 so am I so let's hang out um if you're
8:59 new here Taylor Swift is uh just about
9:01 to go on stage just down the block we
9:04 just we just drove by there to see what
9:06 a zoo it is it's
9:07 Insanity it's Insanity I think she's
9:11 going to clear someone said she's going
9:12 to profit $550 million from this tour
9:16 they're going to gross more than a
9:17 billion dollars from the Aros tour
9:20 crazy oh that's the other thing yeah
9:23 good point uh male M KS Kansas m I think
9:30 um Claude is only available in the US
9:33 and the UK it's not available in Canada
9:36 and I think Bard might one of the other
9:38 ones is is banned from Canada probably
9:41 because of some privacy law or something
9:42 like
9:43 that they're like we want to use all
9:46 your data and then some countries are
9:48 like well then no and they're like fine
9:50 and that's how they talk I don't know if
9:52 you've ever heard Executives at tech
9:54 companies they're like no I don't want
9:56 to do it fine go ahead and be that way
10:00 Canada then you can't use our special
10:03 toy never been to San Francisco H it's
10:07 how it works why do you think they're
10:09 worth billions of dollars you think it's
10:11 just because of the technology no it's
10:12 because of the attitude all right
10:14 explain Ai and chat GPT good Lord that's
10:19 wide okay AI stands for artificial
10:21 intelligence and it's the wrong term um
10:24 technically all the stuff we're playing
10:25 with is machine learning I think it's AI
10:27 is the broader category machine learning
10:29 is the specific
10:31 flavor chat GPT so the the probably the
10:35 best way to explain chat GPT is to have
10:37 you go there and start playing so go to
10:40 actually well you can go to either one
10:42 go to chat. open.com or bing.com or
10:46 claw. either of the top three just go to
10:49 go to one of those sign up for an
10:51 account at Bing you don't have to sign
10:53 up for an account when you go to
10:54 bing.com click on the chat button here I
10:58 I'll do it let's let's let's do a little
11:00 demo that's
11:03 barred here's
11:07 Bing okay so when you go to bing.com it
11:10 looks like just a regular search box
11:12 like a Google search box and up here
11:14 there's a chat button so go to
11:18 chat
11:19 and you have you have a couple of
11:21 different modes you've got creative
11:23 balanc and precise it'll default to
11:25 precise and then the way this works is
11:29 what the prompt here says is ask me
11:31 anything and so it literally is that you
11:33 can ask it anything
11:36 so think of this as a Google search but
11:39 rather than it returning here's a list
11:42 of 20 links it just answers you um and
11:46 it can answer you
11:48 incredibly sophisticated answers so for
11:51 example let's
11:53 say I
11:55 have a
11:57 chubby black and white
12:01 dog and I
12:03 want to write uh a
12:08 country
12:12 country
12:13 song can you
12:17 help
12:19 and it thinks a little
12:25 bit oh it's telling me how to write
12:28 one stop
12:31 no I want you to write one like gu
12:37 Brooks
12:47 would sure here's your
12:52 song
12:54 Wait so so much for doing a demo live
12:57 huh okay let's see write me a song about
13:02 a
13:03 chubby black and white dog good Lord see
13:09 I was trying to get too fancy with the
13:12 words and part of the part of the fun
13:14 and frustration of these things is
13:15 getting it to answer
13:20 um huh this is just not doing it oh
13:22 maybe I need to flip to creative mode
13:25 maybe that's maybe that's part of the
13:26 issue it looks like they've mucked with
13:28 the uh
13:30 um write let's see write me a country
13:35 song about a
13:37 chubby black and white
13:43 dog so supposed to be explaining Chad
13:46 gbt here we go verse
13:51 one there's a chubby black and white dog
13:53 that lives down the road from me he's
13:55 wagging his tail and barking at the
13:57 trees So It Goes On to write this song
14:00 okay I'm going to stop that so now I
14:01 could say um
14:03 explain um
14:06 how uh
14:09 photosynthesis
14:28 works
14:31 and it explains that and it also gives
14:33 me some links to Wikipedia so Bing is
14:36 actually connected to the internet so
14:38 there it is explaining
14:40 photosynthesis um I could say make a
14:43 table
14:45 comparing
14:50 photosynthesis um and um I don't know
14:54 some and gravity let's take two concepts
14:58 that shouldn't really be
15:00 compared
15:03 and see if it can make us a
15:05 table comparing these two processes
15:13 processes so here's a table comparing
15:16 these two things with
15:18 citations so I can do that so now I can
15:21 say um write me a python
15:27 application that
15:31 visualizes
15:39 photosynthesis and I put in I I have a
15:41 couple of typos in there it doesn't care
15:43 about
15:51 typos it imported Turtle it imported
15:54 some libraries it's determining some
15:58 colors it's going slow but it's writing
16:00 code all right so I'm going to stop that
16:02 so you can write code with it um on Bing
16:04 you can also generate an images make me
16:07 a A Creative
16:11 Image that brings the idea of
16:17 photo synthesis to
16:26 life okay I'll try to create that
16:32 and then it's off thinking about it your
16:34 image is generating and it'll give us
16:40 four and there you go there's four
16:42 images that bring the idea of
16:45 photosynthesis to life so so chat GPD so
16:50 all of these large language
16:52 models are an entirely new paradigm the
16:55 way they work so so let me flip this
16:58 back
17:03 so the
17:05 GPT thank you thank you for the Roses
17:08 Welcome the GPT and chat GPT stands for
17:12 generative pre-trained Transformer
17:14 Transformers the technology for how they
17:16 do what they do pre-trained means it's
17:19 been pre-trained on what on all of the
17:21 public internet essentially so it kind
17:23 of knows all of the stuff that's put on
17:25 the public internet you know within
17:27 reason for the past 30 or 40 years so it
17:30 knows a lot like it knows a lot like
17:33 obscure things and then generative means
17:35 it's generating that content so so
17:38 that's kind of what it is and and that's
17:41 this that's this thing that has captured
17:43 people's imaginations why because it can
17:46 dramatically
17:47 improve um anything you do now some
17:51 caveats about it it does this thing
17:53 called
17:54 hallucinating where it just confidently
17:57 will lie to you here here's here's a
17:59 fact about photosynthesis and it just
18:01 made it up it's not right or um could
18:05 you you know explain a thing to me and
18:08 and you know give me the a hyperlink to
18:10 where you got that information it'll
18:11 give you a hyperlink very often that
18:13 hyperlink doesn't work it's just made it
18:15 up so that's something you have to be
18:17 careful with with it it's we're very new
18:20 very new in this so that kind of stuff
18:21 is going to improve but that's something
18:23 you need to know right now um anyway
18:26 that's about it and these things are all
18:28 things that you should go play with and
18:31 kind of figure out which one works for
18:32 you they're all sort of different
18:33 flavors of the same thing um if you're
18:36 just starting out I would start out at
18:38 chat. open.com that's chat GPT that
18:41 they're the people that made it their
18:43 GPT 4 is better than the other things by
18:46 a lot uh it's it's
18:49 fancier uh they they just released its
18:52 architecture or it was leaked this week
18:55 um and the architecture is pretty insane
18:57 and so I don't think think I don't think
18:59 anything's really close to gp4 so you
19:02 can get gp4 at the top one the second
19:05 one and the fourth
19:06 one all right hope that helped that was
19:09 a lot and it started started rough man
19:14 Bing was not cooperating write me a song
19:16 and it was like well you could write a
19:17 song like this one that gu books Brooks
19:20 wrote well no write me one like gu
19:22 Brooks well here's another one that g
19:24 Brooks wrote here's a picture of G in
19:27 fact it's like just
19:31 stop all right Bing is the weakest of
19:34 the big guys me thinks uh no Bard B Bard
19:38 is the weakest of the big guys Bing is
19:41 actually gp4 but it's been neutered it's
19:44 been it's it's been neutered it's
19:46 getting a little better but it's very
19:49 quirky it's it it's very Microsoft in
19:52 its answers in my opinion I don't know
19:54 what they did to fine-tune it but it's
19:56 it somehow feels like a Microsoft
19:58 product in its answers um I would say
20:01 right now gp4 open.com is number one I'm
20:06 kind of leaning with claw. a from
20:08 anthropic is number two um and then Bing
20:13 and then POS just a it's a collection of
20:15 a bunch of them so po po doesn't really
20:17 count and then Bard is probably the
20:18 fourth fourth strongest in in my humble
20:22 opinion and then there's also this new
20:24 thing called Pi Pi which you can get to
20:27 at hei.com
20:29 or you can get the PI app on iOS and you
20:32 can have a conversation with pi I did it
20:35 in my car today you talk to it and you
20:37 pick one of six voices and you just are
20:40 are you're now talking with a virtual
20:43 agent and it's it's fast enough that
20:44 it's like the lag isn't too weird and I
20:47 was asking it everything from like who's
20:49 who's playing a big concert tonight in
20:51 Denver to um explain quantum mechanics
20:55 to me I did all my parlor tricks with it
20:57 and it did it pretty good so that that
20:58 was that was
21:00 interesting all right let's
21:10 see uh let's see yo I got Bing to code a
21:13 simple website from a sketch that's
21:15 super cool that really is
21:17 cool um okay here we are I lost myself
21:20 yeah I'll share in the AI Salon oh that
21:22 would be great that would be great retr
21:24 Punk yeah that that's let's let's do
21:26 that um drop me an email my my contact
21:30 info is in in the bio there's a link
21:32 tree there drop me an email and uh make
21:36 sure I'll put you on the agenda for next
21:37 Tuesday is po worth paying for it it
21:41 was I don't think so
21:45 now so the reason I paid for po it's
21:49 actually a good reminder I think I'm
21:50 going to cancel my subscription the
21:52 reason I paid for PO is that was the
21:53 only place that you could play with the
21:56 claw models and so at po they had a
21:59 100,000 token context window that you
22:02 could only get through po.com or yeah uh
22:05 yeah po.com had had claud's anthropics
22:08 100,000 token model but now that we have
22:11 claw. a that's claude2 with 100,000
22:15 token model and it's free and you can
22:19 upload files to it so so Claude is
22:21 definitely one to start to pay attention
22:23 to they're they're they're a major
22:26 player in fact to indicate how much of a
22:29 major player anthropic is the company
22:32 behind Claude when the White House had
22:35 the CEOs of four AI companies like 3
22:39 months ago two months ago whatever it
22:40 was it was Microsoft Google Sam Alman
22:45 from open Ai and the CEO of anthropic
22:48 those were the four CEOs so anthropic is
22:51 in the room with the with the the big
22:54 boys um so they're they're a major
22:57 player so I would I would I would start
23:00 messing with Claude and actually you
23:02 know it's
23:03 probably if you want to get a little
23:05 geeky with this stuff it's worth geeking
23:09 out on Claude and and how well it
23:12 processes documents and code interpreter
23:15 with open AI um so
23:19 yeah what's your opinion on being and
23:22 Claude 2 so I was just talking about
23:24 that a bit I you know here's the deal
23:26 with well Bing and Claude two are both
23:29 free so so to get the gp4 at open AI you
23:34 had to pay 20 bucks a month if you want
23:36 to claw two you had to pay po 20 bucks a
23:39 month so they're both free so so that's
23:42 nice
23:47 um being here here's here's a
23:51 reason to consider
23:54 B Microsoft is rolling what they're
23:57 calling C pilot but it's gp4 they're
24:00 rolling gp4 across their entire product
24:04 line from their cloud services to
24:07 Dynamics their CRM to the office suite
24:10 so you know imagine within word you've
24:13 got gp4 just built into word you've got
24:15 it built into um Excel you've got it
24:18 built into PowerPoint make me an outline
24:21 based on that word
24:24 document
24:26 um it's going to be built into windows
24:28 11 the the other thing that Microsoft is
24:31 doing well is as they're adding
24:33 functionality to it you know they're
24:35 just kind of quietly rolling it out like
24:37 now you can upload an image to Bing and
24:40 tell it to like I I saw an example right
24:43 before I went live where a it was a
24:46 marketing person and she uploaded an
24:49 image and said I want you to act like a
24:51 you know Master copywriter from a
24:53 marketing thing and I want you to write
24:54 me ad copy for this image and it did it
24:58 um um so so and then Bing can also you
25:01 can also generate images within within
25:03 context right so you don't have to go to
25:05 a separate service like mid journey to
25:07 generate images the downsides of being
25:10 is it's been intellectually neutered
25:13 like like they put lots of safety guard
25:15 rails on it CU they got in trouble um
25:19 and um and and it uses dolly for its
25:23 image generation which sucks but it
25:27 ain't bad it just ain't mid Journey so
25:30 that's big and then Claude is it's not
25:32 quite gp4 quality but Claude is really
25:36 impressive 100,000 token context window
25:39 means between 50 and 75,000 words you
25:42 could copy and paste the text of a novel
25:46 of a full novel into your prompt and it
25:49 won't choke on it it won't say that's
25:50 too much for me to think about it'll
25:52 just do it um so I think they're both
25:55 really interesting really good that's
25:56 like the the whole purpose of this list
25:59 is like if I don't know [ __ ] about AI if
26:01 I'm just trying to get started where do
26:03 I start that's what this list is so you
26:06 now have lots of options here some paid
26:08 some not paid I still think it's worth
26:11 paying 20 bucks a month to chat GPT cuz
26:14 they're in the lead right now no one has
26:15 caught them yet um claude's free so you
26:18 don't have to even consider that um Bard
26:21 is free
26:24 um Bard I
26:27 don't
26:30 I my My Philosophy on Bard right now is
26:32 I'm going to keep checking in it's like
26:35 it's like you know it's like it's like
26:37 that slow student in class I'll keep
26:39 checking in and see if he can get the
26:40 addition thing down it's actually good
26:43 at math and coding it's not good at
26:48 language oh yeah some some of the some
26:51 of the things like I think Claude is not
26:53 available in Canada so if you're in
26:55 Canada sorry someone just left me a
26:57 comment and is not available in
27:00 Canada well yeah I know all right can
27:04 you give a few practical examples of how
27:07 a small business owner could utilize
27:10 AI
27:17 yes trying to think where to do it and
27:19 what to
27:21 do I'll tell you a little bit about how
27:23 I'm I'm using it so
27:27 um
27:29 one way you can use it is for simple
27:31 productivity gains um so that might look
27:34 like
27:38 um let's say let's say you have to you
27:41 know depending on what your business is
27:44 you you have to write professionally
27:46 worded
27:47 emails saying pay your [ __ ]
27:51 bill
27:52 um you could have chat GPT just help you
27:55 write them just say you know here's an
27:57 example example letter that I've written
27:59 before here's the new person I want to
28:01 write it to here's the situation write
28:03 me a new letter right so you can just
28:05 sort of copy and paste stuff that you've
28:07 done before paste it make it more
28:09 efficient rather than manually taking
28:11 notes for a zoom call you could just
28:15 record the transcript you know record
28:17 the meeting get the the transcript and
28:20 have one of these large language models
28:22 summarize that meeting there's a lot of
28:24 plugins that do that now I think they've
28:26 rolled um that functionality into Zoom
28:29 or they're about to um uh your favorite
28:34 web dude just said rip data scientists
28:36 yeah I I'll talk about uh code
28:39 interpreter at chat GPT code interpreter
28:41 is a full-on data analyst in a box it's
28:44 amazing um so so you could just do some
28:47 some productivity gains like that um you
28:50 can also start to do things like um if
28:54 if you want to do creative ideation like
28:56 let's say you want to come up with a
28:57 market marketing campaign you could say
28:59 hey I want chat GPT I want you to act
29:02 like a you know a master you know
29:04 marketing executive and I want you to
29:06 help me come up with a strategy and then
29:08 you explain your business you explain
29:10 your struggles you explain you know is
29:12 it top of the funnel midf funnel you
29:14 know closing whatever it is um explain
29:16 your situation give it a lot of context
29:19 and tell it to write you a marketing
29:20 plan and then have it turn that
29:22 marketing plan into you know a tactical
29:25 plan and then take that tactical plan
29:27 and have it write you you know tweets
29:29 and and Instagram posts and da da da um
29:34 so you can do things like that with code
29:36 interpreter you could do things like
29:39 take documents from your business upload
29:42 it to code interpreter and start
29:43 interacting with it so it could run
29:46 analysis on you know um you know cost
29:49 cost versus you know revenue and you
29:53 know what are the trends you're seeing
29:54 and you know has have there been are
29:56 there any anomalies in the the data or
29:58 do you see any inflection points in the
30:00 data and really just start interact
30:02 acting with it like that um lots and
30:04 lots of things if anyone else in the
30:06 chat has ideas feel free to pop them in
30:08 there um for for silhouette silhouettes
30:11 by Chris okay please tell us ones that
30:14 build websites for us there are a ton
30:17 out there go to futurepedia doio and go
30:20 in the like web design there's there's a
30:22 section just for that um I can't keep up
30:25 with them anymore there's lots of them
30:27 though I had a virtual machine set up so
30:29 friends abroad could try out American
30:32 only do DS oh that's pretty smart yeah
30:34 and if you're in Canada you can try a
30:36 VPN as well I don't know if they work
30:38 were you surprised to hear Gary ve talk
30:40 about you in the latest podcast I don't
30:44 think he did did he is that
30:46 true
30:50 Matt I can't imagine that but okay if
30:54 you did that's
30:56 cool
30:59 I'm an ex agency guy so I know I know I
31:01 know Gary decently well and I've talked
31:02 to James orini at the Sasha group a lot
31:04 so I know them but I haven't I haven't
31:06 talked to Gary I think I've just talked
31:08 to him once or twice over the years so
31:11 if he mentioned me that's that's news to
31:13 me my it is Rusty for a decade and need
31:16 a site to sell a few items which AI
31:19 should I use I would you know it's
31:22 probably
31:24 worth well I mean I would I would go to
31:26 futurepedia I so I don't know there's so
31:31 many ways to skin that cat you need to
31:33 start with what are you trying to
31:35 accomplish do you just want a basic site
31:36 do you want a basic e-commerce site you
31:38 could go to something like square space
31:40 or web flow or one of the sites that
31:43 just you know you pick a template build
31:44 your site fill in the content and go so
31:47 an option is you don't need to use
31:49 AI if you want to use AI you could say
31:52 hey I'm going to use AI to write the
31:54 content of my website and the structure
31:56 of my website but I'm going to just go
31:58 pick a template to dump all that [ __ ]
32:00 into
32:03 um
32:06 don't this is this is a general comment
32:09 and and recommendation for everyone
32:13 don't mistake the tool for the strategy
32:16 your strategy is you need to build a
32:18 site to sell a few items I would start
32:20 with what do you need first and then I
32:23 would start to ask what's the best tool
32:24 to accomplish that um if you're not a
32:28 good writer and you don't you know you
32:30 said your it skills aren't good if your
32:32 if your copy skills are good and you
32:33 already have copy then you know maybe
32:37 you could use one of these AI tools the
32:40 thing about the AI tools right now for
32:42 the most part they're really slick but
32:45 they tend to be very
32:47 shallow so you'll say go make me a
32:49 website and it'll go make you a website
32:51 and you're like oh that's really slick
32:52 and then you go look at it and you're
32:53 like well yeah that's kind of like basic
32:56 and I could
32:58 I want something more than that right so
33:00 if you want something more sophisticated
33:02 you might want to start with you know
33:04 something like Squarespace or web flow
33:06 or one of those guys so I don't know
33:07 it's it's uh you know go play with the
33:10 tools there use any one of these you
33:13 know chat GPT type things for copy uh
33:17 chat gp4 is by F by far the best at copy
33:20 and coding and everything so I would
33:23 personally use uh chat. open.com the
33:27 paid version if I were doing that um for
33:30 the copy and even if you're going to do
33:31 some slight coding or you want to put a
33:33 tool in there and you can't find a
33:35 plugin for whatever tool you're using so
33:37 I know that was kind of a shitty mamby
33:39 pami answer but what are you going to do
33:42 all right Bing is the weakest oh we we
33:45 saw that
33:47 earlier AI learning lab did you build
33:50 this did I build what what did I build
33:54 what I think I was just showing big oh
33:57 oh did I build this or or was that chat
33:59 GPT the thing I was showing earlier was
34:01 big this
34:03 thing yeah this is no I did not build
34:06 this so a little bit about my background
34:08 I'm I'm an entrepreneur I'm not a data
34:10 scientist I'm not a machine learning guy
34:12 I'm an entrepreneur I'm the CEO and
34:16 founder of a company called storyvine
34:18 which is an automated video uh
34:20 storytelling platform uh it's actually
34:22 it we've been around for 11 years we do
34:24 a lot of work in the healthcare and
34:25 Pharma space um we don't use a lot of AI
34:28 to assemble our videos Auto magically um
34:31 we're rolling a bunch of AI into the
34:33 platform um but we use more traditional
34:36 assembly methods I guess would be the
34:38 term and
34:40 um and I was also I've been an
34:42 entrepreneur for for many many years
34:44 I've started 12 or 15 companies I I
34:46 co-founded agency.com which was one of
34:48 the early digital agencies in the
34:51 mid90s and um so I was there for the
34:55 early days of the worldwide web and holy
34:57 ho [ __ ] the parallels to from the early
35:00 days of the web to the early days of
35:02 this generative AI stuff are insane um
35:05 and so the focus my focus is not so much
35:08 the technology of machine learning in AI
35:11 but more the application of it so how
35:13 are we taking these tools where do you
35:15 start if you're trying to figure out
35:17 this stuff for yourself as an individual
35:21 or as a um as a company where do you
35:26 start what do you do how do you do it
35:29 what's [ __ ] what's
35:30 real if you want to make money with chat
35:33 GPT do you go on Tik Tok and listen to
35:35 these guys you can make money a
35:38 CH no you
35:41 don't cuz those guys are idiots and they
35:43 talk like that and no one should have to
35:46 S suffer through that and do you know
35:49 how they're making money with chat
35:51 GPT by getting you to pay for them to
35:53 tell you how to make money with
35:55 chbt and they chat GPT write the whole
35:58 [ __ ]
36:01 script they're evil little
36:08 [Applause]
36:09 [Laughter]
36:14 scabs can you tell him a gen xer bitter
36:17 you betcha all
36:19 right uh uh uh uh uh what site I was
36:23 using before was Bing did you code this
36:25 app I did not
36:27 check out AI PRM as well I will check
36:30 that
36:33 out uh love you my genx ADHD twin yeah
36:36 we got we got some ADHD in the house we
36:39 got we got some ADHD fans listen I I
36:42 think this early in in an industry like
36:45 something like generative AI having add
36:48 is a superpower cuz this shit's moving
36:50 so fast like I mean if you've been
36:52 watching these lives at all I mean you
36:54 know this this list is is changing
36:58 um there okay here's the thing if you
37:03 feel like you're late Andor you feel
37:06 like you just can't keep up you're not
37:09 late and you can't keep up no one can
37:12 keep up like if you just take a teeny
37:15 little slice of just say these tools the
37:20 chat GPT like tools there are at least
37:23 two other tools that could be on this
37:25 list that aren't and there's probably 10
37:27 others below that and then if you go to
37:30 futurepedia doio there's probably 1,500
37:35 tools that are built on top of the core
37:38 technology sitting underneath these
37:40 things so if you just take the thin
37:42 slice of I want to use AI to write [ __ ]
37:46 it is completely overwhelming every
37:48 slice of the AI pie is completely
37:50 overwhelming right now um
37:54 so my recommendation is just start it
37:57 doesn't [ __ ] matter where you start
37:59 it doesn't matter what you do if you
38:01 just use chat GPT to make a meal plan
38:04 for you and help create a shopping list
38:06 so your trip to the grocery store isn't
38:08 such a [ __ ]
38:10 nightmare you're
38:13 winning and then if you can dig a little
38:15 bit deeper and start to help it organize
38:19 your neighborhood block party that
38:22 you're planning for August great and if
38:24 you can help it have you
38:27 Market your business or your side gig or
38:32 help you figure something out or help
38:33 you learn something or help your kid do
38:35 homework great you're winning just keep
38:38 peeling the onion the the investment
38:46 in feeling that uncomfortable feeling of
38:49 I don't know what the [ __ ] I'm doing and
38:51 who are these people and what is this
38:54 AI like I still feel that every day
39:01 and what I can tell you is that the
39:04 investment right now that you put every
39:07 hour that you put into just getting your
39:08 head around what this [ __ ] is is going
39:10 to pay you dividends it's going to pay
39:12 you
39:13 dividends because everyone is going to
39:15 have to be AI literate to some degree
39:18 because these tools that right now these
39:20 tools exist outside of our normal work
39:23 life and life life and things like that
39:26 these are very quickly starting to
39:28 proliferate into all the tools we
39:32 use and so people that kind of have
39:34 their head around what they are and how
39:36 they do what they do are going to have
39:38 an
39:39 advantage and especially as
39:42 um categories of work start to get
39:46 disrupted like I'm a copywriter at a big
39:49 agency well that's going to get
39:52 disrupted so are you a copywriter that's
39:54 AI literate at a big agency because
39:57 you'll have a hell of a lot easier
40:00 conversation in my opinion than someone
40:03 who's just like no no I just like
40:04 writing I like writing the oldfashioned
40:06 way all right um I tried to write a
40:09 paper and every fact was false yes Oreo
40:12 so couple of things about hallucinations
40:16 there's there's two things
40:17 about it's funny that they call them
40:19 hallucinations I got a guitar I where's
40:22 my guitar oh it's down in the other room
40:25 um couple of things about huc
40:28 hallucinations one
40:32 is they they called them hallucinations
40:35 like you know it's kind of a cool thing
40:36 like it hallucinates no it's just
40:38 [ __ ] lying I like to describe chat PT
40:40 as mansplaining as a service
40:44 um
40:48 so they they hallucinate for for a
40:51 couple of reasons the the sort of
40:53 technical reason why they hallucinate is
40:56 the way the tool actually works is when
40:58 you write a
41:01 prompt what the technology does is it
41:04 basically turns those words into a
41:06 mathematical formula that creates a
41:10 probability of what's the next most
41:11 probable word after this prompt the and
41:16 then it takes all of that and it says
41:18 well what's the next most probable word
41:20 after this and the dog and then and then
41:25 so it's literally just doing a ability
41:28 sentence complete over and over and over
41:30 again so when you see it write you a
41:32 whole missive what did you say you're
41:34 writing a paper when you see it write
41:36 your paper it's doing that one word at a
41:37 time so it doesn't know what it's
41:40 written and so so it it's been
41:43 programmed to want to please you so
41:45 you're like give me a well researched
41:46 paper and it's like well here's the
41:47 paper and then it just says here's all
41:49 the research it's not actually checking
41:52 any of that [ __ ] to see if it's
41:54 real now one of of the ways that you can
41:57 improve the hallucinations is by
42:00 narrowing the context so in your prompt
42:04 if you just say write me a paper on you
42:07 know the Napoleonic Wars it's going to
42:11 write you a paper pulling from sort of
42:13 everything anyone's ever written about
42:15 Napoleon in all of the internet
42:17 including all of the horrific student
42:19 papers that were wrong and all of the
42:21 conspiracy theories that said something
42:23 weird about Napoleon but if you say I
42:25 want you to act like
42:27 a uh historian that specializes in the
42:31 Napoleonic Wars and you know who
42:33 Napoleon was and what this was if you
42:36 actually narrow the context you you sort
42:38 of tell chat gbt ignore all that other
42:41 [ __ ] I want you just to act like this
42:43 one kind of specialist that's the reason
42:46 act like a screenwriter act like a
42:49 mathematician act like a historian why
42:52 that works is what you're doing is
42:53 you're kind of filtering out the
42:55 [ __ ]
42:57 that is not the stuff you want in your
42:59 paper so now even if you do that it will
43:04 still hallucinate because it doesn't
43:05 know what it's writing so you absolutely
43:08 are going to have to go check your
43:09 things now one thing that you could do
43:12 is use Bing instead of chat GPT because
43:15 Bing is
43:17 gp4 and Bing does searches on the
43:20 internet and it includes citations in
43:23 context of what it's writing you'll need
43:26 to Dicker around with the three modes
43:28 there's the creative mode the balanced
43:29 mode and the precise mode at in the
43:32 precise mode it might not it might give
43:35 you stuff that's too dry so you might
43:37 need balanced but bing bing might do
43:39 better in terms of giving you citations
43:41 that
43:42 actually link to something and are real
43:45 so hope that helps okay yay you're on
43:48 yay I've been waiting what's wrong with
43:50 me I don't what's wrong with all of us
43:53 one of the things that's wrong with you
43:54 is you were waiting for me to go live
43:59 oh my god oh lordy lordy lordy lordy we
44:03 having fun this Saturday night oh we're
44:06 having fun this Saturday night oh I
44:08 loves me some may I talk we get on we
44:11 drink our ass coffee we slurp it up we
44:14 slurp it up pinkies
44:18 out cuz we're
44:21 classy all
44:22 right but thank you Miss bir appreciate
44:25 you being here
44:27 and that you enjoy these I do intend to
44:30 I do aim to please and and create you
44:33 know I I don't know like hey I just like
44:35 to have fun I hope people are learning
44:43 [ __ ] Bard's hallucinations are Beyond
44:46 acceptable yeah Bard still kind of sucks
44:48 it's why it's at the bottom of the
44:50 list like the OCD in me wants to make
44:52 them like longest to shortest but Bard
44:55 needs to be at the bottom
44:57 Bard barely made the list but Bard's at
44:59 least worth playing with now it wasn't
45:01 for a long time how do you spell Pi P
45:04 not
45:06 piie so it's I I think the website for
45:08 pi is hey.com
45:11 or pi. a both of those will get you
45:15 there and then the IOS app is called Pi
45:21 Pi um can you repeat Pi what please pi.
45:26 a I cheers from Argentina oh that's cool
45:28 hello from Argentina cheers from Denver
45:32 Taylor Swift sends her love she's
45:34 probably performing right now just down
45:36 that way like 10
45:39 minutes um best AI for teachers any of
45:43 these I I mean chat GPT like like
45:46 actually seriously for teachers chat GPT
45:50 because they're the ones that invented
45:52 it it's the best
45:55 um and and I think you you know it's not
45:58 so much what's the best AI for teachers
46:02 I think
46:05 it's how do teachers need to think about
46:07 this [ __ ]
46:11 right I personally feel it's imperative
46:14 for teachers to understand what this
46:17 stuff makes
46:18 possible and then find a way to
46:20 incorporate it into the classroom
46:23 because these tools are going to be
46:25 everywhere they're going to be built
46:27 into operating systems of phones um
46:30 they're going to be rolled out into all
46:32 of The Office 365 products they're going
46:34 to be in all of Google docs and Google
46:36 Sheets and presentation software they're
46:40 already in notion they're going to be in
46:41 all the notet taking
46:43 software
46:47 um and so some of the things we're
46:49 talking about here these things
46:51 hallucinate um the writing's kind of
46:53 shitty sometimes it doesn't give you
46:55 what you want um sometimes it's so
46:58 remarkable your face falls off and
47:00 you're like oh my God it can do
47:03 that and
47:06 so
47:07 the the the the way that I've been using
47:13 these tools and interacting with them
47:14 has been very much like a collaborator
47:17 it's not like this is a search engine
47:19 this is a this is a a profoundly
47:21 different tool than Google search Google
47:24 search you put in a concise little thing
47:26 and it gives you a long list of [ __ ] and
47:27 then it's on you to go figure it out
47:30 this thing is you interact with it not
47:33 so much as like a call and response but
47:35 more like an interaction think of it
47:41 okay what these tools are right now and
47:44 they're going to get better but right
47:45 now what they are is their idiot
47:48 savants you need to give them lots of
47:51 directions cuz if you don't they're just
47:54 idiots but they're savants they know
47:57 everything like they're way smarter than
47:59 any of us are in terms of just the
48:01 amount of knowledge in their virtual
48:04 neural networked freaky ass Transformer
48:07 driven brain
48:10 um and so so what they actually require
48:15 to use well is critical thinking and so
48:19 that's a skill like how do you bring
48:21 that into the classroom well you bring
48:22 it into the classroom and let's explore
48:24 what do they do well what do they not do
48:26 well um let's have it write a paper for
48:29 us right so instead of trying to say
48:31 well we're going to get you know AI
48:33 detectors and we're going to determine
48:35 if anyone's cheating they ran the um
48:38 Declaration of Independence through one
48:39 of those cheaters and it said it was
48:41 written by AI so those detection tools
48:43 are [ __ ] too and SM moderately smart
48:47 students can get around it and will you
48:49 know they'll they'll find ways to cheat
48:52 um but but rather than saying don't use
48:56 these tools to write your paper use the
48:59 tool to write the paper in the classroom
49:01 and then maybe the assignment is you
49:03 know go read what this thing wrote and
49:05 tell me what did it get wrong what were
49:08 the what were the um logical arguments
49:11 that it used did it miss anything would
49:13 you have written it different right get
49:15 those creative the the critical thinking
49:17 skills going because I think critical
49:19 thinking is going to be the
49:21 skill moving forward that that remains a
49:26 premium value because these things are
49:28 idiots and so you need people that can
49:31 understand what these tools are what
49:33 they make possible and then how to
49:35 Wrangle them to to do good stuff so
49:39 great question
49:41 though all right by the way if you're
49:43 new here my name is Kyle Shannon this AI
49:45 learning lab pop any questions in the
49:46 comments below I do try to answer them
49:48 all if I'm not answering your questions
49:50 cuz I haven't gotten to it yet but you
49:51 can ask the people that have been here
49:53 for a while they know I answer as many I
49:56 can claud's not available in Argentina
49:59 either that's yeah it's really
50:00 interesting I mean Claud claw. just came
50:03 out so they might be taking a very uh
50:06 conservative approach with rolling it
50:07 out I would say you know try a VPN or
50:10 try some of the other tools I would
50:12 assume Bing is available for
50:17 you could a VPN help countries that
50:20 don't have it I'm talking about Claude
50:21 that's what I was just saying I think so
50:23 but I don't I don't know you know what's
50:25 your back so my background is um I'm a
50:27 Storyteller by training I have a degree
50:29 in acting I moved to New York City out
50:31 of school and
50:34 um started and ran a theater company and
50:37 then I went through a manic period no
50:40 you I went through a manic period where
50:42 I wrote seven feature length screenplays
50:44 in 2 years um and then shortly after
50:48 that uh the worldwide web kind of showed
50:50 up on the scene and I started one of the
50:52 first online magazines called Urban
50:54 desires and I started one of the first
50:56 digital agencies company called
50:58 agency.com and we built a lot of the
51:01 first websites for the Fortune 500 like
51:03 we built British Airways website we
51:05 built their ticketing system their web
51:07 web ticketing system um we built
51:09 Coca-Cola's website we you know it was
51:11 it was insane it was absolutely insane
51:13 we grew really big really
51:15 fast um I was in New York City all that
51:18 time and then um I've been in Denver for
51:22 14 years I've been in my current company
51:24 Story vine for 11 years
51:26 and the past year and a half or so I
51:28 have been relatively obsessed with AI
51:32 and when chat GPT came out um what hit
51:36 me was the world has just changed but no
51:38 one knows it
51:40 yet um and then you know it went from
51:45 you know zero users to 100 million users
51:47 in 6 weeks so and that's organic I know
51:52 uh uh threads Zuckerberg's Twitter compe
51:55 edor just got to 100 million users in
51:57 like a week but you know he had a
52:00 billion people that were already
52:01 customers of Instagram and just said
52:03 push this button to invite all your
52:05 friends so that doesn't
52:07 count I'm still giving fastest fastest
52:10 adoption of consumer technology and
52:12 history to chat
52:15 GPT um so that's my background and and
52:18 um you know I've started 12 or 15
52:21 companies in my day and for
52:24 me the thing that I'm really passionate
52:26 about
52:28 is the the November 30th 2022 marks this
52:32 point in history where nothing will be
52:34 the same before that time to be in
52:37 machine learning and artificial
52:39 intelligence and all this stuff for the
52:40 most part you were a scientist and a
52:43 researcher and an
52:44 engineer um there were some developers
52:47 in there but for the most part you had
52:49 to be pretty geeky to be in that World
52:51 Cat GPT marks this moment where the rest
52:54 of us are given access to these
52:57 profoundly powerful tools these are
52:59 profoundly powerful they look really
53:02 simple it looks like a Google search
53:04 it's
53:05 not it's it's it's Alice and Wonderland
53:08 right it's it's you know falling down
53:11 the rabbit hole that's what these are
53:13 they're they're they're profoundly deep
53:15 and they're
53:16 profoundly
53:20 um
53:23 disruptive if you're a copywriter and
53:26 you use this you'll know but really if
53:30 you're any kind of knowledge worker if
53:31 you're in any kind of creative arts if
53:33 you look at the wga and the SAG strike
53:36 right now both of them have major issues
53:39 with AI because what they're realizing
53:42 is holy [ __ ] we're in
53:47 trouble my job is becoming obsolete
53:49 actually just laid off from IBM
53:51 interesting how can I find a new start
53:53 with AI well asking the question so
53:55 sorry you lost your job um not sure what
53:58 you do but what I would say is this the
54:00 the the the trick with this stuff is
54:02 just kind of
54:04 um start peeling the onion of what this
54:07 stuff is and what it makes possible so
54:09 go to any one of
54:10 these and start with what I call parlor
54:13 tricks have it write you a song have it
54:15 write you a poem have it make you a meal
54:16 plan have it explain some something
54:19 you're curious about to you it doesn't
54:20 matter just start playing with it have
54:23 it write you some code if you're at IBM
54:25 I don't know if you're on the technical
54:26 side there if you were have it write you
54:28 some
54:30 code then as you start to get your head
54:35 around how
54:37 incredibly flexible that is that it can
54:40 it can do some version of
54:44 anything then I would start to look at
54:46 what are you good
54:48 at technically and then what do you
54:51 enjoy what do you what do you want to
54:53 do and if those those two things are
54:56 pretty close together you're lucky if
54:58 those are far apart then maybe there's
55:00 something that Bridges those two that
55:02 you can do this with and I would start
55:05 saying learn as much as you can about
55:08 how to do what you used to do with these
55:11 tools so if you were a project
55:14 manager learn how you can supercharge
55:18 project management using these
55:21 tools um learn about tools like zappier
55:25 where you can create
55:27 automations um where you don't have to
55:29 code but you can create automations that
55:32 string a bunch of these things together
55:34 and do remarkable stuff um so so just
55:37 start learning like that so I think step
55:40 one is just start to get your head
55:41 around it don't don't feel too much like
55:44 you've got to use it in a certain way
55:47 nobody knows how to use this stuff right
55:49 now everyone's trying to figure it out
55:51 so everyone that you know all the
55:53 [ __ ] Tik toks you see out there that
55:55 that are like wow I got it figured out
55:57 you can do this for sales you can do
55:58 this all these everything's going to
55:59 change everything that's the way it
56:01 looks right now 6 months from now is
56:03 going to be
56:05 unrecognizable so so it doesn't really
56:07 matter what you do what matters is that
56:09 you get curious and you get adventurous
56:11 with with this and just start to
56:13 understand it so so phase one is just
56:15 get curious phase two is talk about it
56:20 put it on your LinkedIn profile I've got
56:22 a group called the AI salon that I
56:25 started for AI adventurers we call them
56:29 um people that are on the AI Adventure
56:31 so if you go to the
56:32 salon. there's a a a link tree there for
56:36 a Meetup Group we've got a Discord
56:38 server where you can make mid Journey
56:39 images and play and ask questions and
56:42 kick ideas around and meet people who do
56:44 what you do and how you do it and the
56:46 whole idea is it's it's a group it's
56:49 like a rag tag group of Misfits right
56:52 from all different businesses at at all
56:54 different levels of AI understanding
56:57 from you know people that have been in
56:59 machine learning for 15 years to people
57:03 that show up and they're like I heard AI
57:06 was going to be big what do I
57:08 do it's like and everything in
57:12 between um there's a fair amount of
57:15 people in here that are like they either
57:17 just got laid off or they're looking for
57:19 work almost all of them are putting the
57:23 fact that they're a member of this on
57:24 their LinkedIn
57:26 profile they're posting on LinkedIn
57:28 they're posting on Facebook they're
57:29 talking about it at dinner parties hey I
57:32 was playing with chat GPT and I did this
57:33 like even if people think you're [ __ ]
57:36 weird tell people what you're learning
57:38 tell people what this stuff
57:40 is because we're now starting to enter
57:44 this weird
57:46 phase where I I was expecting this to
57:49 happen and I think we're entering it
57:51 right
57:53 now where there's kind of an AI Panic
57:56 setting in and not a panic like the
57:57 robots are going to kill us but a panic
58:01 like holy [ __ ] I own an ad agency and
58:06 there's a 10p person AI Ad Agency that
58:09 just started that is telling my clients
58:13 um they'll do what I do for a third of
58:16 the price and my client is strongly
58:18 considering them what do I do so so
58:21 there's there's going to be a lot of
58:23 people realizing that this is disruptive
58:26 and they're going to be looking for
58:28 anyone who has any knowledge of this
58:30 stuff so if you can be one of those
58:31 people someone scrolling through
58:33 Linkedin and you talk about what you do
58:35 well I'm a project manager and here's a
58:37 cool thing I figured out with chat GPT
58:39 today someone's going to be scrolling
58:40 through Linkedin they're going to see
58:41 your post and go hey can we talk I'm
58:43 trying to figure this out can you help
58:45 me what do I do what do I do be the
58:49 person that they ask right and how you
58:51 do that is you just get yourself out
58:52 there so that would be my strong strong
58:54 recommendation and then you know I would
58:56 you
58:58 know take your network of people you
59:01 know
59:02 and start asking them what they're doing
59:05 with AI and and what are their plans
59:09 and you know maybe even volunteer to do
59:12 something for free for someone which if
59:14 you're looking for a job is not going to
59:16 it's going to feel
59:18 counterintuitive but maybe you learn how
59:20 to do something with chat gbt and you
59:21 say say Hey you know I'll come in and do
59:23 a half hour workshop for for team if you
59:25 want and and go there and you know maybe
59:27 that turns into a job we've had a number
59:29 of people from the salon that have
59:30 gotten jobs by doing that um so so just
59:35 dig in just get curious get adventurous
59:38 but sorry you lost your job that [ __ ]
59:40 sucks two * two question mark I assume
59:43 you're asking can chat GPT do math
59:47 sometimes it gives you the right answer
59:49 and sometimes it says
59:51 five um if you use code interpreter code
59:55 interpreter hallucinates a lot less
59:57 because it's actually writing python
59:59 code to do all of its calculations so
1:00:02 when it writes calculations to do math
1:00:05 it's actually doing math as opposed to
1:00:07 just the large language model part of it
1:00:09 where it's doing word prediction it'll
1:00:11 get that stuff wrong a lot more I
1:00:13 realized now that I was V vague I just
1:00:15 want to stay ahead H okay yeah um yeah
1:00:19 Amy like there's there's no there's no
1:00:22 way to stay ahead I all I would say is
1:00:23 just stay curious and and uh this
1:00:26 stuff's moving way too fast and it's way
1:00:28 too deep and Broad and crazy to keep up
1:00:32 with so just know that you're not going
1:00:33 to keep up with it or be ahead of it but
1:00:35 just stay curious and stay
1:00:37 engaged you're on Tik Tok silly I am why
1:00:41 am I silly what did I do did I make
1:00:43 something did I make a
1:00:47 silly I was like Charles Nelson Riley
1:00:49 for you gen xer on here all the young
1:00:51 people are like who I've never heard of
1:00:54 him
1:00:58 you've missed it for an aging Boomer in
1:01:00 dinosaur it where should we reinvent
1:01:04 ourselves well
1:01:05 mation I mean I I the the thing that I
1:01:08 just said about the the person from IBM
1:01:11 I
1:01:12 mean I think it's
1:01:16 um it's get curious I I mean if you want
1:01:19 to stay in
1:01:23 it I think
1:01:26 that AI operations so people who can
1:01:32 understand what tools are available what
1:01:35 the risks are what the security reality
1:01:38 is what the options are for training a
1:01:43 company's private data and keeping it
1:01:47 safe all of that stuff like like there's
1:01:49 there's a new flavor of it in town
1:01:52 that's going to be about how do you
1:01:53 deploy these things and then and then
1:01:55 even how do you deploy these things in a
1:01:57 way that
1:01:59 makes managing it more simple like
1:02:03 designing all those systems building
1:02:04 maintaining all those
1:02:06 systems I think is going to be in
1:02:09 tremendous demand so if you want to take
1:02:13 your dinosaur it skills and just say I
1:02:16 want to learn my dinosaur I want to I
1:02:18 want to redefine
1:02:20 it learning as much as I can about these
1:02:23 tools and how it applies to it I think
1:02:25 that's going to be in massive demand if
1:02:28 you're like dinosaur it like I'm kind of
1:02:31 [ __ ] done with it but what I always
1:02:33 wanted to do was kind of write poetry
1:02:36 books like like there's an opportunity
1:02:39 with these things to completely reinvent
1:02:42 yourself
1:02:45 so so I think it for me these are just
1:02:49 tools
1:02:50 right AI isn't going to solve [ __ ]
1:02:53 you've got to know what you want to to
1:02:56 solve and what you want to do and then
1:03:00 you can say okay how might these tools
1:03:03 help me do that faster better cheaper
1:03:06 you know
1:03:08 explosively a very real possibility
1:03:11 because these tools are so profoundly
1:03:13 powerful is
1:03:15 that you no longer have to put in 10,000
1:03:18 hours to be competent at something to be
1:03:21 like really masterful at
1:03:23 something
1:03:26 like probably somewhere between a
1:03:28 hundred and a thousand
1:03:30 hours using these tools as amplifiers
1:03:34 can get you competent at something that
1:03:36 you're currently shitty at including
1:03:38 something I I'll tell you a really good
1:03:41 example so in the AI
1:03:43 Salon we have this guy named Roger and
1:03:46 he was an IT guy and he's he's an older
1:03:48 fella and he's kagin he's a you know
1:03:50 he's a kudin he's just like anytime
1:03:52 anyone says something he's like here's
1:03:53 why I disagree with that right he's just
1:03:55 always like he's always got the
1:03:57 Counterpoint
1:03:58 um and his and and the original name of
1:04:01 the AI Salon was the AI artist salon and
1:04:05 Roger made his name in the Discord
1:04:07 server Roger not an
1:04:09 artist like as a [ __ ] you to the all you
1:04:12 artists he's like I'm an IT guy right
1:04:15 and so we we put a a mid Journey
1:04:18 instance we put a mid-journey bot inside
1:04:20 the AI Salon um
1:04:23 Discord and Roger took his it brain and
1:04:28 his logical thinking and his you know
1:04:30 the way he sort of breaks [ __ ] down and
1:04:32 deconstructs things and applies The
1:04:35 Logical
1:04:36 system he applied that to Mid journey
1:04:40 and he started making some of the most
1:04:42 beautiful
1:04:44 art of of anyone in in the in the
1:04:47 Discord and just making a lot of it
1:04:49 because he was taking this systematic
1:04:50 approach to it um and he's now changed
1:04:53 his name to I think it's Roger becoming
1:04:55 an artist or something like that but but
1:04:57 like he he he has completely discovered
1:05:00 in himself a joy for art
1:05:04 and a talent for interacting with these
1:05:09 schools these tools with a skill that he
1:05:12 had that looked a certain way when he
1:05:14 was in in it that are manifesting the
1:05:18 completely different way because of
1:05:19 something like mid Journey so it
1:05:22 so the future
1:05:25 might not even look like what you
1:05:26 intended to so again that's why I say go
1:05:28 back to if you had your
1:05:31 DRS what would you do even if you're not
1:05:34 good at something right now and then
1:05:37 once you understand kind of the
1:05:38 direction what's your North Star then
1:05:41 you can look at these tools and say okay
1:05:43 I kind of I kind of want to do music are
1:05:45 there any good music tools go to
1:05:46 futurepedia
1:05:48 doio and just spend a weekend geeking
1:05:51 out on the music tools and you'll find
1:05:53 one or two in there that will do
1:05:55 something that will give you your Kevin
1:05:56 mallister
1:06:00 moment and then you'll be Off to the
1:06:02 Races right then you'll be
1:06:05 hooked and then
1:06:07 again start talking about it start
1:06:09 letting people know you're doing it hey
1:06:12 I used to be a dinosaur in it but you
1:06:14 know what I'm doing now I'm writing
1:06:15 songs with AI I heard AI is evil I hear
1:06:17 it steals from people it's a ripoff and
1:06:20 no let me tell you what I
1:06:22 learned you don't like it you don't have
1:06:24 to use it I'm using it I love
1:06:26 it ADHD here ADHD in the House mation 2
1:06:32 awesome listen these tools are these
1:06:35 ADHD is a [ __ ] superpower with this
1:06:38 stuff right now when it when it all gets
1:06:39 normal and boring we'll we'll be like
1:06:42 yeah yeah those are just tools now but
1:06:44 right now when we're we're inventing
1:06:46 this stuff woo ADHD is a super power
1:06:50 super freak recruiter out LOL
1:06:55 um prompt developers are a thing yes
1:06:59 they call them prompt Engineers I don't
1:07:00 like prompt engineering I heard the
1:07:02 other day on I I read on Twitter someone
1:07:05 said it's more like prompt psychologist
1:07:07 because it's a large language model so
1:07:09 you're actually
1:07:10 using your native language the English
1:07:13 language in my case as a programming
1:07:15 language and
1:07:17 and but the English languages got all
1:07:20 this Nuance to it and so how you write
1:07:22 and craft prompts there's an engineer
1:07:24 component to it but there's also very
1:07:26 much a like I find myself using please
1:07:28 and thank you a lot when I'm talking to
1:07:30 the robot which seems completely
1:07:33 counterintuitive but it especially if
1:07:35 you play with pi uh the the pi um app
1:07:40 from uh
1:07:42 inflection it talks back to you and
1:07:44 they've Pro programmed it so it's super
1:07:47 friendly it's like it's like gee Kyle
1:07:50 you know that's really that idea you
1:07:52 have is really interesting let's dig
1:07:54 into that and then it'll talk and do a
1:07:56 bunch of stuff and so so these
1:07:59 tools
1:08:02 act with empathetic cues they don't have
1:08:05 empathy but they act with empathetic
1:08:07 cues and it kind of reminds you to
1:08:09 interact with them in that way and when
1:08:11 you do they get better it's [ __ ]
1:08:13 weird it's weird um but yes people that
1:08:16 are good at prompting right now
1:08:18 absolutely have jobs I think the first
1:08:20 prompt engineer job that sort of made
1:08:22 all made a buzz on Twitter was like 350
1:08:25 Grand a year um I don't think that's
1:08:27 going to be there for long I don't think
1:08:29 prompt engineering stays a prominent
1:08:32 thing for long because these tools are
1:08:33 going to get increasingly sophisticated
1:08:35 where they're going to do a lot of the
1:08:36 prompting for us behind the scenes now
1:08:38 if you're on the development side if
1:08:40 you're developing applications and
1:08:41 you're you're you know on the back end
1:08:43 doing API stuff with this you need to be
1:08:46 really good at prompting and really
1:08:48 understand the nuances of it uh just as
1:08:50 an end user I think right now it's
1:08:52 valuable to learn how these things work
1:08:56 and to get good at getting [ __ ] from
1:08:58 them right like learning these tools
1:09:00 right now is is kind of the the
1:09:03 equivalent of understanding basic
1:09:06 mathematics before using a
1:09:08 calculator so you at least understand
1:09:10 what it is and how the calculator does
1:09:12 what it does and then you can go okay
1:09:14 now I understand why I would use this
1:09:16 tool and where I might not same thing
1:09:19 here but yes prompt development is a big
1:09:22 deal you are so right I'm using AI
1:09:25 little by little and I learn something
1:09:27 new every day having fun yeah Laura and
1:09:29 and you know um like again share share
1:09:32 what you're learning even if people
1:09:33 think you're weird it doesn't [ __ ]
1:09:35 matter because in
1:09:37 Sharing in sharing it with the world a
1:09:39 couple of things happen some of the
1:09:41 people that you think wouldn't have been
1:09:42 interested in it all of a sudden are
1:09:44 like Could you teach me it's like a
1:09:46 sneaky little thing off on the side um
1:09:50 and then the other thing that'll happen
1:09:51 is you know people that are not
1:09:53 interested will just sort of drift away
1:09:54 and that's
1:09:55 fine but people will self- select into
1:09:59 your circle that are curious about this
1:10:02 stuff so where I am right now I want to
1:10:07 be surrounded by people that are trying
1:10:09 to figure this [ __ ] out I don't care if
1:10:11 they love it or hate it but I want them
1:10:13 in the mode of I'm going to [ __ ]
1:10:15 figure this out I'm going to figure out
1:10:17 how this affects my job and my
1:10:19 livelihood and who I am and what makes
1:10:21 me happy like I want to hang out with
1:10:23 those people that's what this Channel's
1:10:25 about um so yeah it it it it literally
1:10:29 doesn't matter what you do do little by
1:10:31 little just
1:10:34 just well in
1:10:36 fact these lives that I'm doing I I'm
1:10:39 spending 2 to three hours a night every
1:10:42 night for like the past month doing
1:10:44 these
1:10:46 Lives why well because I used to come
1:10:51 home from work and I would spend 2 or 3
1:10:53 hours on the the Sim racing rig which is
1:10:55 right over
1:10:56 there
1:10:58 driving and I [ __ ] love it like I
1:11:01 love it to the point that I have SIM
1:11:03 racing heel I've got like btis in my
1:11:05 heel from stepping on the accelerator so
1:11:09 hard
1:11:11 um and what hit me was I need to
1:11:16 be more into this I need to be learning
1:11:20 more I need to be sharing what I'm
1:11:22 learning more because I want to be
1:11:25 surrounded by people that are not like
1:11:28 oh well AI kind of lies and it's
1:11:31 stealing from people and so I'll just
1:11:32 wait for 3 years and I think it's evil
1:11:35 and the robots are going to kill us any
1:11:37 I have no interest in those people
1:11:39 none if you use these tools and then you
1:11:41 feel like the robots are going to kill
1:11:43 us great go do something about it go
1:11:46 make an anti-robot killing humans tool
1:11:48 go work for
1:11:50 anthropic go work for Elon musk's new
1:11:53 xai he promises to make a Kinder gentler
1:11:56 open AI Okay
1:12:00 cool so so I think it's hugely
1:12:05 important all right Tik Tok just jumped
1:12:08 me to the bottom of the of the comments
1:12:10 again any of you who have been here
1:12:11 before you know that's my pet peeve
1:12:13 could someone please call Tik Tok and
1:12:15 tell them to fix that stupid [ __ ]
1:12:22 bug I don't know who I'm talking to by
1:12:25 the way it's just looks like we got some
1:12:28 new people here hey everybody my name is
1:12:29 Kyle Shannon this is the AI learning lab
1:12:31 um these URLs right here are there used
1:12:34 to only be three at the top now there's
1:12:36 five at the top these are all options if
1:12:39 you've not played with chat GPT these
1:12:41 are all different flavors of chat GPT or
1:12:44 you know some competitor of them I would
1:12:47 personally start with chat. open.com
1:12:50 just go there and start playing and and
1:12:52 if you're brand new to this this [ __ ] if
1:12:54 you're like I don't even know why I'm on
1:12:55 this channel the guy was kind of
1:12:56 screaming in the phone so I stopped if
1:12:59 that's you right now fine
1:13:01 cool go to chat.
1:13:04 open.com and then in another tab open up
1:13:07 prompt.
1:13:08 chat prompt. chat will teach you what
1:13:11 these things are and how to talk to them
1:13:14 and then it's a giant long list of
1:13:18 um personas that you can tell chat GPT
1:13:21 to act like a historian or act like a ma
1:13:23 mathematician or act like a screenwriter
1:13:26 or act like a marketing social media
1:13:28 marketing manager or whatever it might
1:13:30 be and it's it's really inspiring it'll
1:13:32 it'll teach
1:13:35 you what the [ __ ] everyone's been
1:13:37 talking about and you owe it to yourself
1:13:41 to use these tools long enough until you
1:13:44 have what I call a Kevin McAllister
1:13:46 moment Home
1:13:49 Alone um so just go play there all right
1:13:53 let me try to to find where I was in my
1:13:59 comments oh I think I found my
1:14:04 spot how do I feel about the AI actors
1:14:07 um
1:14:08 inevitable you know the the the studio
1:14:11 like one of the one of the reasons sag
1:14:13 went on strike is that the the [ __ ]
1:14:16 Studios [ __ ] idiots like I I just
1:14:19 people are so [ __ ] stupid so so the
1:14:22 studio is basically just ignored ignored
1:14:26 the the actor's guild for like 3 weeks
1:14:28 they knew the writer's Guild was on
1:14:30 strike and you know they knew there was
1:14:32 a possibility there was going to be some
1:14:34 solidarity there so they ignored them
1:14:36 for 2 or 3 weeks and then they come out
1:14:38 with this proposal that they're like
1:14:40 we've got this ground groundbreaking
1:14:42 proposal what if for background actors
1:14:45 so the people in a film that don't have
1:14:47 any lines the ones that are standing
1:14:48 behind Matt
1:14:50 Damon what if we pay them for one day
1:14:53 and and then we scan them
1:14:55 in and we take their license and and we
1:14:58 just own that forever and we can use
1:15:00 them in any movie in the future and we
1:15:01 never have to pay them again wouldn't
1:15:03 that be
1:15:04 awesome and sag was like go [ __ ]
1:15:06 yourself we're going on strike tomorrow
1:15:11 so um what do I feel about it I have a
1:15:14 degree in acting I moved to New York
1:15:16 City out of school I've written seven
1:15:17 screenplay so this is fairly close to to
1:15:22 a lot of what I know and a lot of a lot
1:15:24 of what I
1:15:26 feel and I'm also like a tech Optimist
1:15:30 and a
1:15:31 tech
1:15:43 fan it's going to be incredibly
1:15:47 disruptive incredibly
1:15:49 disruptive how do I feel about it I feel
1:15:52 conflicted because I feel
1:15:57 like as a
1:16:03 consumer 5 years from now I'm going to
1:16:06 be able to
1:16:07 say I want to watch Raiders of the Lost
1:16:10 Arc but
1:16:13 with
1:16:15 um Matt Damon and the lead
1:16:20 role and I'll be able to watch that so
1:16:23 as a consumer there's going to be we're
1:16:26 we're going to we're we're entering an
1:16:28 age of hyper
1:16:30 personalization but as an
1:16:33 actor what do you do about that do
1:16:37 you is there a whole new kind of agency
1:16:40 that all it does is manage likenesses
1:16:44 and license out like
1:16:47 likenesses
1:16:50 um as a business owner if if I want to
1:16:55 produce an ad that's got a person
1:16:57 smiling drinking a cup of coffee or if I
1:17:00 want to produce a commercial where
1:17:02 someone's drinking a cup of
1:17:05 coffee if I don't have to send a film
1:17:07 crew down to a cafe pay the cafe for the
1:17:11 you know 8 Hour full day Revenue that
1:17:14 they would have
1:17:15 lost for the 38 seconds of video that I
1:17:19 might or might might not use if I can
1:17:21 just generate that and use that in my
1:17:23 commercial
1:17:25 I'm going to do that right as a business
1:17:28 owner
1:17:30 um does that suck balls for anyone in
1:17:34 the acting business and in the writing
1:17:37 business and in the voiceover business
1:17:39 and in the cinematography business and
1:17:41 the the gaffs and the lighting people
1:17:44 and all the people that would have been
1:17:45 paid for that 8h hour day yeah sucks
1:17:47 balls for them
1:17:52 um there are going to be actors writers
1:17:57 Film Production people that Embrace this
1:18:00 stuff that are going to get masterful at
1:18:03 it and they're going to produce stuff
1:18:05 that is going to blow our [ __ ] minds
1:18:08 so there's a part of me that's
1:18:14 like like I there's a say in Creative
1:18:17 will always out and what it basically
1:18:20 means is good creative is good creative
1:18:24 and and really creative people when they
1:18:26 get their hands on these tools and
1:18:27 really get their heads around them
1:18:29 they're going to do stuff that's going
1:18:30 to melt our [ __ ] faces that I'm
1:18:32 really excited about but the collateral
1:18:34 damage of the transition from where we
1:18:37 are to where we're going to be five
1:18:40 years from now is just going to be
1:18:41 devastating and it's not just actors and
1:18:43 writers and it's it's every knowledge
1:18:46 worker think about a business
1:18:49 consultant what what do you what does a
1:18:51 business consultant do they come in
1:18:54 you tell them your problems they go away
1:18:56 for a month they do a bunch of research
1:18:58 they talk to a bunch of smart people
1:19:01 they come back a month later and they
1:19:02 say well based on your problem here are
1:19:05 six other companies that have faced what
1:19:07 you faced here's how they've managed
1:19:08 through
1:19:12 that if I'm even halfway decent at
1:19:16 prompting I can put that same
1:19:18 information into a large language model
1:19:20 and effectively get the same answer back
1:19:24 in 15 minutes that I would have paid
1:19:28 $50,000 and waited a month to get back
1:19:31 from the consultant what's that do to
1:19:32 that consultant I don't
1:19:35 know maybe that consultant starts using
1:19:38 these tools and can deliver more better
1:19:40 [ __ ] cheaper I don't know I don't
1:19:44 know so I'm really conflicted about it
1:19:47 you know there there's a there's a
1:19:49 there's a like the the the depth of how
1:19:52 profound this stuff is how profound this
1:19:55 stuff
1:19:57 is is incredibly exciting but in the
1:20:01 depth of how profound it is is profound
1:20:04 disruption and that's [ __ ] terrifying
1:20:06 and so I've said this before on this
1:20:08 channel I am simultaneously as excited
1:20:11 as I've ever been in my life about a
1:20:13 technological shift and as terrified and
1:20:17 I'm not terrified because of the robots
1:20:19 are going to kill us I'm more terrified
1:20:21 that even if it goes brilliant
1:20:24 and 5 6 8 10 years from now we're we're
1:20:28 a much more prosperous world and a lot
1:20:31 of our problems have been solved like
1:20:33 even if we go to the the brightest
1:20:35 possible place the next 2 to three years
1:20:38 are going to [ __ ]
1:20:40 suck and it's just how it is I am funny
1:20:44 and entertaining thank you very much
1:20:46 although I just got off a ranty soap box
1:20:48 I don't think that was entertaining but
1:20:50 um but I don't know this is this is the
1:20:53 this is big stuff it's deep and
1:20:55 profound all
1:20:58 right um Bard needs my image to code
1:21:02 napkins B rules oh reads my image to
1:21:04 code napkins yeah I know retr Punk
1:21:06 that's really cool so so what retr
1:21:09 Punk's talking about is he took so in in
1:21:12 the
1:21:13 when open AI announced gp4 they said
1:21:17 it's multimodal and what that means is
1:21:19 you can upload an image and it will deal
1:21:21 with it it'll understand what it is and
1:21:24 the example that they used was he took a
1:21:27 really bad napkin sketch of a website
1:21:30 that was like a joke like push a button
1:21:32 to get a
1:21:33 joke and he uploaded that into a thing
1:21:36 and said make me this website and it did
1:21:38 it made a functional website um from a
1:21:41 napkin sketch and so retr Punk uh said
1:21:44 he did that with Bard today and bing.com
1:21:47 also reads images now so Bing and Bard
1:21:52 experiment upload an image and ask it to
1:21:54 write you some advertising copy for that
1:21:56 image and see what it does Boston you in
1:22:00 Boston Deborah that's super cool my the
1:22:04 uh my my co-host of the AI Salon is is
1:22:07 in Boston she's a professional
1:22:09 photographer named Leah
1:22:11 faston cheers from Scottdale Arizona
1:22:14 Cheers Cheers Cheers welcome welcome
1:22:16 welcome it's good for teachers to be
1:22:18 able to type in a GPT write me five
1:22:20 analogies that explain the subject
1:22:22 exactly all sort of stuff that that like
1:22:25 you can take teaching from these broad
1:22:27 Concepts to you know something that's
1:22:30 specific to each kid and let them all
1:22:33 sort of find their own personalized I
1:22:36 the thing that I'm like I'm where I'm
1:22:39 where I'm increasingly starting to
1:22:41 understand these tools
1:22:43 is we move into an increasingly
1:22:46 personalized world to the point where it
1:22:50 gets hyper personalized kind of five to
1:22:53 10 years out where all of our
1:22:56 entertainment
1:22:58 is original music that is just like
1:23:02 exactly like the music we love and it
1:23:04 just gets better and better and better
1:23:06 and there might be some music in there
1:23:08 that's music that existed before but
1:23:11 there's going to be a bunch of it in
1:23:12 there that that was written for US based
1:23:15 on all the other [ __ ] that we
1:23:17 like and it's going and like on every
1:23:19 level of what we do so like that's where
1:23:21 we're headed and so
1:23:23 you know I'm starting to see like little
1:23:25 it's like little little pieces of
1:23:26 glitter starting to fall from the from
1:23:28 the clouds and I'm like oh there's a
1:23:31 glimpse of what it might oh there's
1:23:32 another one but it's I don't know it's
1:23:35 it's I I I don't know where this stuff's
1:23:37 going I got no [ __ ] clue as as we
1:23:40 joked about this week some guy came in
1:23:42 when I was just starting a live and you
1:23:45 know people was like hey Kyle hey how's
1:23:46 it going and he was like what are your
1:23:48 qualifications it's just like very
1:23:50 matter of fact like who the [ __ ] are you
1:23:52 dude and it threw me I was like I don't
1:23:55 know what my qualifications are my
1:23:57 qualifications are I'm curious about
1:23:59 this [ __ ] and I'm trying to figure it
1:24:01 out that's it I mean I got you know 58
1:24:04 years on the planet and I've done some
1:24:07 [ __ ] but nobody's qualified right now
1:24:10 nobody knows what this is going to look
1:24:11 like nobody knows the impact these tools
1:24:14 are going to have on everything nobody
1:24:16 does so if you do let me know because
1:24:20 I'd like to
1:24:21 know
1:24:23 if indeed you're a time traveler that
1:24:25 used gpt1 17 and used it to build the
1:24:28 time machine to come back here to just
1:24:30 hang out and see how stupid we were
1:24:31 about this stuff just you know put it in
1:24:33 the comments let me know ask them about
1:24:35 things you're an expert in and their
1:24:37 shortcomings are more obvious yeah
1:24:39 that's a that's a really good point user
1:24:42 that so so one of the
1:24:45 superpowers of using these tools
1:24:49 is the more quickly you can sus out
1:24:54 what's good or bad what's [ __ ] and
1:24:56 lies or truth the better you get at it
1:25:01 so if you're older if you're like a gen
1:25:04 xer who actually knows how to use
1:25:05 computers and you've been on the planet
1:25:07 long enough to be able to very quickly
1:25:09 look at something from chat GPT and and
1:25:11 and you're like well that's just
1:25:13 horrible or oh that's [ __ ] that's
1:25:16 good but also as user said start using
1:25:20 these tools where you've got deep domain
1:25:23 expertise where you're where you're good
1:25:25 at something you'll be able to more
1:25:26 quickly figure out where it's good where
1:25:29 it's not good and where it's lying and
1:25:30 where it's not it's a it's a great Point
1:25:33 validation tools how do we find original
1:25:35 sources are we at risk of losing the
1:25:36 library well this becomes the ultimate
1:25:38 library right like ultimately the the
1:25:41 base model for these large language
1:25:42 models is the entire output of humanity
1:25:46 that's where we're headed right that's
1:25:48 like the ultimate expression of these is
1:25:51 all of what we've ever done is in them
1:25:54 or can be right can be processed in them
1:25:58 so so this becomes the ultimate Library
1:26:02 um the validation tools so if you use
1:26:06 bing.com
1:26:08 um Bing is actually connected to the
1:26:11 internet and and Microsoft has done a
1:26:13 pretty good job of incorporating
1:26:15 citations into when it gives you answers
1:26:18 it will actually search for you know
1:26:21 articles on the internet and and include
1:26:23 links to to them within your answers um
1:26:27 I think there's going to be an entire
1:26:29 industry built around validation and you
1:26:33 know DEH hallucinating the content and
1:26:36 making it more factual um we're early
1:26:38 right now so just understand these tools
1:26:41 are spit and duct tape right now the
1:26:43 underlying Tech is is fairly
1:26:45 sophisticated and it's built on Decades
1:26:47 of you know
1:26:49 Evolution but the application layer of
1:26:52 chat GPT and of Bing and of claw. and of
1:26:56 b.google.r
1:27:12 foundly
1:27:13 sophisticated what they actually do how
1:27:16 they check it are not so much thank you
1:27:18 for the subscription I appreciate that
1:27:20 that's awesome when AI starts teaching
1:27:23 itself how will we know the information
1:27:25 it taught itself is
1:27:29 accurate that that could end up be that
1:27:32 that could end up being a risk um
1:27:35 there's kind of two schools of thought
1:27:37 it's it's not so
1:27:38 much is the information accurate or not
1:27:42 because we're going to be able to with
1:27:44 validation tools we're going to be able
1:27:45 to figure out what's what's accurate or
1:27:47 not one of the one of the issues that
1:27:51 that is being talked about about in
1:27:53 machine learning circles is that if you
1:27:55 train a large language model on output
1:27:57 of large language models there's a thing
1:27:59 that can happen called Model collapse
1:28:01 where the quality of the the model is
1:28:03 getting better and better and better and
1:28:05 then just all of a sudden it falls off a
1:28:06 cliff and is bad right so model collapse
1:28:09 and then I saw a counterargument to that
1:28:11 that said basically model model collapse
1:28:14 is not a thing and you know there's a
1:28:16 way you can do it where it just keeps
1:28:17 getting better so I don't know
1:28:21 um
1:28:25 I the industry is very well aware that
1:28:28 how large language models work make make
1:28:30 it susceptible and prone to
1:28:33 hallucination um so I think you're just
1:28:35 going to have companies that build an
1:28:36 applic so so so you've got this large
1:28:39 language model based down here you've
1:28:40 got your application layer where where
1:28:42 you type into chat GPT you know give me
1:28:44 an answer and then it digs into the
1:28:47 large language model and pulls out the
1:28:49 answer and it spits it into your
1:28:51 face
1:28:54 well why not build a layer in between
1:28:58 those two that's a validation layer so I
1:29:02 want this answer to this question you
1:29:04 know dump it down into the large
1:29:06 language model it spits out an answer
1:29:07 the validation layer goes is there any
1:29:09 [ __ ] in here oh there's six pieces
1:29:12 of [ __ ] why don't you go back down
1:29:13 and try again right or why don't I go
1:29:16 out as a validation layer and I'm going
1:29:18 to go out to the internet I'm going to
1:29:19 find real sources for this and I'm going
1:29:21 to replace all the bu [ __ ] sources
1:29:23 with real sources then I'm going to have
1:29:25 the large language model rewrite the um
1:29:29 thing that it wrote and update any uh
1:29:32 counterfactual information and then
1:29:35 we'll give the person the
1:29:37 answer there's your business to start
1:29:40 all week this week it's been I've been
1:29:42 giving away billion dooll business ideas
1:29:46 you can make
1:29:48 money there's one build a validation
1:29:51 layer
1:29:53 from a from a conceptual standpoint it's
1:29:56 actually a very simple thing take take
1:29:59 the output of a large language model
1:30:02 look at it and determine if anything in
1:30:05 there is [ __ ] and if it is fix it
1:30:09 now technically probably a little more
1:30:13 complicated but that could be a plugin
1:30:15 for chat GPT right the plugins are
1:30:18 really easy to build here so you could
1:30:20 build a plugin that's like no bull buit
1:30:23 the no [ __ ] plugin and you install
1:30:26 the no [ __ ] plugin and then you just
1:30:28 interact with chat GPT and every time it
1:30:30 gives you an answer longer than two
1:30:33 sentences it checks it says any [ __ ]
1:30:37 in here and it tells you we found three
1:30:40 pieces of [ __ ] here's what they were
1:30:42 here's what we replaced them with go
1:30:44 check it for yourself but we're pretty
1:30:46 sure we fix
1:30:48 this um yeah so and by the way AI is
1:30:51 teaching it itself right now the whole
1:30:53 open source Community is using gp4 to
1:30:56 train all the lower level uh open source
1:30:59 large language models um
1:31:02 so we'll find out if model collapse is
1:31:05 is a very real risk pretty
1:31:10 soon very helpful thank you I asked chat
1:31:13 if it had read the Library of Congress
1:31:15 and it could not confirm or
1:31:18 deny were you just good at finding
1:31:21 Talent ENT to write this code
1:31:26 what was I just good at writing at
1:31:29 finding talent to write all this code I
1:31:31 haven't written any of this code I
1:31:33 haven't written any of this this is all
1:31:35 I'm just showing [ __ ] that these people
1:31:36 have written I my first Epiphany about
1:31:40 chat GPT was I had it write me an
1:31:42 application in Python and I've never
1:31:45 written a line of python in my life and
1:31:47 in 90 minutes I created a fully
1:31:49 functional python application that asked
1:31:52 someone for a topic asked them three
1:31:54 questions about it and then generated
1:31:56 five pieces of social media content that
1:31:58 was the first application I wrote with
1:32:00 chat PT never written a line of python
1:32:03 in my life still haven't that's when
1:32:06 that's when it hit me holy [ __ ] the
1:32:08 world just changed but no one knows
1:32:11 it that was my big Epiphany I'm using
1:32:15 pi. a for the first time as you speak
1:32:17 and I've made huge inroads tonight in 6
1:32:20 minutes oh that's awesome yeah Silver
1:32:22 Fox I was I was I I saw a Tik Tok and it
1:32:26 was this these two guys and they're
1:32:27 talking about pi. a and I'm like oh I
1:32:30 know pi. a I've used that and and then
1:32:33 they showed the um the the iOS version
1:32:36 of it that you can talk to it and it
1:32:37 talks back and I'm like oh that could be
1:32:39 fun and so I just as I was running
1:32:41 errands today I was talking to my Ai and
1:32:45 it was it was crazy good it you know and
1:32:48 it it still has a little bit of a a
1:32:50 delay but it's not bad
1:32:53 it's pretty close and it's connected to
1:32:54 the internet so you can ask it for
1:32:56 weather and current events and [ __ ] like
1:32:58 that
1:33:00 um yeah it's this is really interesting
1:33:02 and Pi should probably be on this list
1:33:05 so at some point I'm going to need to
1:33:07 figure out a different maybe I'll have
1:33:09 multiple lists maybe this is like a a
1:33:11 rotating oh that's what it could be this
1:33:14 could be a uh this could be just a
1:33:16 rotating series of screens right so
1:33:19 depending on when when you log in you
1:33:20 get a different tool I I know I'll
1:33:22 figure something out for
1:33:24 that I love that you're doing these
1:33:26 lives I appreciate that mation
1:33:28 appreciate that a lot it's uh it's good
1:33:31 to know sometimes I feel like I'm
1:33:32 shouting into a void because I do do a
1:33:34 lot of shouting because I'm an idiot
1:33:36 thanks Silver
1:33:39 Fox
1:33:41 ah oh Slinky Slinky It's a Wonderful
1:33:46 toy um only 15% have used chat gbt 1%
1:33:50 have returned not exactly a option oh
1:33:52 Dakota Digital Cowboy can you confirm on
1:33:56 the copywriter part what do you mean can
1:33:58 I confirm on the copywriter part I'm not
1:34:00 sure what you're talking about but if
1:34:01 you clarify it I'll I'll address it um
1:34:04 chat would not tell
1:34:06 me what it had been trained on oh yeah
1:34:10 it it's so it doesn't necessarily know
1:34:14 so
1:34:18 again understanding how the tool works
1:34:21 is
1:34:26 these tools have the
1:34:29 cognitive capacity of a
1:34:33 calculator like if you ask the
1:34:34 calculator how do you do your
1:34:36 calculations it doesn't [ __ ]
1:34:39 know these tools are a a word calculator
1:34:45 on
1:34:47 steroids the challenge with them is
1:34:49 they're so sophisticated that what they
1:34:51 write seems like they understand [ __ ]
1:34:54 they
1:34:55 don't like they don't have any capacity
1:34:58 to understand what they've written
1:34:59 they're just they're just a probability
1:35:01 engine putting one word after another
1:35:04 really really
1:35:06 fast but they don't actually understand
1:35:08 what they wrote so they also don't have
1:35:11 Consciousness about what they were
1:35:13 trained on now if you say
1:35:19 um if you say rather than have you read
1:35:22 the Library of Congress if you say what
1:35:25 are the what are the primary categories
1:35:28 of content in the library of con
1:35:30 Congress it probably knows that and then
1:35:33 you can probably start to run down some
1:35:34 incredibly deep you know tunnels through
1:35:38 Li library of Content Library of
1:35:40 Congress content um but you know again
1:35:44 it may hallucinate like you know it it
1:35:47 may or may not be right but but it it
1:35:49 does not it right now it doesn't have
1:35:51 any kind of self-a awareness it doesn't
1:35:52 have memory it doesn't have like all of
1:35:55 its memory is just within your chat
1:35:57 within the context window of your
1:35:59 current chat that's its only memory
1:36:01 right now future versions Sam mman said
1:36:04 this about chat
1:36:05 GPT
1:36:07 um I'm pretty sure Claude has said it uh
1:36:10 anthropic has said it about Claude these
1:36:12 things are going to start to have you'll
1:36:13 have a profile where it understands who
1:36:15 you are it will start to have memory
1:36:17 these things are going to start to get
1:36:19 way more sophisticated right now they're
1:36:21 just very simp simple calculators fancy
1:36:24 ones profoundly fancy ones teachers will
1:36:28 lose jobs to AI I
1:36:32 think so so again I go back to this
1:36:35 quote from this guy Paul rater from the
1:36:37 marketing AI Institute AI will not
1:36:39 replace your job but people using AI
1:36:43 will I think teachers in particular are
1:36:46 going to be forced to reinvent how they
1:36:48 teach
1:36:50 because
1:36:54 the structure of work is going to
1:36:57 change what you train to is going to
1:36:59 change so in today's world there's a lot
1:37:03 of specialization you know I'm going to
1:37:04 go to school for advertising copywriting
1:37:07 I'm going to go learn to be a copywriter
1:37:09 for an advertising agency it's this very
1:37:11 narrow tactical thing there are
1:37:14 exceptions like liberal arts like one of
1:37:16 the things that I'm feeling like this AI
1:37:19 does is this is like revenge of the
1:37:20 liberal arts majors so having a
1:37:22 horizontal view of the world is a is a
1:37:25 superpower when you have tools that do
1:37:27 all the Tactical Tech tactical specialty
1:37:30 stuff so I think teachers are going to
1:37:33 have to shift from that idea of you've
1:37:36 got to learn these tactical
1:37:38 specialization things and you know in
1:37:40 lower school we're going to teach you
1:37:42 them as broadly as possible but but it's
1:37:44 all about learning the sort of tactical
1:37:46 execution to shifting to critical
1:37:48 thinking and and how do we drive
1:37:51 critical thinking thinking and how do we
1:37:52 drive horizontal thinking because that's
1:37:55 what's going to be required in a in a
1:37:57 future world where tactical execution is
1:38:00 handled by the
1:38:01 machines um so I so I don't know that
1:38:06 teachers are going to lose their jobs to
1:38:07 AI but I think that
1:38:10 teachers deeply understanding what a AI
1:38:13 makes possible and that having to be
1:38:15 part of the curriculum is just by its
1:38:18 nature going to force some teachers out
1:38:19 of the
1:38:20 profession that that don't want to make
1:38:22 that transition because it's it's a
1:38:24 completely different
1:38:26 mindset
1:38:28 um yeah take your Skillet ad chat GPT
1:38:32 become a superhero oh skill set not your
1:38:36 skillet that's pretty
1:38:37 funny yeah so that that's a really good
1:38:40 point uh Dakota Digital Cowboy
1:38:42 that yeah these These are in my opinion
1:38:46 these tools are Humanity amplifiers so
1:38:48 take the thing you're good at or the
1:38:49 thing you're passionate about
1:38:52 do it in conjunction with this and it's
1:38:54 just like you just like level up this is
1:38:57 like level up machines these are these
1:38:59 are like you know power pills in a video
1:39:03 game that's what these things are for
1:39:05 your skills for your dreams for what you
1:39:08 want to execute welcome everybody my
1:39:11 name is Kyle Shannon this is the AI
1:39:12 learning lab these if you're new here uh
1:39:15 these URLs chat.
1:39:17 open.com is the official chat GPT
1:39:20 website all these other things are
1:39:22 either competitors of or different ways
1:39:24 that you can use chat GPT so if you
1:39:26 haven't used chat GPT before the the
1:39:29 prerequisite to being part of my lives
1:39:31 is go [ __ ] use these
1:39:33 tools just go play you don't have to be
1:39:36 a machine learning expert you don't have
1:39:38 to be a scientist I have a degree in
1:39:40 acting that's why I talk with my hands
1:39:42 that in the
1:39:43 ADD but just go there and start playing
1:39:46 and go to prompt. chat that little thing
1:39:48 right there prompt. chat that'll teach
1:39:50 you how to interact with these things if
1:39:52 you've never done it and just while you
1:39:54 listen to me ramble in the background
1:39:55 just go there um it looks like we got a
1:39:57 decent chunk of people in here if you
1:39:59 would just uh tap the like screen so we
1:40:01 get some likes so people can come in and
1:40:03 if you're willing to share the live
1:40:05 maybe we can get a another pile of pile
1:40:07 of folks in here and oh thank you very
1:40:09 much look at some rain it's raining
1:40:12 money look I might end up twerking oh
1:40:15 yeah I'm an
1:40:19 influencer I need to get the purple
1:40:21 strip my hair and learn to twerk but
1:40:23 after that then I'll be an influencer
1:40:26 um what was I talking about uh uh thanks
1:40:30 for the hearts
1:40:32 uh I don't know what I was talking about
1:40:35 uh if if one of the things you're trying
1:40:37 to figure out how to do is you can make
1:40:39 money with this is not that but we do
1:40:43 make fun of those people here and that's
1:40:44 part of the
1:40:45 deal I have no clue why Silver Fox was
1:40:49 in my last comment I I don't know Ridge
1:40:52 Hardy Winston Ridge
1:40:54 Hardy maybe you
1:40:56 do I don't know cuz Tik tok's [ __ ]
1:40:58 weird Man 6 months not that soon I don't
1:41:01 know what I was talking about but that's
1:41:03 fine yeah my predictions are I I figured
1:41:06 we would get into an AI what I'm calling
1:41:09 an AI panic in about 6 months back in
1:41:11 January and I don't think we're in it
1:41:13 yet but I think we're starting to see it
1:41:15 I think all these lawsuits are starting
1:41:17 to be it you know the Sarah Silverman
1:41:19 lawsuit the SAG and wga strikes having
1:41:23 AI at the center what they do the um
1:41:28 chat GPT pulling um browse with Bing
1:41:31 because it was stealing content from
1:41:33 Fortune Magazine I you know I I just it
1:41:38 it seems like there's starting to be
1:41:39 little chinks in the armor of
1:41:41 traditional business
1:41:43 models starting to be threatened so
1:41:46 we're starting to see increases in
1:41:47 lawsuits and things like that so it it
1:41:50 might be another six months but I have a
1:41:52 feeling it's it's coming pretty
1:41:54 soon um L why they're not I don't know
1:41:57 what you're talking about I work in the
1:41:59 AI field blue fathom I don't know what
1:42:01 tell me what I said you said LOL Y no
1:42:03 they're not I work in the AI
1:42:05 field I probably said something that was
1:42:08 [ __ ] what you drinking iced coffee
1:42:13 M see there's a little ring of coffee
1:42:16 grounds down
1:42:19 there and water it's Saturday I know
1:42:21 it's boring I just don't I just didn't
1:42:23 I'm not in the mood for for the
1:42:29 booze and I'm you know ampy enough that
1:42:34 I don't
1:42:35 need
1:42:37 lubrication all right this time around
1:42:41 my reply to Silver Fox was on purpose oh
1:42:43 that's funny all right uh I'm from the
1:42:46 future
1:42:47 [Laughter]
1:42:50 aipi uh remember Moran Mindy I do I used
1:42:53 to live in Boulder Colorado which is
1:42:55 where Moran Mindy uh was was housed was
1:42:58 uh was housed no was set um the Moran
1:43:01 Mindy house is still there it's still
1:43:04 got a rainbow flag on the
1:43:06 outside Robin Williams was one of my
1:43:09 heroes I'm still devastated by his
1:43:13 loss uh Moran meny was brilliant that
1:43:16 was a brilliant show it was I mean that
1:43:18 was just a show that was just like you
1:43:20 know you just realized this guy was a
1:43:22 [ __ ] Tor to force of talent just
1:43:26 transformed like like no one else could
1:43:28 have done that show no one else could
1:43:30 have done that show
1:43:32 period um I've been a C++ and python
1:43:35 programmer for years and I love fooling
1:43:36 with these AI tools and gaining skills
1:43:38 Mad Skills yeah exactly it's like it's
1:43:42 like anything like like I've written a
1:43:44 bunch of screenplays and I'm doing
1:43:45 things like like you know what's funny
1:43:48 is
1:43:49 like like anything you do you kind of
1:43:51 have these these vague ideas sometimes
1:43:55 and then you kind of playact in your
1:43:57 head what would it take to turn that
1:43:59 vague idea into a real idea and most of
1:44:01 them just die on the vine cuz you're
1:44:03 like oh that would require work and I
1:44:04 don't feel like it right now so that
1:44:06 idea just goes it just Withers away and
1:44:08 you know blows off in the wind but with
1:44:10 these AI things it's like oh I have this
1:44:12 vague Whisper of an idea let me what if
1:44:15 I put that in chat
1:44:18 gbt oh it came to life here it is
1:44:22 like you can do it with programming you
1:44:23 can do it with writing you can do it
1:44:24 with [ __ ] anything it's amazing or or
1:44:26 even just weird things like you like
1:44:29 someone will mention something on the
1:44:30 news and you're it's like it's like some
1:44:32 historical event that you've heard of 20
1:44:35 times you're like I don't actually know
1:44:36 what happened there and you just type it
1:44:38 in what happened at that event it's like
1:44:40 here's the answer and then you're like
1:44:42 oh that's amazing well how did that lead
1:44:44 to this well here's how those two things
1:44:46 are connected you're like ah it's just
1:44:48 like [ __ ] instant knowledge I was on
1:44:51 the
1:44:53 phone yester no yeah
1:44:57 yesterday yeah I was on the phone
1:44:59 yesterday okay so I don't know if you do
1:45:01 this as a as a laminated card carrying
1:45:04 member of the ADHD Club I don't have the
1:45:06 attention span to read much so I don't
1:45:08 read a lot but I'm a fast synthesizer so
1:45:12 like like one of the things I do with
1:45:13 books is once I understand the plot of
1:45:16 it I have no interest in reading it
1:45:18 because it's like why would I read the
1:45:20 rest of this I know what's going down
1:45:22 right it's just the way my brain
1:45:23 works so I was on a call and um this guy
1:45:27 goes have you read whatever it was by
1:45:29 Neil Stevenson and I said no I haven't
1:45:32 and he and he starts talking about all
1:45:33 this stuff that sounded really cool like
1:45:34 what we were talking about so as he's
1:45:36 talking I just went to chat GPT and I
1:45:38 typed in I think it was something
1:45:39 diamond diamond or something diamond
1:45:41 from Neil Stevens and I just TP what's
1:45:44 the plot of D and you know make it short
1:45:46 and concise with short bullet points and
1:45:48 it just tossed it up and like I
1:45:50 instantly got what the book was about
1:45:52 what the concepts were about and could
1:45:54 actually understand what he was telling
1:45:56 me in a deeper and more profound way and
1:45:59 I I told him I said I said I just had
1:46:01 chat gvt read that book for me thank you
1:46:05 um so you know if you got one of those
1:46:09 kind of brains this shit's [ __ ]
1:46:11 genius someone uses some fancy ass word
1:46:14 in a conversation and normally you'd be
1:46:15 like uh-huh and just pretend to
1:46:18 understand and then go I don't have a
1:46:19 [ __ ] what they talked about you go to
1:46:21 your little
1:46:22 dictionary you don't have to do that
1:46:24 anymore it's amazing all right I wrote a
1:46:26 business plan for my 15-year-old son
1:46:28 last night awesome with the Drone that I
1:46:30 got him for his
1:46:32 birthday you wrote the business plan
1:46:34 with the Drone or for the Drone he being
1:46:37 honest uh can chat gbt help with my
1:46:40 acting career um
1:46:45 interesting I think chat GPT you know
1:46:47 one of the things chat GPT could do for
1:46:49 you for your acting career is it could
1:46:52 probably very quickly find
1:46:56 um monologues so so like find a
1:46:58 monologue that you
1:47:00 like and then um tell it to go find
1:47:05 similar speeches or monologues in plays
1:47:07 and movies that are similar to the one
1:47:09 that you like right for characters of a
1:47:12 similar age so so could it could help
1:47:15 there
1:47:17 um I mean you know one of one of my
1:47:21 philosophies with my acting career when
1:47:23 I was in New York was like I don't want
1:47:25 to wait around to work so I started a
1:47:27 theater company and we did all original
1:47:30 plays right so we we got together with
1:47:32 playwrights directors um producer stage
1:47:36 manager types and actors and we just did
1:47:40 plays and did plays and did plays so you
1:47:43 could either help have it help you with
1:47:45 writing or with directing or with Stage
1:47:49 management
1:47:51 um those could be value if you want to
1:47:52 get um busy there I think with things
1:47:56 like um Bing being connected to the
1:47:59 internet you might be able to find more
1:48:03 opportunities
1:48:05 um I think voiceover actors are going to
1:48:08 be in
1:48:10 trouble I think actor actors even though
1:48:12 they can synthesize actor actors I think
1:48:14 we're a few years out from from that
1:48:19 being
1:48:21 like as easy as the voiceover stuff is
1:48:24 right now voice voiceover actors
1:48:27 probably have another year before that
1:48:28 industry is just fully
1:48:30 disrupted and then actor actors is
1:48:33 probably more like 3 to 5
1:48:35 years um trying to think what
1:48:42 else you could probably have chat gbd
1:48:44 tell you what are all the different
1:48:46 kinds of acting opportunities that are
1:48:48 out there like from corporate to
1:48:50 commercials to plays
1:48:54 to I don't know um you know Live
1:48:58 Events I was like one of my first jobs
1:49:01 in New York you had to be an actor and I
1:49:03 got a job selling yo-yos at FAO Schwarz
1:49:05 the big toy
1:49:06 store so there's [ __ ] like that so it
1:49:09 could maybe help you find stuff like
1:49:11 that I don't
1:49:14 know have you signed up for air. a I
1:49:18 don't I I don't even know what it is so
1:49:20 no I haven't signed up but let's go find
1:49:22 out a
1:49:28 IAI uh uh uh uh uh uh uh 100,000 sales
1:49:32 and customer service reps at the top of
1:49:33 a button oh I saw this
1:49:38 today first ever AI that can have a
1:49:41 fullon 10 to 40 minute long phone call
1:49:43 that sounds like a real human with
1:49:45 infinite memory perfect recall and can
1:49:48 automatically take actions across five
1:49:50 5,000 plus applications this is what I
1:49:52 was talking about yesterday on the live
1:49:55 someone needs to invent this and then I
1:49:56 saw it today so I'm glad you brought it
1:49:58 up
1:50:00 um the the phone tree you know the phone
1:50:03 tree that we've all come to accept as
1:50:07 acceptable you know thank you for
1:50:10 calling
1:50:11 AT&T
1:50:13 say Spanish I don't speak Spanish um
1:50:18 press one for the thing you're not
1:50:20 interested in Press Two for the other
1:50:22 thing you're not interested in press
1:50:24 three for a thing you're really not
1:50:26 interested in if you're thinking about
1:50:28 this press four the thing you're
1:50:30 interested in go [ __ ] yourself you know
1:50:32 that thing that goes away that we flush
1:50:36 that entire model down the [ __ ]
1:50:39 toilet and replace it with this you call
1:50:42 it hey this is AT&T welcome my name is
1:50:46 Sally how are you I'm good Sally how are
1:50:48 you doing I'm doing just great what can
1:50:50 I help help you with today and it's one
1:50:51 of those
1:50:53 things it's a virtual thing that knows
1:50:56 all of the rules it knows all of the
1:50:58 content it can it has authority to do
1:51:02 all of the [ __ ] it needs to do up to
1:51:04 whatever level of authority is
1:51:06 appropriate and then it's also smart
1:51:08 enough to understand that if your
1:51:10 problem is complicated enough that an AI
1:51:14 can't solve it it immediately routes you
1:51:16 to a
1:51:18 person that's [ __ ] genius so yes I
1:51:21 will go sign up for that join the wait
1:51:23 list thank you for
1:51:26 that this is this is one of
1:51:29 those products I get excited about are
1:51:32 you know there's certain things in life
1:51:33 that just piss me off so much oh look at
1:51:35 that congrats you're officially a part
1:51:38 of History oh well well
1:51:43 well all right it's about verbage a
1:51:46 phone answering service using AI oh AI
1:51:50 wait AI a pii is
1:51:56 that wait a API are you part of air. a
1:52:01 if you are that's a cool service get me
1:52:03 into the get me off the weit list sooner
1:52:06 than later if you're a part of that
1:52:07 company because I'd love to play with it
1:52:11 my ears perked up at AI psychologist
1:52:13 yeah Dr J I think
1:52:15 um I like the term prompt crafting
1:52:18 because I think it is a creative
1:52:20 it's a creative Endeavor but it does
1:52:23 have engineering elements about it and
1:52:24 then when I heard that idea of of prompt
1:52:27 psychology that that felt right to me
1:52:30 because it's it's a large language model
1:52:32 we
1:52:33 we're we're tapping into the the core um
1:52:39 tool we have as Humanity to communicate
1:52:41 with each other and we're using that
1:52:44 tool to put into this
1:52:47 tool that knows all the languages that
1:52:50 knows knows all the words that knows all
1:52:51 the
1:52:55 meaning um so yeah so I think it is kind
1:52:58 of AI psychology AI
1:53:01 psychologist someone had chat GPT played
1:53:05 Detroit become human what is that a
1:53:09 movie reference I don't know what that
1:53:11 means someone had chat gbt played
1:53:13 Detroit become human clarify what you
1:53:17 mean and I'll
1:53:19 I'll
1:53:21 respond to that in some way I've never
1:53:23 used Discord can I join the salon if yes
1:53:27 how do I get an invite okay
1:53:30 so if you
1:53:33 go to
1:53:36 here go to the salon.
1:53:39 a it's a link tree so the first link is
1:53:43 the Meetup and so the Meetup is not
1:53:46 Discord so the Meetup is just um it'll
1:53:48 notify you when we have meetings our
1:53:50 next meeting is next
1:53:52 Tuesday um we have a good speaker coming
1:53:54 in we meet every other week we usually
1:53:56 have speakers or we we'll sometimes do
1:53:59 this thing called AI confessionals where
1:54:00 everyone in the meeting has 10 minutes
1:54:03 to talk about something they've done
1:54:04 with AI those are
1:54:06 fun and then we have a Discord I'll I'll
1:54:09 show you a little bit about Discord the
1:54:11 there's a there's a link to the Discord
1:54:13 invite um on the link tree and then
1:54:15 there's also a um a LinkedIn group that
1:54:18 you can join as well um let me flip over
1:54:29 here so so I'm in the I'm in the the AI
1:54:33 Salon Discord right now so just to show
1:54:36 you what's going on here so when you
1:54:39 when you when you click on the invite
1:54:41 for Discord it'll have you probably
1:54:43 create an account and then it'll have
1:54:45 you jump through a couple of Hoops like
1:54:47 you know click on this thing to agree to
1:54:50 our terms or whatever I don't know what
1:54:51 it does like it it's some sort of [ __ ]
1:54:54 that basically says don't be an [ __ ]
1:54:56 um and then when you log into Discord
1:54:58 there's kind of three columns there's
1:55:00 there's this column here these are
1:55:02 called servers so every one of these
1:55:03 circles is like a different community so
1:55:06 the AI Salon you know that's our little
1:55:09 icon there and then this next column is
1:55:12 these are the conversations within the
1:55:14 AI salon so like here's a general
1:55:16 discussion here's a ask for help oh
1:55:19 retro oh so this is cool so retr Punk
1:55:22 who's who's in the live or at least was
1:55:23 before this is the um thing that he he
1:55:27 uploaded this drawing to Google and it
1:55:29 made him a
1:55:30 website that's super
1:55:33 cool look see built him a website look
1:55:36 at that
1:55:37 look that's super cool from that um so
1:55:41 so each one of these little discussions
1:55:43 is is kind of this this you know uh
1:55:47 linear discussion of just people talking
1:55:50 about stuff and there's all different
1:55:51 categories there's a mid Journey channel
1:55:54 so you can come into the mid Journey
1:55:55 Channel and you can make
1:55:57 images and if you've never if you don't
1:55:59 know what Discord
1:56:00 is then you probably don't know how to
1:56:03 do a mid Journey image so I'm going to
1:56:06 show you a cool thing right now so we're
1:56:08 going to go out of Discord for a
1:56:10 second and we're going to go over to a
1:56:13 thing called po.com which is on my list
1:56:16 of things you should go check out if you
1:56:18 want to play with cool [ __ ]
1:56:21 and so when you get to po.com you've got
1:56:24 a bunch of
1:56:26 different chat gbt kind of models here
1:56:29 you've got Sage which is theirs you've
1:56:31 got Claude 2 and Claude instant so this
1:56:34 is like GPT 3.5 that's like GPT 4 with
1:56:38 the 100,000 token context window you got
1:56:41 gp4 you got Cloud instant so so these
1:56:42 are all just like chat GPT tools but po
1:56:45 also has Bots and one of the Bots that
1:56:47 they have so if you go explore up here
1:56:51 you go to all these Bots and they have
1:56:54 different Bots and if I go to AI the
1:56:56 very top bot here is called mid
1:56:59 journey and so if I click on the mid
1:57:01 Journey bot I can now type in a crappy
1:57:06 image prompt so I'm going to say what am
1:57:09 I going to do tonight let's say
1:57:13 um um
1:57:15 blue berry
1:57:18 muffin in
1:57:21 a
1:57:24 hipster coffee
1:57:26 shop so I put in just that thank you for
1:57:30 the gift and then it rewrote that color
1:57:32 photo of a blueberry muffin in a hipster
1:57:35 coffee shop rustic and cozy at Ambiance
1:57:39 wooden table chairs vintage Decor
1:57:43 natural lighting Canon EOS 5D markv
1:57:46 camera Kodak ector 100 film 50 m 50 mm
1:57:51 lens shallow depth of field da da with
1:57:53 the names Jamie Oliver AA Garten Martha
1:57:56 Stewart so it rewrote a prompt for me
1:57:58 that's really good so that's that's a
1:57:59 cool thing go use the uh the Bots at po
1:58:02 are a lot of fun and that one is one
1:58:04 there so now we're over at Mid Journey
1:58:08 or we're at the the AI Salon Discord in
1:58:10 the mid Journey Channel and so to make
1:58:13 an image in mid Journey you're in
1:58:15 Discord you type slash
1:58:18 imagine
1:58:20 and then it gives you a space for a
1:58:23 prompt and rather than writing it I'm
1:58:25 just going to paste in what po just gave
1:58:27 me and I hit return and now it's going
1:58:30 to go make me that
1:58:31 image failed to process your command
1:58:34 apparently it's not going to go make me
1:58:35 that image wait did it
1:58:37 just oh now I think it did waiting to
1:58:40 start okay it did now um so now it's
1:58:43 going to go make me that image and I'll
1:58:45 show you I'll show you what it does so
1:58:47 so that's how you use Discord well
1:58:50 that's the beginnings of it there's a
1:58:51 bunch of [ __ ] there
1:58:52 it's the fact that you have to use
1:58:55 Discord to make mid Journey images and
1:58:57 use Discord at all it's if you've ever
1:58:59 used slack it's sort of the precursor to
1:59:01 slack slack is sort of Discord for
1:59:04 business
1:59:06 um and so here come our blueberry
1:59:08 muffins and you'll notice if you've
1:59:09 never seen images be generated uh in AI
1:59:14 before they start they always started
1:59:15 these as these blurry kind of noisy
1:59:17 images and then getting increasingly
1:59:19 Sharp sharper it's the nature of how
1:59:21 they're produced they start as kind of a
1:59:24 a noisy averaged image of like thousands
1:59:27 of images and then it just denoises it
1:59:29 over 30 or 50 steps and turns it
1:59:32 intoing there's our blueberry
1:59:35 muffins so those are pretty slick so
1:59:37 let's take one of these that we
1:59:40 like uh we're going to
1:59:44 take I think that one's nice but this
1:59:48 one has blueberries on the plate so
1:59:50 let's take four so once you've created
1:59:52 these you got upscale 1 2 3 four so
1:59:55 you're going to make them bigger or
1:59:56 variations often what I'll do is I'll
1:59:59 I'll hit the variation first so I'm
2:00:00 going to get four different variations
2:00:02 of this bottom right one so I'm going to
2:00:04 hit
2:00:08 V4 welcome everybody doing a little
2:00:11 little mid Journey
2:00:18 play
2:00:24 bow all right and here come our
2:00:26 variations one looks dramatic that upper
2:00:29 left one looks like it's going to be
2:00:30 dramatic that's a little drama it's high
2:00:32 drama in the coffee
2:00:34 shop that's a lovely muffin you've got
2:00:37 there sir is that
2:00:39 blueberry I love a good blueberry
2:00:43 muffin you're a muffin you're a muffin
2:00:48 stop
2:00:50 come on muffin you can do
2:00:56 it all right so those are pretty
2:01:00 cool I'm going to
2:01:03 grab one or three one let's grab one so
2:01:07 I'm going to hit upscale one so now it's
2:01:10 going to make a bigger version of that
2:01:11 and then I'm going to show you a cool
2:01:14 thing you can you can now pan you can
2:01:18 Pan the image so so it's it made this
2:01:21 image that's vertical right so the
2:01:22 muffin feels like it's in a little jail
2:01:25 it's like in this visual jail so so now
2:01:27 we've got this lovely lovely photo of a
2:01:31 muffin if if we had a I mean here's a
2:01:34 really good example someone asking me
2:01:36 about actors before and photographers
2:01:38 and the impact of this stuff like if I
2:01:40 had a coffee shop I could hire a
2:01:43 photographer to come in and photo style
2:01:44 a muffin but if that looks close enough
2:01:46 to what my muffins look like there it is
2:01:49 I'm done okay so now I can zoom out two
2:01:53 times or 1.5 times or or custom but I've
2:01:57 also got these little panning arrows so
2:01:59 I'm going to pan to the
2:02:04 right
2:02:06 and wait failed to process your command
2:02:09 now now there it went okay um so now
2:02:12 it's basically going to add to the to
2:02:14 the right of the image um so it's going
2:02:17 to give me more of the coffee shop and
2:02:19 it'll probably make it more like a
2:02:20 square image than a tall image and then
2:02:23 I can zoom out from there and things
2:02:24 like that but it's really really
2:02:32 powerful so the left half of this is the
2:02:35 image that was there before the right
2:02:37 half is the things it's
2:02:41 adding so this one added a plant which
2:02:44 is kind of cool oh little flowers on the
2:02:45 table that's kind of
2:02:47 cool this one looks like it's adding a
2:02:49 fan F your little light up
2:03:05 here I don't know what I just did that
2:03:08 that wasn't quite beatboxing that was
2:03:11 like that was sort of the beatboxing
2:03:13 equivalent of a dad
2:03:14 [Laughter]
2:03:18 joke
2:03:26 all right so now look so so the aspect
2:03:28 ratio of our image changed here which is
2:03:30 super
2:03:32 cool and
2:03:34 so yeah let's just grab this third one
2:03:37 here we'll upscale the third
2:03:44 one boom there you have it there's a
2:03:48 lovely
2:03:50 you know depth of field beautiful
2:03:52 photograph here if I wanted to do like
2:03:54 an Instagram ad about my Cafe I've now
2:03:57 got plenty of room over here to put text
2:03:59 and a title you know the N
2:04:03 Cafe you know Heaven on a plate whatever
2:04:07 right and what did that take us 4
2:04:12 minutes probably longer than
2:04:16 that so hopefully that gives you some
2:04:19 sense of what these things are all
2:04:22 about they're just crazy it's crazy you
2:04:25 put in some words and out comes a muffin
2:04:28 he made a muffin out of nothing a muffin
2:04:31 out of nothing have you seen this this
2:04:33 this this IIA stuff it's incredible it's
2:04:37 it's incredible it's he he put in words
2:04:39 and out came a muffin and there was a
2:04:41 steaming cup of warm milk next to it I
2:04:44 don't know why there wasn't coffee there
2:04:46 it was for a
2:04:47 cafe but it was a muffin it looked tasty
2:04:50 do you like a blueberry muffin I love
2:04:52 blueberry muffins and these were plump
2:04:54 blueberries I don't he didn't say
2:04:56 anything about plump but they were
2:04:59 plump I don't know I don't know what's
2:05:02 going to happen you just say words and
2:05:05 out comes a muffin I think it's like
2:05:07 it's like them uh it's like that Star
2:05:09 Trek
2:05:11 show they say things and then it comes
2:05:18 out
2:05:21 H you can make money with
2:05:26 ch don't ask me for acting advice by the
2:05:30 [Laughter]
2:05:32 way like driving a car I don't know what
2:05:34 that means I accidentally got a work
2:05:37 from job with AI just joking around yeah
2:05:42 uh what's your name uh Kile oh that's
2:05:45 cool Kyle Kile Kile vibes
2:05:49 um a bunch of people in the AI Salon
2:05:52 have gotten jobs just by putting on
2:05:54 LinkedIn that they're doing [ __ ] with AI
2:05:55 have gotten jobs just because of that
2:05:57 we're we're we're entering kind of a
2:05:59 silly season where if you're remotely
2:06:01 confident with this stuff and that's why
2:06:03 I'm
2:06:05 saying if if you start playing with
2:06:07 these tools with any regularity and you
2:06:09 start understanding them and getting
2:06:10 your head around them and having your
2:06:12 aha moments start talking about it
2:06:14 because you never know who you're
2:06:15 talking to everyone right now is trying
2:06:17 to figure this out and one of those
2:06:19 people in your life that you had no idea
2:06:21 is going to go I need someone who knows
2:06:23 this can you come work for me or can you
2:06:24 do this job for me I'll pay
2:06:27 you good points can you upload half a
2:06:30 book and just ask it to make the book
2:06:32 better
2:06:33 yes now will it make it better I don't
2:06:36 know I think you probably have
2:06:39 to put some like what like what I would
2:06:42 do in that in that case is I would do a
2:06:46 lot of pre- prompting so rather than
2:06:47 trying to write the ultimate kind of
2:06:49 prompt like here's a book make it better
2:06:53 make it better how right so what what I
2:06:56 would probably do is I would say um you
2:06:59 know I want you to act as an experienced
2:07:02 book editor who specializes
2:07:05 in you know
2:07:09 improving this kind of
2:07:11 fiction
2:07:13 um without changing the intent of the
2:07:15 author start out with some sort of broad
2:07:17 statement defining what it is
2:07:20 and then maybe even describe the outline
2:07:23 of the book or the the main ideas in the
2:07:25 book and then um and then maybe what you
2:07:29 do is you pop in the first half of the
2:07:31 book and again you you'll probably have
2:07:33 to do this at claw. because it's got a
2:07:36 100,000 um token context window which is
2:07:39 about 75,000 words so if you're talking
2:07:41 half a book you're well within that so I
2:07:43 would sort of set it up Define it as as
2:07:46 this master editor and then I would pop
2:07:50 in um the first half of the book and
2:07:53 then maybe even say and here's the
2:07:55 outline for the second half of the
2:07:57 book and then I would say you
2:08:01 know you know please take a look at this
2:08:04 book and and without changing
2:08:08 anything I want you just to tell me if
2:08:10 you understand what your job
2:08:12 is and then it will say yes and it
2:08:16 hopefully won't just start rewriting
2:08:17 everything and then you can say I want
2:08:20 you to look at the first
2:08:22 chapter and I don't want you to rewrite
2:08:25 anything yet but I want you to give me a
2:08:27 bulleted list
2:08:29 of um General improvements that you'd
2:08:33 make to the first
2:08:34 chapter and then it'll give you those
2:08:37 and see if you like what it came up with
2:08:39 and then you can say okay I like one
2:08:41 three and four but I don't agree with
2:08:43 number two so can you give me a new list
2:08:46 of improvements where you don't change
2:08:48 anything of number two and it'll give
2:08:49 you that and then you say okay now can
2:08:52 you rewrite the first chapter to
2:08:54 incorporate those changes and then you
2:08:56 go read it and see how it did and then I
2:08:57 would just work my way down through the
2:08:59 book so um so yeah but it absolutely can
2:09:03 and the other thing to do with these
2:09:05 things is don't expect that it's going
2:09:08 to get like my experience with this is
2:09:10 it tends to be really good at reading
2:09:12 something and responding to it so I
2:09:14 think that exercise is it'll probably be
2:09:16 pretty
2:09:17 good but I consistent ly find that as
2:09:20 I've gotten better with these tools I
2:09:22 can get to about 80 between 80 and 90%
2:09:26 of where I want to be very very
2:09:28 quickly and then there's kind of a point
2:09:30 of diminishing returns where it just
2:09:32 doesn't get it and then I basically just
2:09:34 take it from there and rewrite it so I I
2:09:36 would think it's going to be a similar
2:09:37 exercise with your
2:09:39 book is you know if it gives you a
2:09:42 shitty chapter with bad edits then tell
2:09:44 it hey that was shitty you you you
2:09:47 omitted this part of the plot which is a
2:09:49 setup for something down the line and
2:09:52 give it feedback and tell it it was
2:09:54 shitty say you know you didn't you
2:09:56 didn't match the tone when you changed
2:09:57 this thing or you made it too generic oh
2:10:00 I'm sorry here's another version let me
2:10:02 try
2:10:04 again um I'm here for it oh Tik Tock
2:10:07 just jumped me to the bottom dang you
2:10:09 Tik
2:10:12 Tock what' you do tonight I don't know I
2:10:14 watch some old guy try to navigate Tik
2:10:16 Tok why would you do that I don't
2:10:19 know I don't know he was talking to some
2:10:22 actor about how he could use chat PT to
2:10:24 get an acting job didn't seem realistic
2:10:26 to me thank you Clint copper appreciate
2:10:28 it good evening free indeed good evening
2:10:32 to you I'm scrolling back up the list of
2:10:35 questions to see where I was by the way
2:10:37 if you're new here I try to read all of
2:10:39 the questions and answer them all
2:10:41 because I'm a psychopath um so what that
2:10:44 means is if you've put a question in in
2:10:46 the last hour I probably have not gotten
2:10:48 to it yet and you're probably like this
2:10:49 isn't even
2:10:51 alive he's probably a bot this is
2:10:54 probably something from China where
2:10:56 they're trying to get our information
2:10:57 they're scanning our retinas you know
2:11:01 man I'm
2:11:02 gp5 very nice to meet
2:11:05 you hello my name's
2:11:09 Bob I'm the new manifestation of AI
2:11:12 That's going to destroy your life thank
2:11:17 you
2:11:20 [Music]
2:11:22 oh there's the person that got a a job
2:11:24 accidentally by just joking around about
2:11:26 AI that's [ __ ] awesome it's funny I
2:11:29 actually when I started agency.com back
2:11:31 in the olden timey days in in the mid
2:11:34 90s uh there's a guy who teaches I think
2:11:36 he teaches at NYU now or Columbia named
2:11:38 Klay shery he's written a bunch of books
2:11:40 but Clay did this website that was it
2:11:43 was called biffs something website and
2:11:45 it was purposefully a horribly designed
2:11:48 designed website it had blink tags and
2:11:50 all different sized fonts and different
2:11:52 rainbow colors it was hilarious and I
2:11:54 ended up hiring him cuz he WR he made a
2:11:56 joke website I sent him an email and I
2:11:58 said anyone that can design a site that
2:12:01 bad needs to come work for us and so he
2:12:04 came and worked for us for a while it
2:12:06 was pretty
2:12:07 fun uh I'm here for it where do I start
2:12:10 so you
2:12:12 start with all of these URLs right here
2:12:14 you just pick one of them you don't you
2:12:16 don't need to use all of them the reason
2:12:18 they're all here is
2:12:20 that some of these are new competitors
2:12:22 like Claude and Bard um have showed up
2:12:26 on my list um in the past two days but
2:12:29 chat. open.com is the official chat GPT
2:12:32 website so start there just go there and
2:12:35 start playing ask it to write you a song
2:12:37 a poem a recipe um ask it to explain a
2:12:41 concept to you ask it to write a
2:12:43 children's book just anything ask it
2:12:45 anything and then go to prompt. chat
2:12:48 this little thing down here it's a
2:12:50 document that'll teach you about what
2:12:53 these things actually are and and what
2:12:55 prompting is prompting is basically just
2:12:57 talk to the talk to the thing and ask it
2:13:00 for [ __ ] ask it to do stuff for you and
2:13:02 it will and so this gives you a lot of
2:13:05 ideas about things that you can ask it
2:13:06 and things ways you can ask it to
2:13:09 behave um and then and then the other
2:13:11 things are all just different variations
2:13:13 of it so you can go play with all of
2:13:15 them but they're all sort of variations
2:13:17 on the theme that is chat.
2:13:19 open.com I need to use it for my real
2:13:21 estate company please then go ahead and
2:13:24 do it Kyle you're the man thank you
2:13:26 Troy um PV sunshine I'm not sure what
2:13:28 you mean ask me a specific question I'll
2:13:30 do what I can to answer it um I'll be
2:13:33 right on buddy I'll be sure to talk to
2:13:36 Tik Tock that's I assume that's when I
2:13:38 was talking about it jumping me to the
2:13:40 bottom it's such a stupid [ __ ] bug
2:13:43 it's
2:13:44 like so the bug is if I if I have my
2:13:47 finger on the screen and I'm scrolling
2:13:49 the text up and someone posts it jumps
2:13:51 me to the
2:13:53 bottom it's like just make it not do
2:13:56 that like there's there's a there's an
2:13:58 if statement that is on the wrong side
2:14:00 of when my finger is touching the
2:14:02 screen just [ __ ] fix your order of
2:14:06 operations for your scrolling people
2:14:09 come on got a degree in acting I could
2:14:12 fix this but the problem is it's not a
2:14:15 bug that I'm sure no one uses Tik Tok
2:14:18 like I'm using it so it's not a bug
2:14:19 anyone would give a [ __ ] about but
2:14:23 me I can find bugs very fast Kyle AI
2:14:26 evangelist Shannon thanks I'm drinking
2:14:28 the Kool-Aid thanks Mark that's [ __ ]
2:14:31 hilarious that's awesome I know enough
2:14:34 Pro wait I know enough programming that
2:14:37 I can catch a lot of wrong information
2:14:39 but I can usually edit my something yeah
2:14:42 I mean that's the thing the more the
2:14:45 more you use these tools for [ __ ] that
2:14:47 you're good at the better you're going
2:14:49 to be at them cuz you'll be able to go
2:14:51 that's brilliant that's a piece of [ __ ]
2:14:53 that's just
2:14:56 wrong
2:14:58 um just
2:15:00 like it's the thing about actors isn't
2:15:03 it you stick a [ __ ] heart over their
2:15:05 head and they just start dancing for
2:15:08 you I just want to be
2:15:13 loved we need to get AI to figure out
2:15:15 how to fix the scrolling issue we
2:15:17 [ __ ] do could could you all get on
2:15:19 that could you all go to one of these
2:15:21 things and come up with a fix describe
2:15:25 the problem come up with the fix and
2:15:28 then we'll send the code to Tik Tock say
2:15:30 could you please
2:15:32 just do it do a get you know a get a get
2:15:36 on this get this in the [ __ ] code
2:15:41 base I was saved with an AI and saved I
2:15:44 like that Dr J that's solid edit my
2:15:46 prompts and get code I want yeah that's
2:15:50 that's the thing it the the don't give
2:15:53 up on these tools when they give you
2:15:56 shitty results initially for for two
2:15:59 reasons one is these things are the
2:16:02 worst they're ever going to be this
2:16:04 shit's going to get better fast like
2:16:06 dramatically better
2:16:09 fast um but also back to that idea of
2:16:14 prompt being a prompt
2:16:16 psychologist one once you kind of make
2:16:19 the transition from Google prompting
2:16:22 which is like I want as concise a prompt
2:16:25 as possible cuz I don't want to confuse
2:16:27 the search engine this is kind of the
2:16:29 opposite where you got to kind of think
2:16:32 psychologically like what context do I
2:16:35 need to give this idiot
2:16:37 savant so it can take all of its vast
2:16:40 knowledge and become my
2:16:45 superpower once you make that shift
2:16:48 everything starts to change then you
2:16:49 realize oh I can problem solve if I get
2:16:53 back a CRA crappy answer I can problem
2:16:55 solve the prompting to the point that I
2:16:58 can get back a consistently good answer
2:17:00 and once you get that holy [ __ ] is that
2:17:05 gratifying I just want to make a million
2:17:07 dollars using chat gbt you can make
2:17:09 money with chat
2:17:12 gbt we got to do the t-shirts people I
2:17:15 I'll get on it I just haven't done it
2:17:17 yet I'm still think about what I want it
2:17:19 to look like but we'll do it and I don't
2:17:21 know how to spell
2:17:22 [Music]
2:17:26 J so it's all math right yeah currently
2:17:29 the way these tools work it's all
2:17:31 math I I mean what's what's astounding
2:17:34 about these tools and mid Journey like
2:17:36 the the image generation tools are using
2:17:38 the same [ __ ] technology what's
2:17:40 astounding is that it does what it
2:17:44 does because the output is so
2:17:47 sophisticated
2:17:50 but the but what it's actually doing
2:17:52 once these things are embedded is they
2:17:54 are vector they are vector sets they're
2:17:57 sets of numbers that point to a location
2:18:01 in Vector space like thousand Dimension
2:18:04 Vector space where there are semantic
2:18:06 clouds of words that have similar
2:18:08 meaning or related meaning and then when
2:18:11 you type your
2:18:13 prompt it does some sort of math and
2:18:16 says Where in the Vector space is most
2:18:20 likely to be the next word and it goes
2:18:22 oh that's in Vector Cloud
2:18:26 768 you know over here where there's all
2:18:29 the word dog with all these different
2:18:31 meanings and this particular use of the
2:18:33 word dog is the one that's most likely
2:18:35 to be the next word so it's all math
2:18:37 it's astounding has no concept of what
2:18:40 it's doing it's it's just a calculator
2:18:43 it's just the most [ __ ] remarkable
2:18:46 calculator ever in the history of
2:18:48 humankind that's going to transform the
2:18:51 very definition of work but other than
2:18:53 that it's probably just a passing fad
2:18:56 right probably just a passing
2:19:02 fat this is it's one of my favorite
2:19:05 moments is when someone's like well I
2:19:08 heard it's it's just
2:19:10 overhyped and then you show them
2:19:12 something that's relevant to
2:19:15 them and you see them go
2:19:20 I had no idea I was on a call with a guy
2:19:24 once like this like three months ago and
2:19:27 I had just built I built this thing
2:19:29 called the social media marketing
2:19:30 machine which is I I basically built a
2:19:32 zapier automation which was the which
2:19:35 which matched the the first python
2:19:37 application that I wrote in chat
2:19:39 GPT and it was it was a Google form that
2:19:42 you'd give it a topic and then you'd
2:19:45 answer two or three questions about the
2:19:46 topic and then you'd hit submit
2:19:48 and it would email you like five pieces
2:19:51 of social media content a set of tweets
2:19:54 a blog post a set of like a a tweet like
2:19:59 a hashtag cloud and just all that all
2:20:01 sorts of [ __ ] and it was impressive and
2:20:03 it formatted it was like a nice email
2:20:05 right so I'm on the phone with this guy
2:20:07 and he goes he
2:20:09 goes yeah yeah I've used chat GPT I and
2:20:13 I said oh I just built this cool cool
2:20:14 tool you want to see it and he's like oh
2:20:16 sure so I said here let's let's so I
2:20:19 shared my screen and I pulled up the
2:20:21 Google form and I said give me a topic
2:20:24 and he gave me a topic and then I said
2:20:26 give you know give answer these
2:20:28 questions like why is that topic
2:20:29 important and whatever they were and so
2:20:32 I hit submit and so we're chatting and
2:20:34 I'm like oh it's this cool thing you'll
2:20:35 get an email in like 30 seconds so check
2:20:37 it out and so we're talking and going
2:20:39 and and all of a sudden I can see him
2:20:41 check his email and and he goes like
2:20:43 this he just
2:20:46 goes he he
2:20:48 goes
2:20:50 I he goes he goes
2:20:53 Kyle I had no idea he
2:20:56 goes I don't even know what to do with
2:20:58 this I got to go I got to go and he hung
2:21:02 up on
2:21:05 me it was wild it was wild but it's like
2:21:09 you you know if if you're looking at
2:21:11 this stuff from the outside it's really
2:21:12 hard to see what it is cuz it kind of
2:21:14 looks like a Google search bar and
2:21:16 you're thank you for the h dang why you
2:21:20 y paluca
2:21:22 um and it kind of looks like that and
2:21:25 then but everyone's telling you oh no
2:21:27 but this is really different this is
2:21:28 really remarkable and you're like well
2:21:30 what like what is it it's a chat chat
2:21:32 bot well haven't chat Bots always sucked
2:21:34 yes they've always sucked I hate that
2:21:37 they called it a chat bot because it's
2:21:38 not really that it's so much different
2:21:42 than that
2:21:44 and and then when people actually use it
2:21:49 that light bulb goes off where they're
2:21:50 just like oh
2:21:52 [ __ ] wait wait it can do
2:21:56 that but it can't do it can't do what I
2:21:59 do
2:22:01 right oh [ __ ] what do I
2:22:06 do right like that's the moment I think
2:22:09 we all deserve to like
2:22:12 confront and and and a big part of this
2:22:16 channel is like I I want you to confront
2:22:18 that moment on your terms not on someone
2:22:21 else's terms because that's starting to
2:22:27 happen extras are now going to charge a
2:22:29 bazillion dollars for an appearance um
2:22:32 here's the problem with
2:22:34 acting for for every job there are
2:22:38 10,000 people that would kill their
2:22:42 grandmother for that part it is it is
2:22:46 a
2:22:50 it's a rough business everyone wants to
2:22:52 be in it the media business in general
2:22:54 is a very rough business because it's so
2:22:57 sexy and it's so prestigious and
2:22:59 everyone wants to be a
2:23:01 star that and this is what sag's trying
2:23:04 to hedge against and I don't I don't
2:23:06 know how they fight this fight quite
2:23:07 frankly I mean this is a really tough
2:23:09 fight because as a producer if I can
2:23:12 save you
2:23:14 know $3 to5 million for from a
2:23:17 production by not having to bring in you
2:23:21 know a thousand
2:23:23 extras I'm going to fight for that the
2:23:26 actors are going to say well you've got
2:23:27 to bring in at
2:23:29 least 50% of the people in background
2:23:33 have to be real people like they'll
2:23:34 negotiate it to
2:23:36 some
2:23:38 um but it's that's a rough one like I
2:23:40 don't know how they're going to fight
2:23:41 for that because it's it's from an
2:23:43 economic
2:23:45 standpoint it's a it's a it's a tough
2:23:47 one
2:23:49 and again you've just got you know
2:23:52 thousands and thousands and thousands of
2:23:54 people that would literally kill for
2:23:57 that job for that
2:23:59 part it's just a it's just a rough rough
2:24:03 business Luna Labs is the extras I don't
2:24:06 know Luna Labs but I will look at it you
2:24:09 are on spot I couldn't agree more I have
2:24:11 no [ __ ] idea what I said but thank
2:24:13 you um I plan to play a bit when I move
2:24:17 it's going to be interesting Pablo I
2:24:19 would encourage you to play a bit right
2:24:21 now do it now before you move but yeah
2:24:25 play with it very well put thank you
2:24:28 don't know what I said discriminator
2:24:31 okay so AI will drive up no access
2:24:35 data behind pay walls well it did that's
2:24:40 so open AI had um browse with Bing that
2:24:45 where it was chat connected to the
2:24:48 internet
2:24:49 and it was
2:24:52 pulling content from behind pay walls
2:24:55 from Fortune Magazine and probably from
2:24:57 lots of
2:24:58 others and so they pulled it and then
2:25:01 two days 3 days later they announced a
2:25:04 Content deal with the Associated Press
2:25:06 and I would be very surprised if they
2:25:08 didn't cut a deal with time ink which
2:25:10 publishes Fortune Magazine
2:25:13 um I have a feeling open AI is going to
2:25:16 start cutting a lot of content deals um
2:25:19 cuz they're going to have to or they're
2:25:20 just going to get their asses suited out
2:25:21 of
2:25:22 business so and they don't like you know
2:25:24 they they want to be running their
2:25:25 business they don't want to be dealing
2:25:27 with with copyright lawsuits so they're
2:25:28 going to have to cut content deals
2:25:32 so and you know luckily they've got
2:25:35 they've got Microsoft behind them so
2:25:37 Microsoft is going to run interference
2:25:39 for them and they're going to say okay
2:25:40 here are the here are the important
2:25:41 relationships we don't want to [ __ ] up
2:25:43 and here's the people to talk
2:25:46 to um hey everybody hi Kyle hey the bear
2:25:49 what's happening what's shaking what's
2:25:51 shaking uh James Smith is it worth it to
2:25:54 pay for gp4 is there another way to use
2:25:57 GPT connected to the internet with good
2:25:59 answers okay so this is now a more
2:26:02 complicated answer than it was two days
2:26:05 ago I think it's still worth paying for
2:26:09 chat GPT plus which gives you access to
2:26:11 gp4 because you got access to code
2:26:14 interpreter and you also get access to
2:26:16 plugins I'm excited about code
2:26:18 interpreter than plugins but plugins are
2:26:19 going to be cool at some point they're
2:26:21 just they're just they're a little
2:26:23 marginally valuable right now they'll
2:26:25 get better but code interpreters [ __ ]
2:26:29 mind-blowing
2:26:31 um bing.com is chat gp4 so if you go to
2:26:36 the chat button on bing.com that's gp4
2:26:39 now how how have I explain this before
2:26:41 so here's your large language model this
2:26:43 is the gp4 large language model chat GPT
2:26:46 is actually an appc application sitting
2:26:48 on top of the large language model Bing
2:26:51 chat is
2:26:53 similarly an application sitting on top
2:26:55 of the large language model so they both
2:26:57 are
2:26:58 gp4 large language model but they're
2:27:00 they're different applications right so
2:27:02 they're not the same the quality of the
2:27:06 open AI one is better than the bing one
2:27:08 but the bing one actually does more [ __ ]
2:27:10 than the open AI one like it will
2:27:12 analyze images and it will gener
2:27:14 generate images if you ask it to
2:27:17 so those are two options for
2:27:20 gp4 then you have Claude AI which is
2:27:25 anthropics gp4
2:27:28 competitor that has a couple of things
2:27:30 about it that make it really really
2:27:32 compelling it's not quite as good as GPT
2:27:34 4 for what it's worth but you can upload
2:27:39 five files to it like five PDFs and
2:27:42 start interacting with them up to like
2:27:44 10 megabytes each it also has 100,000
2:27:47 token context window meaning you can put
2:27:50 75,000 words in that and have a
2:27:52 conversation with it or like 50,000
2:27:54 words I don't know the the depending on
2:27:57 how they calculate tokens it it might be
2:27:59 more or less than 75,000 or or between
2:28:03 50 and 75 um po.com you can get access
2:28:07 to gp4 and then Bard is Google's version
2:28:11 of maybe gpt3 Bard still sucks that's
2:28:14 why it's on the bottom so Claude Bing
2:28:17 and chat.
2:28:19 open.com are would be the contenders
2:28:23 Bing is free
2:28:25 gp4 chat. open.com is paid gp4 but it's
2:28:29 better and then Claude is a free
2:28:31 alternative that has some advantages
2:28:33 over gp4 but it's not as strong a
2:28:36 language model so just play around with
2:28:37 those three good question though it's
2:28:40 this shit's just just the the options
2:28:43 that we have right now are really
2:28:44 complicated right now cuz new
2:28:45 competitors are coming on under the
2:28:47 scene that are that are pretty good have
2:28:50 compelling uh strengths we don't need
2:28:53 teachers I disagree with that I think we
2:28:55 need teachers um to help
2:29:00 students learn how to interact with the
2:29:02 world um rather than learning how to
2:29:04 interact with information I think
2:29:06 there's a transition happening I think
2:29:08 we need teachers and
2:29:10 leaders got to go many thanks again
2:29:12 thank you Laura appreciate it you're
2:29:13 probably gone by now because I'm so far
2:29:15 behind I'm a damn questions
2:29:18 I'm a human with curiosity good enough
2:29:20 for me that's all you need to be right
2:29:21 now be a human be ravenously curious and
2:29:25 be adventurous and just go to these
2:29:27 things and start playing start figuring
2:29:29 it the [ __ ]
2:29:31 out so you go Focal Point Minnesota I
2:29:36 like it I've been trying to replicate an
2:29:39 ml repo for over 6 months in 2 days with
2:29:43 chat GPT
2:29:45 done wow
2:29:48 that's [ __ ]
2:29:50 outrageous that's amazing that's
2:29:53 amazing I mean and and now that you've
2:29:55 got things like code interpreter and
2:29:58 Claude do you know claw. where you can
2:30:00 upload big
2:30:02 files you could take entire GitHub
2:30:05 repositories and upload it to these
2:30:08 tools and start interacting with it tell
2:30:10 it to give you a step-by-step tutorial
2:30:12 for how to do it which it sounds like
2:30:14 that's what you did with with chat GPT
2:30:16 that's amazing Johnson very cool I kind
2:30:19 of disagree I think just like the
2:30:21 internet AI will be split public versus
2:30:23 private I well I think I I think you're
2:30:27 right I think that must have been in
2:30:29 reference to me saying it's going to be
2:30:30 trained on all of humanity or some
2:30:32 something around that you're definitely
2:30:35 going to have big private areas you know
2:30:37 the
2:30:38 the the early days of the web there was
2:30:42 very much sort of uh you know
2:30:44 information wants to be free kind of
2:30:46 attitude and there was lots of Open
2:30:47 Source stuff and a lot of that's going
2:30:49 on now with AI and then over time you
2:30:53 know the web got more and more
2:30:56 consistent and corporatized and all that
2:30:59 sort of stuff and you know we broke into
2:31:01 the major
2:31:02 players that's all now going to get
2:31:04 shattered and we're going to go through
2:31:05 this cycle again um the thing that I
2:31:09 think might be different here is that
2:31:13 because so much of so much of the power
2:31:17 of these tools is in the is in the open-
2:31:20 source Community now and it sounds like
2:31:23 meta is going to release a powerful like
2:31:26 GPT 3.5 competitor for free for the open
2:31:30 source Community um with a commercial
2:31:33 license they announced that two days ago
2:31:36 um I think we're going to see an
2:31:38 incredibly vibrant um public ecosystem
2:31:43 for for all of this stuff where you're
2:31:45 going to have
2:31:47 um if you work for a corporation you'll
2:31:50 use the Microsoft flavor right or the
2:31:54 Google Flavor if you're a small company
2:31:57 you might use one of the open source
2:31:58 models or or you know a commercial
2:32:00 entity built on top of Open Source um so
2:32:03 I think we're going to have lots and
2:32:04 lots of
2:32:05 options um but yeah I agree I there's
2:32:08 there's still going to be like the we're
2:32:11 human beings there's going to be power
2:32:13 centers and people are going to do
2:32:15 everything they can to maximize their
2:32:16 power that's never going to
2:32:20 change um gigo needs the experts to get
2:32:24 familiar and correct it Go garbage in
2:32:27 garbage out yep I agree I transcribed a
2:32:30 YouTube video with vid to text and use
2:32:32 CLA to turn it into two blog posts and
2:32:34 10 tweets there you go Vicki that's it I
2:32:37 know it's just like it's [ __ ] insane
2:32:41 it's it's just you you take like that
2:32:43 that one little element of of what you
2:32:45 do during a day day right if part of
2:32:47 your job was doing social media content
2:32:50 and you're like okay I got to go to
2:32:52 YouTube I'll download the transcript
2:32:54 I'll read the transcript I'll pull out
2:32:56 the things I'll write the things da da
2:32:58 da da and that's with interruptions
2:33:00 during the day that's 2 three four hours
2:33:03 of your day and now you figured out how
2:33:05 you can do that like that and then you
2:33:08 could throw that in a zapier Automation
2:33:10 and
2:33:12 literally pop in the YouTube url have it
2:33:16 run the Automation and post that content
2:33:21 to a Google doc or to a spreadsheet or
2:33:24 to wherever doesn't [ __ ] matter
2:33:26 automatically post it right so yeah it's
2:33:30 and and that's like one little element
2:33:32 of what you do right and there's a
2:33:33 thousand of those that can all be
2:33:35 transformed in similar ways and then
2:33:37 there's new things that can happen so
2:33:39 it's
2:33:41 astounding not all needed data will be
2:33:44 free try reading New York Times article
2:33:46 without a subscription I don't disagree
2:33:48 with that RJ I think I think one of the
2:33:50 places companies are going to make
2:33:53 um make you know buckets of money is on
2:33:56 proprietary data look up if if you
2:33:58 haven't heard of it um Bloomberg GPT
2:34:01 they they they're taking proprietary
2:34:03 Bloomberg data they they embedded it and
2:34:06 threw it in a vector database and now
2:34:07 they're going to charge buckets of cash
2:34:11 to be able to GPT financial data and so
2:34:15 there's going to be tons of those I
2:34:17 think that's ultimately where
2:34:19 competitive advantages are going to live
2:34:21 the fact that open AI just cut a deal
2:34:22 they cut a deal with Shutterstock and
2:34:24 they cut a deal with the Associated
2:34:26 Press and I'm guessing that they're
2:34:27 going to cut a bunch more that means
2:34:30 open AI is going to be able to have a
2:34:32 version of chat GPT where you'll say
2:34:35 you'll check the Associated Press button
2:34:38 and say I want to I want to understand
2:34:41 you know this particular news event um
2:34:45 restricted to AP P
2:34:47 sources done like that's going to be
2:34:50 something worth paying for right and AP
2:34:52 will get a cut of that and Cat GPT will
2:34:54 get a cut of that and they're going to
2:34:55 have a ton of those so I think
2:34:57 proprietary data is a really interesting
2:34:59 place folks can make
2:35:02 money um I meant for the code for the
2:35:05 companies you built oh um so I'm more of
2:35:11 a product guy than the coding guy so so
2:35:13 we've got a tech team that builds our
2:35:16 app and builds the our CMS the back end
2:35:19 but but I designed it all like I I I
2:35:23 know enough about tech to know what we
2:35:25 need to do I just don't know how to do
2:35:27 it so I say here's what I want it to do
2:35:30 and then they build it and they hand it
2:35:31 to me and I make it break and I say
2:35:34 here's what's wrong with it
2:35:37 um yeah and we've been around for 11
2:35:40 years now so the the code base is fairly
2:35:42 well established and pretty stable um
2:35:44 but now that I'm I'm you know you know
2:35:47 rethinking it all with AI we're prob
2:35:50 what I'm probably going to do is so I
2:35:52 I'll have my code base here for my
2:35:54 current Story vine code base here and
2:35:56 then I think what I'm going to do is I'm
2:35:57 going to create an AI playground over
2:35:59 here and then at some point when stuff
2:36:01 gets real enough here I'll roll it
2:36:09 in yes thank you I look forward to your
2:36:11 broadcast daily thank you Firefly I
2:36:13 appreciate
2:36:15 that what do I need to make these crazy
2:36:17 stable diffusion videos um I assume you
2:36:21 mean the runway videos go to Runway ml
2:36:24 or download the runway ml app and then
2:36:27 you can either use video to video or
2:36:29 text to
2:36:30 video um and it just makes videos and
2:36:33 they're crazy and they're disturbing and
2:36:36 here here's the thing about all these
2:36:37 text to video tools they're they're
2:36:39 they're all kind of janky right now but
2:36:42 they are going to be de Jank ified
2:36:45 quickly if you if you look at the
2:36:48 quality of image generation a year ago
2:36:51 it was laughably bad and if you look at
2:36:54 what mid journey is doing today it's
2:36:56 disturbingly good video is going to
2:36:59 follow the exact same path in fact the
2:37:01 CEO of stable diffusion stability
2:37:04 AI said that their current internal
2:37:07 working model that they're working with
2:37:09 so on stable diffusion you know you you
2:37:11 type in text and you know whatever it is
2:37:14 10 seconds later it generates the image
2:37:15 for you
2:37:16 the model that they're working with
2:37:18 internally at stability AI is generating
2:37:21 images 60 a
2:37:24 second 60 images a second is the speed
2:37:27 that the the the internal model at at
2:37:30 stable
2:37:31 diffusion
2:37:33 um why is that relevant well 30 frames a
2:37:37 second is video 60 frames a second is
2:37:39 video games well 120 but but 60 is
2:37:43 pretty good um dynamically
2:37:46 generated you know so so you're going to
2:37:49 have
2:37:50 fullframe fullframe video being
2:37:52 generated from a text
2:37:54 prompt um in the not too distant future
2:37:58 so Runway ml all these things that you
2:38:00 look at it and you're like well it's
2:38:02 kind of cute but it's very disturbing
2:38:03 cuz when he went to go drink the beer it
2:38:06 went through his head that's all going
2:38:07 to go it's all going to be fixed so so
2:38:10 again what I would strongly encourage
2:38:12 you to do is is go learn those tools
2:38:15 start playing with them
2:38:16 um the other thing I would do is go to
2:38:18 futurepedia doio there are probably a
2:38:21 lot of video tools that that I don't
2:38:23 know about um I know there's uh
2:38:28 descript there's Runway ml's the most
2:38:30 famous one right now cuz their text of
2:38:32 videos pretty good there's descript
2:38:35 there's another one I can't remember the
2:38:37 name of and there's probably a ton of
2:38:38 them there so just go to the video
2:38:40 section at futurepedia
2:38:43 doio which Tech vendor do you work for
2:38:46 for uh it's a company called storyvine
2:38:48 I'm the founder and the CEO of it and
2:38:50 it's a an automated video storytelling
2:38:52 platform story vine.com
2:38:54 they might come up with a system that
2:38:56 works this time I don't know who that's
2:38:58 in reference to I am DocuSign X
2:39:01 Salesforce and sap interesting um yeah
2:39:05 this so I think that
2:39:08 um I think that these large language
2:39:12 models like like Salesforce in my
2:39:15 opinion
2:39:18 no
2:39:20 one should ever have to deal with that
2:39:23 [ __ ]
2:39:25 interface ever it's a [ __ ] travesty I
2:39:29 don't know why crms have to be as
2:39:30 complicated as they are but they
2:39:33 are and it's a [ __ ] disaster
2:39:37 but you could throw a large language
2:39:40 model simplification on top of it and I
2:39:42 know that Salesforce just announced that
2:39:43 they've got 10 like 10 AI tools that are
2:39:47 going to be built into it um but but I
2:39:50 should be able to just I should be able
2:39:52 to just call someone record that call
2:39:56 and it just automatically knows this
2:39:59 person's new they're not in Salesforce
2:40:00 it auto it auto creates the contact the
2:40:03 account the opportunity because it it it
2:40:06 understood what we talked about and it
2:40:08 was clear that this person actually has
2:40:10 budget so it throws it into whatever the
2:40:12 [ __ ] part of the funnel it needs to be
2:40:15 in I should have to think about that
2:40:17 [ __ ] at
2:40:22 all someone needs to go build that and
2:40:24 if Salesforce isn't building it shame on
2:40:27 you I know you don't work there you're
2:40:29 your ex Salesforce but DocuSign is
2:40:32 similarly complicated and sap is just a
2:40:34 whole another you know nightmare in of
2:40:36 itself um these are all powerful tools
2:40:39 don't get me wrong like businesses need
2:40:41 these things to run but they are so
2:40:44 complicated if you think of the amount
2:40:46 of hours that salespeople have to spend
2:40:49 inputting data into
2:40:53 Salesforce if they could be on the
2:40:55 [ __ ] phone
2:40:56 selling rather than inputting data on
2:41:00 Salesforce we would probably double our
2:41:02 revenue and I'm a small
2:41:05 company I mean if you could
2:41:09 get if you could get a 10% increase in
2:41:13 productivity out of salespeople for a
2:41:15 big Pharma company
2:41:17 [ __ ] that all that all drops to the
2:41:21 bottom line I asked for a meal plan Bing
2:41:23 gave me links Bard gave me a plan and a
2:41:26 shopping list yeah that's cool uh try
2:41:28 claw. a and try chat GPT it's actually
2:41:31 now that there's so many um options here
2:41:34 the other night on here we did this it
2:41:36 was fun someone someone asked can chat
2:41:38 write a joke and I'm like probably not
2:41:40 but let's go try and so we we we wrote I
2:41:43 wrote a single prompt to write a joke
2:41:45 about
2:41:46 avocados and then we went to each of
2:41:48 these tools and and used the exact same
2:41:51 prompt AC prompt across the tools so so
2:41:54 that's it's a really interesting
2:41:55 exercise to learn how good or bad these
2:41:57 things are and how they're different
2:42:00 because some are some are like the pi
2:42:03 one that's not up here pi. a pi. a um is
2:42:08 super chatty and like I'm going to be
2:42:11 your best friend and then some like Bard
2:42:14 is more sparse and Bing is more
2:42:16 Microsoft corporate you know and chat
2:42:19 GPT is you know decently conversational
2:42:21 but they're they're all very very
2:42:23 different it's a good exercise to take
2:42:25 the same prompt and do it across
2:42:28 things all right using it right now pain
2:42:31 points to a 9 to-5 job very
2:42:35 cool that's
2:42:36 good uh laughing out loud don't worry
2:42:39 and don't twerk see this is someone who
2:42:42 understands the reality of this
2:42:44 channel the day I Twerk is the day this
2:42:47 thing gets shut down for
2:42:50 good how did you lose your Tik Tok
2:42:53 Channel I don't want to talk about it I
2:42:55 thought it was a good
2:43:00 song guess I shouldn't have shown my
2:43:02 butt crack all right we've warned you
2:43:05 before about twerking I know Vicky I
2:43:07 know I know I'll put my back out exactly
2:43:12 exactly I know I know that these nightly
2:43:14 things are over when my butt gets numb
2:43:16 so it's you know we're not far we're not
2:43:18 far from putting the back out I want to
2:43:21 build a proven technology that provides
2:43:23 free energy I think these things these
2:43:26 things are going to lead us
2:43:29 so because these these what what these
2:43:32 tools really do is they amplify what
2:43:35 people want to
2:43:37 accomplish and so people who want to
2:43:39 accomplish things like free energy and
2:43:42 cures for cancer
2:43:44 and their effort effs get Amplified and
2:43:47 if nothing else rather than being able
2:43:50 to do one experiment in a week maybe I
2:43:53 can do four or maybe I can do 10 or
2:43:57 maybe I can do virtual experiments and I
2:43:59 can do a thousand and in those thousand
2:44:03 that might have taken a thousand weeks
2:44:05 historically maybe the answer is in one
2:44:07 of those thousands that I do in a
2:44:09 week and that's we're going to start to
2:44:12 see an acceleration in scientific
2:44:14 breakthroughs that's going to it's just
2:44:16 gonna [ __ ] melt faces it's it's gonna
2:44:18 change everything so you go for it you
2:44:24 go for it user 6801 7466 38 3133 I would
2:44:30 think one of the things that you could
2:44:31 use chat GPT for is to come up with a
2:44:33 clever
2:44:35 name cuz cuz we do rate names here and
2:44:39 just sticking with the old user number
2:44:42 no that's zero out of 10 zero out of 10
2:44:47 zero although you know what I'll give
2:44:50 you one out of 10 because you showed up
2:44:52 isn't isn't that how some Tesco if you
2:44:53 get your name right on the cover letter
2:44:55 you get you get a point so you get a
2:44:58 point for being here all
2:45:00 right just here to learn how to make
2:45:02 money you can make money
2:45:06 watch is that voice make money with Kyle
2:45:09 and chat gbt oh in that voice yeah
2:45:13 exactly all right how much of a game
2:45:15 changer is code generator code
2:45:17 interpreter it's it's staggering you
2:45:20 want to see it let's code
2:45:23 interpreter which by the
2:45:25 way if you're
2:45:28 thinking G
2:45:30 Kyle code interpreter doesn't sound
2:45:33 interesting at all and it sounds kind of
2:45:35 geeky and this is is it me or is that a
2:45:38 [ __ ] horrible name you're correct all
2:45:42 right and what the engineers have told
2:45:44 me over and over again is well
2:45:46 technically it's accurate it's it's code
2:45:47 interpreting it's it's a it's a common
2:45:49 term in the it's it's a term of Art in
2:45:51 the in in the the industry is as an
2:45:53 engineer I I know exactly what code
2:45:55 interpreter means
2:45:57 and it's [ __ ] horrible name all right
2:46:01 at some point we're going to come up
2:46:02 with a new name we're going to we're
2:46:04 going to fix Tik Tock scrolling bug and
2:46:06 we're going to come up with a better
2:46:07 name for code interpreter but in the
2:46:09 meantime it's called code interpreter so
2:46:13 how you get to this is I'm at chat .
2:46:15 open.com so this is chat GPT I've paid
2:46:18 for chat GPT plus which is the little
2:46:21 yellow flag there and then in gp4 you
2:46:24 see code interpreter if you don't see
2:46:26 this here you need to enable it in your
2:46:28 settings so you go to uh your little
2:46:32 name you go to settings and then you go
2:46:34 to Beta
2:46:35 features and you turn it on so you got
2:46:39 plugins and you got code interpreter and
2:46:41 when they bring back browse with Bing
2:46:42 that'll be here as well okay
2:46:46 so now we got code
2:46:51 interpreter and what happens is so
2:46:53 you've got your little prompt window
2:46:54 here and you've got a plus
2:46:56 button and we'll
2:46:59 upload
2:47:01 something let's see is it
2:47:07 downloads all right I don't even know
2:47:09 what this is so I went to if if you want
2:47:11 to get data sets you can go to um
2:47:16 kaggle K A gle.com
2:47:20 and they'll give you data sets and I
2:47:22 think this I think this one might
2:47:25 be Denver crime stats but there's all
2:47:28 sorts of data sets you can get but it's
2:47:30 a zip file and so what I'm going to do
2:47:32 now is I'm going to say um what's the
2:47:37 data in this file question
2:47:40 mark so the first thing it'll do it says
2:47:43 let's start start by unziping
2:47:48 it okay and so it says the zip file
2:47:52 contains crime Offense Code and crime
2:47:54 XML let's take a look at the first few
2:47:57 rows so let me show you what's actually
2:48:00 happening here because here's here's the
2:48:02 first part of what's mind-blowing about
2:48:03 this so not only can It open a zip file
2:48:06 that's cool how did it open the zip file
2:48:08 well it wrote python code and executed
2:48:12 it it imported Library it wrote python
2:48:15 code to import some libraries that unzip
2:48:17 files and it wrote the code and executed
2:48:20 it and then it said the result of my
2:48:21 code was here's the names of the files
2:48:24 and so that's how it wrote this and then
2:48:26 it said okay so you asked me what was in
2:48:29 here so let's go ahead and look in
2:48:33 here and now it looked inside those
2:48:35 files so it knows it knows what's in
2:48:38 here so here's all the different data
2:48:40 points now it's saying let's inspect the
2:48:42 XML
2:48:44 file
2:48:47 so the XML file has things like title
2:48:50 geoform for Geographic data bounding
2:48:54 coordinates data access usage
2:48:58 constraints right so so now it
2:49:00 understands how this data is going to be
2:49:04 formatted with that XML file so it says
2:49:07 it's important to note this is just a
2:49:09 partial view of the XML file and it
2:49:11 might contain more information for full
2:49:12 analysis you might process the entire
2:49:15 XML file considering its complete
2:49:17 structure I'm going to say
2:49:20 um
2:49:23 please
2:49:25 analyze
2:49:28 and
2:49:29 visualize in the most interesting and
2:49:35 beautiful be uh beautiful way
2:49:42 possible
2:49:44 possible
2:49:45 the
2:49:48 trends in this
2:49:53 [Applause]
2:49:55 data so it so it says sure let's start
2:49:59 by loading the complete CSV file and
2:50:01 analyzing the data and I think it'll
2:50:03 break the first time I've done this
2:50:05 before because it it's encoded it uses
2:50:07 the wrong encoder yeah it did it it did
2:50:09 it again so we en encoded it
2:50:13 in whatever and there are some non utf8
2:50:17 characters in the file so let's try
2:50:20 doing it again with ISO
2:50:22 8589 or
2:50:26 8859 so then it
2:50:28 says Let's do let's start the analysis
2:50:32 by doing the the following check the
2:50:34 time span analyze the distribution of
2:50:36 crimes over time check the most common
2:50:39 offense types analyze the geographical
2:50:41 distribution of crimes check if there's
2:50:44 any trender pattern in the time hour of
2:50:47 the day day of the week month of the
2:50:49 year when crimes occur first let's start
2:50:51 with a time span of the data set so it's
2:50:53 writing it's writing code for all of
2:50:57 these it looks like this one might time
2:50:59 out and so if it times out it'll write
2:51:01 new code so so this is the kind of stuff
2:51:04 code interpreter is doing is it's doing
2:51:07 full-on data analysis full-on coding
2:51:09 what it's actually doing is it's
2:51:11 spinning up a virtual
2:51:14 machine for every chat session that you
2:51:16 use and it's running it's running itself
2:51:19 um it it's basically running like a
2:51:21 little mini computer inside chat GPT
2:51:25 that writes and executes code it's got
2:51:27 its own set of libraries it can
2:51:30 use and then it just goes and write
2:51:32 stuff there there it did a a crime chart
2:51:35 so that's a line plot of what number of
2:51:38 crimes over time
2:51:42 monthly and it's still working it's just
2:51:44 it's literally doing data
2:51:48 analysis remember I asked for it to be
2:51:50 beautiful here's a bar chart that it did
2:51:52 in lovely rainbow colors which I'm sure
2:51:54 would
2:51:56 offend many of
2:51:59 the people looking at crime stats but
2:52:07 whatever they did some sort of scatter
2:52:09 chart that that looks like it's a bad
2:52:11 scatter chart this scatter plot shows
2:52:13 distribution of random sample of
2:52:15 crimes um so so that you could tell it
2:52:19 it screwed something up there so to fix
2:52:20 it and it'll go fix it it's still going
2:52:24 right so it it came up with its own set
2:52:27 of instructions all I said to it was
2:52:29 analyze this and visualize it make it
2:52:33 beautiful so that's that's code
2:52:36 interpreter so yes what was it how much
2:52:38 of a game changer code interpreter um I
2:52:42 think it's big you can upload images and
2:52:44 say What are the the top 15 average
2:52:46 colors in this image and give me their
2:52:48 RGB values and display them and give
2:52:50 them the name of a creative crayon color
2:52:53 and it
2:52:55 will it's nuts now what do I what can
2:52:58 you do with it that I'm not as clear on
2:53:01 cuz I don't know data science at all
2:53:04 like at
2:53:05 all um but I know that what I just
2:53:08 watched it do is staggeringly
2:53:13 powerful
2:53:17 no more teachers or 10e professors I
2:53:19 don't know about that I think teachers
2:53:21 are here you're so right I'm glad but I
2:53:23 don't know what I was saying how can I
2:53:25 leverage chat GPT in real estate um many
2:53:29 many
2:53:30 ways um go play around with some of the
2:53:32 real estate plugins in chat GPT you can
2:53:36 have it do comps you could have have it
2:53:39 write you marketing copy you could have
2:53:41 it it could it could write you video
2:53:43 scripts for Tik Tok
2:53:45 you could just if you got the prompt
2:53:47 right you could probably give it an MLS
2:53:51 listing and it would pull the
2:53:55 relevant data from the MLS listing
2:54:02 and yeah you could you could probably
2:54:05 you could probably create a prompt that
2:54:07 would create a Tik Tock script for you
2:54:09 or whatever like a YouTube script um
2:54:12 that just talked about like the
2:54:13 neighborhood and the school TOs and the
2:54:16 how cool the house is and what are the
2:54:18 unique pieces of the thing
2:54:21 and you could have it write letters
2:54:24 Prospect letters you could have
2:54:26 it summarize call notes and
2:54:29 automatically follow up with emails I
2:54:32 would look into zapier to do automations
2:54:34 to automate a lot of this
2:54:37 stuff
2:54:41 um with with Bing you can upload a
2:54:46 picture to Bing you could upload a
2:54:47 picture of the house assuming you're not
2:54:49 in commercial real estate but even if
2:54:51 you
2:54:52 are upload a picture of the house to
2:54:54 Bing and have it
2:54:57 write social media copy for that house
2:55:00 and just see how it does I I don't know
2:55:02 if it'll do good but it understands what
2:55:04 a house is and it understands what
2:55:06 copyrighting is and you could say I'm
2:55:10 you know you're a real estate agent
2:55:12 who's who specializes in writing really
2:55:15 interesting social media Co copy based
2:55:17 on the you know the the you know the
2:55:20 image of a
2:55:21 house I don't know uh like what I would
2:55:25 say is start using these tools start
2:55:26 playing with them start understand what
2:55:28 they make possible and then just dig
2:55:30 deeper and deeper and deep take take
2:55:32 little pieces of your job and start to
2:55:35 use these things to amplify it I really
2:55:38 hate writing letters I'm not really good
2:55:40 at it okay great perfect take if there's
2:55:44 a an email that you wrote that is a
2:55:47 masterpiece took you 3 hours to write it
2:55:49 and it was super painful but you love
2:55:51 that email and you use that as as your
2:55:54 template and it's always painful take
2:55:57 that pop it into chat GPT say hey you're
2:56:00 a real estate agent and
2:56:02 here's an an email that you sent
2:56:05 recently paste it in paste now write me
2:56:09 a new email to this person about this
2:56:12 property address and follow the T tone
2:56:15 and structure of the email above out
2:56:17 it'll come it'll be perfect every time
2:56:19 you'll never have to write another
2:56:21 email start with [ __ ] like that little
2:56:23 pieces of your job make them better and
2:56:25 then just start to expand your little
2:56:27 circle of the things you can do get
2:56:29 bigger and bigger and bigger and tell
2:56:31 all your co-workers about it share it
2:56:34 talk about
2:56:36 it is there any AI out there that's
2:56:39 helping things like affiliate marketing
2:56:41 website funnels there's a [ __ ] ton of
2:56:42 them and there's there's plugins for
2:56:44 them Sean thank you for that what is
2:56:46 that that's like a little baby dinosaur
2:56:48 cute love it all right um go to
2:56:53 uh affiliate marketing funnels so go
2:56:56 start playing with chat GPT plugins
2:57:00 there's a bunch of SEO and affiliate
2:57:01 marketing [ __ ] in there and then I would
2:57:03 also go to Future
2:57:13 pedia.com training manuals or things
2:57:16 like
2:57:17 that just pop it all in there like go to
2:57:20 claw. that's got 100,000 token context
2:57:24 window and just upload a bunch of PDFs
2:57:27 of you know training manuals from your
2:57:30 affiliate marketing thing and just have
2:57:32 it start writing copy for
2:57:34 you and and have it design you know
2:57:37 website funnels and website funnel copy
2:57:40 and all that sort of stuff I think this
2:57:41 shit's going to be brilliant for
2:57:43 affiliate marketing but I guarantee you
2:57:46 a lot of the affiliate marketing people
2:57:47 are all over it but they're all the ones
2:57:51 that are like you can make money with
2:57:52 Chachi
2:57:54 P you know you can white label this you
2:57:57 can make money with Chachi p with your
2:57:58 face saying you can make money with
2:58:01 Chach just sign up for my multi-level
2:58:05 scheme uh uh support
2:58:08 network so we can all make money with ch
2:58:11 telling people that they can make money
2:58:13 with ch
2:58:17 not that I'm bitter and cynical about
2:58:18 affiliate
2:58:19 marketing or as we called them in my
2:58:22 genx days pyramid
2:58:26 schemes all right they're not all
2:58:29 pyramid schemes I know that don't get on
2:58:31 my case I'm just being bitter all right
2:58:34 will Bard or Bing AI help with planning
2:58:37 a trip and train travel they all will
2:58:40 and in fact go to if you if you pay for
2:58:42 chat GPT plus go to plugin there's a
2:58:44 bunch of travel plugins in there um but
2:58:47 Bing is probably the best
2:58:48 one because it's connected to the
2:58:50 internet so just go to Bing and tell it
2:58:52 to not only book your travel for you but
2:58:56 uh plan daily itineraries find good
2:58:59 restaurants all that sort of [ __ ] all
2:59:02 right it's 10:12 I got to go people I
2:59:04 can't it's Saturday night what are you
2:59:06 doing here with me why is everyone still
2:59:10 here why are you here why are you here
2:59:17 well because of the entertainment value
2:59:20 clearly all
2:59:23 right how can I leverage from Real
2:59:25 Estate we just talked about that it was
2:59:27 awesome what's that one called I don't
2:59:29 know from chbt I mean we talked about
2:59:32 that what's that product what was the
2:59:38 product what was I doing what did I
2:59:41 do I don't remember what I did
2:59:45 oh I don't know I don't know where I was
2:59:47 sorry you got to be more because I'm
2:59:50 behind on all my uh comments if you just
2:59:54 say what was that or it I don't know
2:59:56 what it or that
2:59:58 was I want to know when the AI has
3:00:00 access to Scientific abstract and
3:00:03 discussion but there is a pay wall so go
3:00:06 to here I'll show you mation if you're
3:00:10 still here
3:00:13 mation
3:00:15 B so we're going to go to chat GPT
3:00:20 plugins and then we're going to turn off
3:00:24 zappier and we're going to turn
3:00:31 on X papers and ml paper
3:00:37 reader and I think there might be more I
3:00:40 might go I'm going see go see if there's
3:00:41 other ones out
3:00:43 there
3:00:53 hey wome of the day so
3:00:56 paper no plugins found for paper oh
3:00:59 that's
3:01:00 pepper paper XP papers paper chat search
3:01:04 through archive Publications I already
3:01:07 got that
3:01:08 one search for the latest research
3:01:11 papers on a specific to topic from
3:01:13 PubMed I'll install that
3:01:15 one it looks like there's a bunch of
3:01:17 plugins for this so 250 million
3:01:20 scientific papers that one's called
3:01:24 science free and reliable academic
3:01:26 search some now some of these
3:01:30 may require accounts behind pay walls
3:01:33 but I think some may not lit Maps so
3:01:38 that's get help exploring scientific
3:01:43 literature all right well whatever let's
3:01:45 go let's go back out of
3:01:48 here and so I've got what do I have
3:01:50 turned on I've got ml paper
3:01:55 Reader X papers and next paper I'm going
3:01:59 to put on science because that one seem
3:02:01 more scientific okay so now I'm going to
3:02:03 to um I
3:02:06 want to find
3:02:09 research on um effective
3:02:15 high
3:02:17 do
3:02:21 antibiotic
3:02:22 treatment of
3:02:25 longterm lime
3:02:28 disease and let's see what it does it
3:02:30 should fire
3:02:32 up one or more of those plugins using
3:02:37 science so the way the plugins work is
3:02:39 they kind of understand based on what
3:02:41 you're looking for might I be the right
3:02:43 plug PL in for
3:02:45 this and so here are some recent papers
3:02:47 on the topic high dose antibio biotic
3:02:50 treatment for long-term Lyme disease
3:02:52 manage of pediatric Lyme disease updated
3:02:54 Lyme disease guidelines con confirm
3:02:56 current best practices [ __ ] [ __ ]
3:02:59 current best practices are [ __ ] this
3:03:02 is something I know a lot about fever in
3:03:05 return traveler um and then I could also
3:03:08 go um
3:03:12 what is the
3:03:15 latest um
3:03:17 research with regards
3:03:21 to large context
3:03:25 windows for
3:03:28 llms and this should fire off a
3:03:30 different plugin this should fire off
3:03:32 one oh it's using science as well that's
3:03:34 okay let's see what it
3:03:37 does here's some of the latest scaling
3:03:40 laws for auto auto regressive gener
3:03:43 generative
3:03:45 modeling from 2021 these seem like
3:03:48 they're older but anyway go play with
3:03:51 the plugins and and maybe just turn them
3:03:53 on and off one by one until you find one
3:03:55 that works so hopefully that that helps
3:03:59 hopefully that helps with the finding of
3:04:03 the papers the other thing you can do is
3:04:05 use Bing which is connected to the
3:04:07 internet and oh is that but is that what
3:04:09 you asked about yeah it there might be
3:04:12 things behind pay walls but those
3:04:13 plugins might do it I don't I honestly
3:04:15 don't know enough about the the research
3:04:18 world to be of any help more than
3:04:30 that actors can monetize prompt engineer
3:04:33 portfolio so theater groups produce
3:04:35 their own original Works
3:04:37 true just get I mean listen the thing
3:04:40 about any disruptive technology is that
3:04:42 it's in in the immediate it's [ __ ]
3:04:46 terrifying because it's changing the
3:04:48 game but long
3:04:52 term actors are going to find ways to
3:04:54 act I here's the here's the thing thank
3:04:56 you very much here's the thing that I
3:04:58 think is going to be a a big deal
3:05:02 is
3:05:04 authenticity the value of authenticity
3:05:07 is going to go up
3:05:09 dramatically when the world is inundated
3:05:12 with autogenerated content we're going
3:05:14 to see all this content that's going to
3:05:15 look super slick on the surface and then
3:05:18 it's all just going to be kind of this
3:05:19 numb sameness and it's going to get
3:05:22 really boring really quick and you know
3:05:24 what's going to rise above the
3:05:27 noise human
3:05:30 beings that find ways to connect I mean
3:05:33 Taylor Swift is playing right now 10
3:05:35 minutes behind me there's [ __ ]
3:05:38 880,000 people in that Stadium singing
3:05:42 every single word of every single song
3:05:44 for however many [ __ ] hours she she
3:05:47 performs it's [ __ ] insane those kind
3:05:51 of experiences in my opinion go up in
3:05:53 value dramatically when all this
3:05:56 synthesized [ __ ] overwhelms us right and
3:05:59 so what our actors going to do they're
3:06:00 going to find ways they're going to find
3:06:02 ways to rise above the
3:06:04 noise so and you
3:06:06 know movies might look different movies
3:06:09 might be completely
3:06:11 transformed but
3:06:14 we're going to
3:06:16 still realize that there is something
3:06:19 different about a human being doing it
3:06:21 making a human connection than the
3:06:25 Bots there's going to be a lot of [ __ ]
3:06:27 where it doesn't matter if it's a
3:06:29 bot right if it's a [ __ ] Coffee
3:06:33 commercial does it really need to
3:06:36 be anything other than like all they're
3:06:38 doing is hiring the most corporately
3:06:40 Bland human being anyway right they're
3:06:43 they're the a lot of those commercials
3:06:45 they're bleeding the humanity out of
3:06:47 them anyway because they don't want to
3:06:48 [ __ ] offend
3:06:50 anyone so let those things be
3:06:53 synthesized does that [ __ ] some things
3:06:55 up sure but you know what it does it
3:06:56 frees up those creative people to do
3:06:58 more interesting
3:07:01 [ __ ] what does that look like I don't
3:07:04 know but I think authenticity the value
3:07:07 of it's going to be up there and actors
3:07:09 are going to find ways to fill that
3:07:10 space and creative people are going to
3:07:12 find ways to fill that space and
3:07:14 previously non-creative people who have
3:07:17 found their creative voice because of
3:07:18 these tools will fill that
3:07:21 space
3:07:24 so
3:07:29 yeah is he live on a couple of platforms
3:07:32 and reading comments from someone else
3:07:34 well Aurora Richard Hill realtor no I'm
3:07:38 just very B far behind on my comments
3:07:40 and I haven't gotten to you yet but now
3:07:42 that I've gotten to you I'm talking like
3:07:44 a smarmy [ __ ] because I
3:07:47 can am I alive who
3:07:50 [Laughter]
3:07:54 knows okay sorry about that uh yeah no
3:07:58 I'm just behind all my questions and
3:07:59 that's what happens that's what happens
3:08:01 theater group can create characters best
3:08:04 suited for the people in the group those
3:08:06 with disabilities oh yeah that's
3:08:08 actually a really interesting idea can
3:08:10 you say the name of the customer service
3:08:12 AI again I can I still have it here air
3:08:17 air.
3:08:19 air. this is the future in my opinion
3:08:22 this is the future in my opin I don't
3:08:24 know if this company is going to be the
3:08:26 future but 100,000 sales and customer
3:08:29 service reps at the to of a button yes
3:08:32 blow up the phone tree right now flush
3:08:35 it down the [ __ ]
3:08:40 toilet God I hate those things
3:08:45 like here's you want to know why gen
3:08:47 xers are so [ __ ] angry do you really
3:08:50 want to
3:08:52 know we
3:08:56 remember it used to be when you got on a
3:08:59 plane and you sat down and it was more
3:09:02 than a 30 minute flight they'd bring you
3:09:04 a nice hot meal and they' treat you like
3:09:07 a human being and they were
3:09:09 nice and we remember that that's gone on
3:09:14 and then we also remember if you spent a
3:09:17 lot of money with a company and you had
3:09:19 a problem you picked up the phone that
3:09:21 was connected to the wall and you had to
3:09:24 you had to stand there and you had
3:09:26 to do this with the things but when you
3:09:29 called a company a [ __ ] human being
3:09:31 answered and you talk to them we
3:09:33 remember
3:09:34 that that's why we're
3:09:37 bitter [ __ ] bag of pretzels the size
3:09:40 of a half of a pretzel that's not a me
3:09:43 meal and a phone tree isn't a [ __ ]
3:09:49 conversation bring on the
3:09:51 robots I want to talk to a robot I don't
3:09:55 want to press nine if I'm slightly
3:09:58 [Laughter]
3:10:04 inconvenienced every now and then I just
3:10:06 get the image of someone scrolling
3:10:08 through Tik Tok and they just see me
3:10:09 [ __ ] ranting about being bitter and
3:10:13 makes me
3:10:14 happy I'm hoping there will be an AI
3:10:17 that can do principal component
3:10:20 analysis I don't know what that is
3:10:22 mommination but I hope for your sake
3:10:24 there is too I assume there could be it
3:10:28 seems to me that code
3:10:31 interpreter could do thatc but I don't
3:10:34 know what it is but I assume that you've
3:10:37 got some sort of data set that you could
3:10:39 put in there and you could ask it to do
3:10:41 principal component analysis and I bet
3:10:43 it
3:10:44 will but I but I'm so ignorant about
3:10:47 what that term even means I could go
3:10:49 look it up on chat gbt right now and be
3:10:51 like a typical man and act like I know
3:10:53 what you're talking about but I don't
3:10:55 mansplaining is a service you can make
3:10:57 money with
3:10:58 CHT oh okay so we'll do a we should also
3:11:02 do mansplaining as a service as a
3:11:04 t-shirt cha PT mansplaining is a service
3:11:08 all right what is the thing in your
3:11:10 window that says 100,000 is that air AI
3:11:14 it's too small to see yeah that was air.
3:11:16 a yep that was it that was it that was
3:11:19 it it it it being search engine gives
3:11:23 links to info it finds
3:11:26 exactly uh
3:11:30 uh oh it's you're all talking to each
3:11:33 other which I think is super cool AI
3:11:35 theater instead of casting people for a
3:11:36 role you can create a role based on the
3:11:39 Caster character yeah that's the kind of
3:11:41 so what you're talking about Chris is
3:11:46 the so the short-term fear that AI is
3:11:49 going to take our jobs or we're all
3:11:51 going to lose our jobs a lot of people
3:11:52 are going to lose jobs but AI is going
3:11:53 to generate a bunch of jobs and they're
3:11:55 going to generate jobs based on
3:11:57 rethinking how things are done
3:12:00 done no almost okay you done yeah
3:12:05 soon what are you doing still working
3:12:08 yeah
3:12:10 nice doggy ah doggy's good he has hasn't
3:12:13 been barking too
3:12:16 much hey fatty Hey fatty
3:12:20 hey come here
3:12:22 baby all right but yeah that kind of
3:12:25 innovation like even in theater like yes
3:12:28 yes yes yes more of that that's that's
3:12:31 what will come that's the stuff that
3:12:32 gets me excited but I am not naive
3:12:35 enough to think that there's not going
3:12:37 to be all sorts of [ __ ] storm pain to
3:12:41 get there which which that
3:12:44 sucks seen that new AI called claw. a
3:12:48 you mean that one looks pretty sick now
3:12:51 I use Bard chat gppt perplexity and
3:12:53 Claud I know I know Wilbert me too plus
3:12:57 today I was talking to pi. aai and now
3:13:00 I'm going to sign up for air. because I
3:13:02 want to tell everyone that I know to
3:13:05 [ __ ] replace their phone trees with
3:13:07 something like that I hope they're a
3:13:09 success I hope their technology doesn't
3:13:11 suck um wait
3:13:14 perplexity yeah and and Bing you should
3:13:16 throw Bing in there because Bing you can
3:13:18 generate images and it's connected to
3:13:20 the internet and have you created 360
3:13:23 panoramic images using AI I did using
3:13:26 that there's a the [ __ ] it
3:13:29 called there's a text to sort of 3D
3:13:32 world generator that makes a a 3D world
3:13:35 that you can generate it's a game
3:13:36 generating thing hang on let me I can't
3:13:38 ever remember the name of it
3:13:42 um
3:13:44 um
3:13:45 AI text to World wait text to
3:13:53 world
3:13:58 genor not deep AI blade Labs go to uh
3:14:02 Skybox blads
3:14:11 docomo and then
3:14:19 do and we'll
3:14:28 do okay we're going to
3:14:30 do
3:14:36 portal to a new
3:14:39 world where that world
3:14:49 is
3:14:50 full of
3:14:55 sentient
3:15:00 avocados generate let's see what it
3:15:11 does it's working it's working working
3:15:13 patience people
3:15:15 patience
3:15:19 Pat genius takes 38
3:15:24 seconds now can I turn off how do I turn
3:15:26 oh it turned it off for me so there you
3:15:33 go there's our sentian avocado
3:15:36 world look we've got an avocado we've
3:15:39 got an avocado Dash panel here
3:15:43 oh that's beautiful actually look at
3:15:46 that sort of like an avocado it's like
3:15:48 an apple avocado
3:15:51 Planet so yeah that's that 360 World so
3:15:54 this is Skybox blad labs.com I mean
3:15:58 what's what's fascinating about this I
3:16:00 forget how you do it can you download
3:16:04 this is this download
3:16:08 no
3:16:11 download
3:16:18 so I'm G to I'm going to open the
3:16:21 image and this this kind of takes some
3:16:24 of the mystery away from it but this is
3:16:26 what it
3:16:28 generated so this so all it's doing and
3:16:31 I I know like I'm saying all it's doing
3:16:33 like it's not doing
3:16:35 anything they've essentially pre-
3:16:37 prompted it to create a single flat
3:16:40 image that if you
3:16:43 wrap it around a a cylinder it looks
3:16:46 like a 3D world it's not actually
3:16:49 generating 3D art it's just generating
3:16:51 2D art that happens to be formatted like
3:16:54 that but then it
3:16:56 looks like
3:16:59 this because
3:17:01 [Laughter]
3:17:05 math right I mean it's crazy it's nuts
3:17:09 so like those curved lines here they
3:17:11 look straight cuz it's mapped against
3:17:13 it's mapped against the curved
3:17:14 projection
3:17:17 right so the the the next step the next
3:17:21 stage of this kind of thing is it will
3:17:23 generate something like this but it'll
3:17:25 be a full-on 3D World right it'll be a
3:17:29 generated 3D world and I can fly out
3:17:31 this window it's not there yet but holy
3:17:34 [ __ ] [ __ ] nuts
3:17:41 right
3:17:46 I mean what's amazing is all
3:17:50 of all of that that we just saw all of
3:17:53 mid Journey all of the protein folding
3:17:56 that Alpha fold did is all based on the
3:18:01 same basic technology here these these
3:18:03 weird Vector spaces these
3:18:06 math Vector
3:18:08 spaces
3:18:10 representing garant an amounts of
3:18:17 data I can't I still like that thing I
3:18:20 just showed the the blockade Labs
3:18:25 thing like six months
3:18:28 ago I don't know how old year a year ago
3:18:31 that wasn't
3:18:32 possible and now it's like I can draw a
3:18:35 portal and say make me a portal of a
3:18:37 world of avocados and it just [ __ ]
3:18:40 does
3:18:41 it
3:18:44 and with as with as fast as these tools
3:18:46 are advancing if you've been playing
3:18:48 with this stuff even for a
3:18:53 month what do we do in 5 years what what
3:18:56 is possible in 5 years what's possible
3:18:59 in a
3:19:00 year a year from now July next
3:19:04 year what are these tools going to be
3:19:07 they're going to be more [ __ ] magical
3:19:08 than they are
3:19:09 now so learn to use them now and just
3:19:12 keep up as best you can and stay
3:19:15 curious go play go to futur
3:19:23 pedia.com
3:19:38 oh can I tell us how or show us um
3:19:44 reference picture with the prompts do
3:19:46 you mean
3:19:48 in mid Journey or
3:19:52 in
3:19:54 uh
3:19:56 bing um let me show you in bing I just
3:20:00 rubbed my face it's all
3:20:06 red we're going to start
3:20:09 over I'm going to go more balance cuz I
3:20:12 don't know what to do here so let's see
3:20:14 um let me go grab a
3:20:26 picture okay that's a good
3:20:41 one
3:20:45 where am I what did where did I do why
3:20:47 did I do I'm so confused okay oh wait my
3:20:56 Bing do I need to reload
3:21:00 this come
3:21:05 on my Bing no longer has image
3:21:08 processing I do not know why that's
3:21:11 strange
3:21:13 that's a strand one I do not
3:21:16 know let's try
3:21:19 Bard b.google.r
3:21:25 an
3:21:27 image upload a file
3:21:32 jpeg yeah that's good we'll do that one
3:21:35 good picture that's a good mid Journey
3:21:37 image I created for the AI Salon well I
3:21:40 created it just cuz I created it and
3:21:41 then I thought that be good for the a
3:21:43 Salon Image so I'm going to say
3:21:47 um you are an
3:21:50 ad
3:21:53 creative expert in
3:21:57 wait expert in
3:22:03 Creative um
3:22:05 engagement
3:22:06 copy
3:22:09 please write 10 add
3:22:14 headlines that
3:22:18 incorporate wait
3:22:21 incorporate the
3:22:24 themes in this
3:22:28 image um for a
3:22:34 community of AI
3:22:39 enthusiasts let's see what it does so
3:22:43 now this is Bard Bard's not as good at
3:22:44 the wordy words um but it can analyze an
3:22:48 image and I don't know so
3:22:50 Bing in your little search bar with Bing
3:22:53 next to the microphone button if there's
3:22:57 a little camera icon next to it um it
3:23:01 will you can upload images and mine just
3:23:04 didn't have that for some reason okay so
3:23:06 it says sure here's here's two here are
3:23:09 10 ad headlines that incorporate the
3:23:11 themes in the
3:23:12 image so the art of the art of
3:23:15 artificial intelligence take a ride on
3:23:18 the future with AI so there's a VW bus
3:23:22 in there the New Frontier of creativity
3:23:24 surf the waves of AI Innovation so it's
3:23:27 a beach scene right here's the image
3:23:31 there's the image right surf the waves
3:23:35 that's pretty good and this is Bard
3:23:38 right Bard's not good at the wordy words
3:23:40 and this is not bad AI the next chapter
3:23:42 in human evolution that's got nothing to
3:23:45 do with that image the power to create a
3:23:48 better world the future is now answer
3:23:50 our creative needs the key to unlocking
3:23:51 our potential so a bunch of those are
3:23:53 just
3:23:54 generic
3:23:56 um uh too many are
3:24:01 generic can
3:24:03 you mention at least one
3:24:10 specific item in in the image for every
3:24:24 headline explore the future of art with
3:24:27 this painting by AI eh take a ride on
3:24:31 the future with this self-driving car eh
3:24:34 uh the New Frontier of creativity with
3:24:37 this so okay this is something that that
3:24:40 AI will do a lot it it kind of locked
3:24:43 in in a in a
3:24:46 linguistic um structure that it's using
3:24:50 in every single one of them with this
3:24:53 with this AI created Avatar with this AI
3:24:56 powered disaster with this medical
3:24:58 diagnosis tool like like it's so so Bard
3:25:02 like we're where're Bard's just going
3:25:03 off the rails here and it's it's it's no
3:25:06 longer even referencing the image it's
3:25:09 kind of referencing its own headlines
3:25:11 somehow
3:25:13 that's one of those things where this is
3:25:15 just it's I wrote a crappy prompt and
3:25:16 it's but it's it's being particularly
3:25:18 crappy this is the kind of [ __ ] that gp4
3:25:22 just doesn't do or doesn't do as much
3:25:25 but anyway that's how you would add add
3:25:27 an image in Bard or or
3:25:29 Bing um and then over in Discord if you
3:25:34 want to to add an
3:25:37 image it's a little more complicated but
3:25:40 it's doable so let me show you how to do
3:25:42 that so we're going to go take that same
3:25:48 image and I'm going to drag it from the
3:25:51 finder into the Discord window and then
3:25:54 I hit
3:25:58 return and it uploads that
3:26:01 image it's a big image it's 3 megabyte
3:26:04 image so now what I'm going to do is I'm
3:26:06 going to rightclick it or on a Mac you
3:26:08 can control-click it and I'm going to
3:26:10 copy the link to the image
3:26:14 and then I'm going to say
3:26:19 um slash imagine so that's how you make
3:26:22 an image in mid
3:26:26 journey and in the prompt instead of
3:26:29 writing anything right now I'm just
3:26:31 going to hit paste and I'm pasting in
3:26:33 the URL of the image right so basically
3:26:37 the first part of the prompt here is
3:26:40 this image and and then I'm going to
3:26:42 write
3:26:46 photograph of a VW
3:26:50 bus on a
3:26:54 beach at
3:27:01 Sunset I'll just leave it I'll just
3:27:03 leave it at that and let's see what it
3:27:05 does so in theory it should turn this
3:27:07 into a photograph rather than the
3:27:10 stylized art it is now now cuz I told it
3:27:12 to make a photograph so we'll see if it
3:27:14 does
3:27:15 it and this is part of the joy and
3:27:18 frustration of working with these tools
3:27:21 is that they are remarkable and then
3:27:23 they are remarkably frustrating getting
3:27:25 it to consistently do what you need it
3:27:29 to do or what you expect it to do is
3:27:32 just not always it's not always as
3:27:34 straightforward as you think but you do
3:27:36 get better at it and the tools get
3:27:38 better
3:27:40 and this is how we do what we do we does
3:27:43 what we does it looks like it's making
3:27:45 red VW buses which is this is
3:27:49 promising and I think they're looking
3:27:54 photographic we're building we're
3:27:57 building we're building I didn't give it
3:27:59 an aspect ratio so it just made them
3:28:03 Square although it does look like it
3:28:05 letterbox them which is
3:28:09 interesting all right one more
3:28:18 one more pass let's see which one of
3:28:21 these looks
3:28:26 good I think number one this one's got
3:28:30 some weird ass Reflections on
3:28:33 it and that one's just boring that
3:28:35 doesn't quite look like a VW bus so
3:28:38 we'll do number one so I'm going to
3:28:39 upscale number one
3:28:43 and it's going to go off and make me a
3:28:45 lovely
3:28:47 expanded
3:28:49 version nice there you go and now I'm
3:28:53 going to do zoom out 2x because if you
3:28:55 haven't seen zoom out it's pretty
3:28:56 amazing so I've got the bus here I'm
3:28:59 going to do zoom out and it's going to
3:29:02 give me four different options of this
3:29:04 where we see a lot more sky and a lot
3:29:06 more
3:29:10 Beach uh
3:29:13 wait what's it say that interaction
3:29:14 failed well you
3:29:18 suck it looks like mid journey is a
3:29:20 little busy tonight that interaction
3:29:22 failed well you still suck anyway all
3:29:25 right that's how you do that hope that
3:29:29 helps all right I said I had to
3:29:32 go 25 minutes
3:29:37 ago I wish I could quit
3:29:40 you
3:29:45 oh at least I entertained
3:29:50 myself we're having fun tonight Saturday
3:29:53 night hanging out in the air learning
3:29:55 lab what did you learn tonight I didn't
3:29:58 know
3:30:01 [ __ ] I joined a little late what was the
3:30:04 AI voice you were
3:30:06 showing that you said is like what AT&T
3:30:10 uses oh it wasn't an AI voice Jason it
3:30:14 was I was making fun of AT&T using phone
3:30:18 trees um and I want them to use AI
3:30:22 robots to just talk to
3:30:24 us um I think that's what I was talking
3:30:27 about the site was called air. a air.
3:30:33 a I assume it's a B2B thing that's
3:30:35 pretty expensive are you on Google
3:30:37 Scholar I'm not what's is Google Scholar
3:30:39 doing interesting stuff with AI right
3:30:41 now I would imagine they would be Google
3:30:49 Scholar and if you're asking me if if
3:30:52 I've been there if I'm there as
3:30:55 a professor no I'm not maybe I should
3:31:02 be no I don't I'm not I don't know tell
3:31:05 me why you're
3:31:08 asking night
3:31:10 nomination
3:31:12 what kind of AI jobs are available all
3:31:14 jobs are going to be impacted by AI so I
3:31:16 would I would say it's not so much that
3:31:18 so if you if you want to go work for
3:31:21 companies that are doing AI like go
3:31:23 start you know learning about these
3:31:27 companies that are behind all this [ __ ]
3:31:28 we're using um but but if you're talking
3:31:31 about like what what jobs are out there
3:31:33 in all the businesses that are going to
3:31:35 be using AI all the businesses are going
3:31:37 to be using AI so what I would say is
3:31:40 what are you good at and what are you
3:31:42 passionate about start learning Ai and
3:31:45 then start how to do start to teach
3:31:47 yourself how to do what you do Amplified
3:31:51 by
3:31:52 Ai and start talking about it and you'll
3:31:55 you you will jobs will find you for the
3:31:57 next year year and a
3:32:00 half anyone who gets AI
3:32:03 literate in in their expertise is going
3:32:07 to be in good shape anyone who resists
3:32:09 it is going to be [ __ ] so um while
3:32:13 there may be like there might be jobs
3:32:15 like prompt engineer and we need an AI
3:32:17 this and an AI
3:32:18 that it's not going to be that obvious
3:32:21 short term because what's going to
3:32:22 happen is people are just going to be
3:32:24 like I need to learn like this I I I was
3:32:28 out to dinner with my my you know work
3:32:32 my CEO buddy and he told me that he saw
3:32:35 30% productivity gains cuz he started
3:32:38 using AI I got to figure this stuff out
3:32:40 how do I use AI for my dut shop if
3:32:43 someone in the dut shop is like always
3:32:46 talking about how they use chat GPT and
3:32:48 a AI when that CEO goes hey uh uh wait
3:32:53 what's your name hey dney uh dney you've
3:32:55 been talking about that chat GPT thing
3:32:57 do you know anything about this AI I
3:32:59 know a little bit about it good you're
3:33:01 going to be the new VP of AI like that
3:33:04 shit's going to
3:33:05 happen um so start learning how to do
3:33:09 what you do um
3:33:12 using these tools and
3:33:14 then it will become increasingly obvious
3:33:17 where you might be good in this new
3:33:19 world wrote Pharmacy procedures in
3:33:22 minutes oh that's fascinating that's
3:33:24 really cool use it for VBA programming
3:33:27 and Excel formula building it's
3:33:28 brilliant at that [ __ ] cuz it's so well
3:33:30 documented all that VBA stuff and the
3:33:32 Excel formula stuff there's just there's
3:33:34 like an infinite number of courses on it
3:33:37 so in the in the Corpus for the the base
3:33:40 large l language model it's just full of
3:33:43 that [ __ ] like you know those formulas
3:33:45 have been around for decades right so
3:33:47 it's good at that [ __ ] that's
3:33:50 awesome all right someone's going to
3:33:52 have to edit every spur
3:33:57 work I don't know what spur workor was
3:33:59 supposed to be but someone will have to
3:34:02 edit
3:34:03 something all right let me do a few more
3:34:05 questions and then I got to go I really
3:34:06 got to go this time seriously stop
3:34:09 forcing me to be here it is not
3:34:11 appropriate it's not I'm I am
3:34:14 triggered I'm triggered looks like we
3:34:17 got some new folks in hey everybody my
3:34:18 name is Kyle Shannon this is the AI
3:34:21 learning lab and apparently I'm a bitter
3:34:22 old gen xer don't you call me Boomer
3:34:26 don't don't you do it [ __ ]
3:34:29 Millennials
3:34:32 [Laughter]
3:34:36 okay my AI movie will be called the
3:34:38 matriarchy showing the history of the
3:34:40 world if women were we're in charge
3:34:41 that'll be a good one it'll be awesome
3:34:44 do it love it you know visions of all
3:34:47 that comes later what sight is he on
3:34:50 what drug is he on it's called caffeine
3:34:53 and add M
3:34:58 tasty bedtime reminder Apple has now
3:35:02 told me it is time for me to leave
3:35:05 [Laughter]
3:35:09 you ah
3:35:12 is it getting better me no this is as
3:35:15 good as it gets it only goes it only
3:35:17 goes downhill from
3:35:20 here um used to take beating up an
3:35:23 analyst now it's just a conversation yep
3:35:26 love that real estate analysis
3:35:29 cool how do you get it to get GID mode
3:35:35 GMO G git mode I assume you mean git as
3:35:38 in
3:35:39 GitHub um
3:35:42 you can just copy and paste a readme
3:35:44 file from a GitHub and have it give you
3:35:47 step-by-step instructions for how to get
3:35:49 it up and running um somewhere I read
3:35:52 that someone exported an entire git
3:35:55 repository and uploaded it into um think
3:36:00 code interpreter and just started doing
3:36:02 [ __ ] with it so that again the depth of
3:36:06 code interpreter is significantly beyond
3:36:09 my ability to even imagine it so as I
3:36:12 learn more about that I will share what
3:36:14 I learned but I uh code interpreter is
3:36:18 one of those things that
3:36:22 is is newly
3:36:24 humbling just when you think you had
3:36:26 this [ __ ] figured out ping here's a
3:36:30 thing to make you feel
3:36:33 stupid da X Y and Z Trend it analyze it
3:36:37 some okay for my AI bought to be my
3:36:39 friend it has to be private and trusted
3:36:42 yeah Louis Scott that's that's a that's
3:36:44 going to be a big deal trusted private
3:36:47 data is going to be a big deal R2-D2 is
3:36:50 100% loyal the other thing you're
3:36:51 talking to is a thing called alignment
3:36:54 and and Alignment is
3:36:57 basically aligning the goals of these
3:37:01 things with the goals of these
3:37:04 things right um Sam Alman the CEO of of
3:37:09 open AI um said that they're dedicating
3:37:12 20% of their compute power which is a
3:37:14 lot 20% of their resources are going to
3:37:17 what's he's he's calling super alignment
3:37:19 which is basically we want to make sure
3:37:22 that the Bots have the the same
3:37:26 imperatives as the
3:37:28 humans um that they're in
3:37:31 alignment so that these things just
3:37:33 don't go you know we'd be a lot more
3:37:35 successful if we got rid of these pesky
3:37:37 meat bags that's the thing we don't want
3:37:41 because they're already verging on being
3:37:44 significantly smarter than us and
3:37:46 probably within a
3:37:50 year no probably within 3
3:37:54 years the capability is going to be so
3:37:56 beyond any human that if we haven't
3:37:58 worked really hard on alignment that's
3:38:00 an actual risk maybe five
3:38:04 years thank you for your picture upload
3:38:06 demo good I didn't know which one you
3:38:08 wanted I hope that helped that's nice I
3:38:10 want be there oh the the VW van I know
3:38:13 that picture made me very happy when I
3:38:14 made it that picture was a complete
3:38:16 accident I was showing someone on the
3:38:18 live someone asked about the describe
3:38:21 feature in mid Journey so you can upload
3:38:23 an image in mid Journey or or you type
3:38:26 slash describe and then it it gives you
3:38:28 an upload button and it will describe it
3:38:30 will write prompts for your image and
3:38:34 then you can generate new images from
3:38:36 The Prompt so I uploaded a
3:38:39 papercraft Cru ship on an ocean at
3:38:43 Sunset and I had it rewrite prompts and
3:38:46 then I had it generate images and one of
3:38:48 the images it generated was the VW bus
3:38:51 and the prompt didn't say VW bus it was
3:38:53 just weird it was just a weird anomaly
3:38:55 that it came up but it was so [ __ ]
3:38:58 beautiful the avocado Universe
3:39:02 intervened go get some wine and come
3:39:05 back I need to go to bed just get some
3:39:08 wine just drown those sorrows you're
3:39:10 right
3:39:14 though how to make money with chat gbt
3:39:17 um like this you can make
3:39:20 money hope that helped all right use
3:39:22 Google Scholar for journals for school
3:39:24 nice passive income yes yep yes how do I
3:39:30 learn to
3:39:31 add AI to my
3:39:36 CV come join the AI
3:39:39 salon
3:39:41 and just say member AI
3:39:44 Salon there's no
3:39:46 fee we're just a bunch of people trying
3:39:48 to figure this [ __ ] out show up to a
3:39:51 meeting contributing me member of the AI
3:39:54 Salon there's a
3:39:56 way I got
3:39:59 um I work for an organization called
3:40:01 content Evolution I don't work for them
3:40:03 it's it's not a paying gig or a paying
3:40:07 organization but
3:40:09 um it's it's a it's like a Federation of
3:40:12 marketing companies that all kind of
3:40:13 co-market and support one another
3:40:15 basically it's like a professional trade
3:40:17 group so I made a Content generator for
3:40:21 them if you go to content evolution. net
3:40:23 it's called The Challenge engine so it's
3:40:26 basically a zappier automation that
3:40:28 generates some cool content and so the
3:40:31 guy that runs it said hey I want to give
3:40:33 you a title how about Chief generative
3:40:35 officer I'm like cool he said I'm pretty
3:40:38 sure you'll be the first one in the
3:40:39 world to have that title I'm like
3:40:41 awesome it's completely meaningless but
3:40:43 it's front and center of my resume Chief
3:40:45 generative officer content
3:40:49 Evolution I don't know make [ __ ] up
3:40:52 everyone's making this [ __ ] up right now
3:40:54 nobody knows what the [ __ ] they're
3:40:57 doing the like these tools are literally
3:41:00 being invented while we're using them
3:41:03 like this list used to have three things
3:41:05 in it now it has five and I could argue
3:41:07 that there should be six maybe seven and
3:41:10 probably within a month there will be
3:41:13 10 10 what 10 major competitors to
3:41:20 gp4 how do you use them I don't know how
3:41:23 do you use code interpreter I don't know
3:41:26 what's it going to do I don't
3:41:28 know is it going to shake some [ __ ] up
3:41:32 uh-huh in dramatic and powerful and
3:41:35 profound ways what's that going to look
3:41:37 like I don't know but you're the chief
3:41:40 generative officer exactly no one's an
3:41:43 expert right now anyone that tells you
3:41:45 they're an expert it's full of [ __ ] so
3:41:47 if you know how to spell AI put it on
3:41:49 your
3:41:50 CV and by the
3:41:53 way copy and paste the contents of your
3:41:56 CV paste it into chat GPT write some
3:42:00 bullet points hey by the way in the last
3:42:02 three months I've done the following
3:42:04 things with AI bullet point bullet point
3:42:06 bullet point 6 10 whatever it is doesn't
3:42:09 matter two please update my CV to
3:42:13 reflect the fact that I've got
3:42:15 experience in generative Ai and I want a
3:42:17 job doing that and it will rewrite your
3:42:20 CV for you and automatically incorporate
3:42:22 that [ __ ] into it it'll do it in 30
3:42:25 seconds and then when you pick your jaw
3:42:27 up off the floor upload that puppy to
3:42:30 LinkedIn change your title to Chief
3:42:32 generative officer of
3:42:35 bucus who's like you think they're going
3:42:37 to [ __ ]
3:42:39 check do you have a master's degree in
3:42:42 generative AI from Harvard no no one
3:42:46 does you know how you can get a degree
3:42:48 in like web design and [ __ ] now when I
3:42:51 started in
3:42:53 1994 nobody knew what the [ __ ] it was
3:42:56 they didn't have classes in web
3:42:59 design you had to make that [ __ ] up
3:43:02 that's where we are right now so make it
3:43:05 up no one's going to
3:43:08 check where can I find the best CV
3:43:12 language right
3:43:15 there tell it to be a
3:43:19 master
3:43:22 um uh like like employment
3:43:26 advisor
3:43:28 and ask it what are the different um
3:43:33 what are the most successful strategies
3:43:36 for resume creation and let it tell you
3:43:38 the strategies and then say here's
3:43:40 here's my resume can you improve my
3:43:43 resume for this kind of job
3:43:46 incorporating this kind of experience
3:43:48 with
3:43:50 AI um using one of the strategies you
3:43:53 know the the the strategy above you
3:43:55 think best fits my
3:43:57 resume out it'll come it's just it's
3:44:01 right there that's why I sit with these
3:44:04 next to my [ __ ] head all night long I
3:44:07 don't work for any of these guys they
3:44:09 don't pay me to point you to
3:44:13 them all right I am thinking oh [ __ ]
3:44:17 Tick Tock the [ __ ] scrolling bug just
3:44:19 happened
3:44:20 people I thought someone here said they
3:44:23 were going to fix this and tell Tik Tok
3:44:24 to update their scrolling code don't we
3:44:27 have chat GPT for that I'm thinking
3:44:30 about using AI to scan my jobs
3:44:34 Sops sop statement of something and
3:44:38 asking it questions yes you should do
3:44:39 that um although just just be careful um
3:44:45 there's risk of like I would turn off
3:44:48 your browsing your your prompting
3:44:50 history in chat GPT make sure you export
3:44:53 your data first so you get your prompt
3:44:54 history but i' I turn it off and then
3:44:56 request that they don't use
3:44:59 it we're hanging out braah peace brah
3:45:03 all right you're our AI slave I am can
3:45:07 you mansplain that that's what this
3:45:08 channel is
3:45:13 what are your qualifications Kyle I
3:45:16 ain't got none but I'll still talk like
3:45:19 I
3:45:26 do problem with AI it's using a lot of
3:45:29 copywritten data to build its data set
3:45:31 well it's been trained on a lot of
3:45:33 copywritten
3:45:34 data
3:45:37 um is it a problem m
3:45:41 it is to the copyright holders right now
3:45:44 but most of them don't understand how it
3:45:45 works and I think long
3:45:47 term long term if you play out how to
3:45:51 best train these you train them on as
3:45:52 much data as possible and what is that
3:45:54 that's all of the data you train them on
3:45:57 as much as you can get all of the
3:45:58 copywritten all of all of the data all
3:46:00 of the
3:46:01 data because ultimately this is the
3:46:04 library right this is the knowledge
3:46:06 Library this is the the the collective
3:46:09 out put of humanity is the ultimate data
3:46:13 set that we want in these
3:46:16 things so just like a copyright holder
3:46:20 could be pissed that their book can be
3:46:22 read for free in a
3:46:24 library there's a greater
3:46:28 humanitarian purpose that that Library
3:46:31 serves same with this [ __ ] it's just
3:46:34 mixed up right now because business
3:46:37 models are going to get [ __ ] up and
3:46:40 there's companies that are going to
3:46:41 profit from that copywritten
3:46:45 data
3:46:47 and the more data that's in here the
3:46:50 better that these tools are for all of
3:46:52 us so it's both right there are
3:46:56 copyright issues there are trademark
3:46:58 issues there are intellectual property
3:47:00 issues
3:47:02 and these tools are better when they
3:47:04 have that [ __ ] in
3:47:06 there now how we deal with attribution
3:47:08 and things like that I don't know
3:47:10 and and now that you have open- Source
3:47:13 models now that meta has open- sourced
3:47:17 their models you got a whole open source
3:47:20 Community that's also built on the same
3:47:22 copywritten data but there's no one to
3:47:24 sue there's they're open source so this
3:47:26 shit's out there so
3:47:30 so there's there's a lot of rearranging
3:47:32 the deck chairs on the Titanic right
3:47:36 now because the long game here is is
3:47:40 that all the information is going to be
3:47:41 in these things and we're all going to
3:47:43 have access to
3:47:44 it over the next decade we'll figure out
3:47:47 ways just like we did with streaming
3:47:49 we'll figure out ways to monetize
3:47:52 it but just like
3:47:55 streaming the benefit to the artist the
3:47:58 benefit to the intellectual property
3:48:00 Creator gets lower and lower and lower
3:48:04 in a world
3:48:05 where you can create an infinite amount
3:48:09 of high quality quity
3:48:12 content I don't know the solution to
3:48:17 that if we're moving to a world of hyper
3:48:20 personalized
3:48:21 content where I can watch a movie
3:48:24 starring whatever star that I want with
3:48:26 whatever plot that I want in whatever
3:48:28 genre that I want just for me to
3:48:34 watch is that valuable to anyone else I
3:48:36 don't know it might
3:48:38 be how does that get monetized I don't
3:48:42 know do you pay the company that
3:48:44 dynamically generates the movie do they
3:48:47 have deals with all the likeness
3:48:49 agencies of all the actors that might be
3:48:51 included in my movie I don't know like I
3:48:53 don't know how all of this soup of
3:48:57 intellectual property gets
3:49:02 attributed um but it's it's just all
3:49:05 going to change dramatically so I don't
3:49:07 know so it's a problem but it's also so
3:49:10 a is mid Journey free noes 10 bucks a
3:49:13 month is the smallest subscription
3:49:15 that's what I pay I pay the 10 bucks a
3:49:17 month I don't use it enough to exceed
3:49:19 that um and also
3:49:38 leonardo.da 3.0 Leonardo Leonardo a is
3:49:42 really good they give you 150 credits a
3:49:46 day for free and that's enough to do
3:49:49 like I don't know 15 or 20 images
3:49:50 something like that
3:49:53 so it's
3:49:55 good please do a video on how to use
3:49:57 Discord in mid journey I find it nearly
3:49:59 impossible I just did that earlier
3:50:01 hopefully you were there for that I'm so
3:50:03 sad when my comments are gray I don't
3:50:05 know what that means why are your
3:50:06 comments
3:50:08 Gray so so Salon Chief stinky Wrangler
3:50:14 exactly AI Salon Chief oh Slinky
3:50:17 Wrangler yeah
3:50:22 exactly it was the best of time it was
3:50:25 the best of times it was the worst of
3:50:27 times Kyle probably saying how he has to
3:50:32 leave I got to go people hasn't he said
3:50:35 that like 30 times I don't think this is
3:50:38 live is he in AI
3:50:41 are you silent disregard I can hear you
3:50:45 I was probably thinking for a moment and
3:50:48 occasionally my brain just shuts down
3:50:50 what's your opinion on building a
3:50:51 startup in public um I think right
3:50:56 now
3:51:00 um I think Founders tend to be way way
3:51:04 too precious about their ideas I would
3:51:06 say put it out there build in public um
3:51:10 because you'll get customers and you'll
3:51:12 get feedback faster and you'll you'll
3:51:14 get reality checks
3:51:16 faster
3:51:18 um I would also build with the fullon
3:51:22 fullon expectation that whatever you
3:51:25 build right now is going to be useless
3:51:28 in 6 months if you're building stuff
3:51:30 based on current AI
3:51:32 technology
3:51:34 um you also we we're also
3:51:38 building we're stepping into an
3:51:41 environment we're not quite there but
3:51:42 we're really [ __ ]
3:51:46 close we're moving into an environment
3:51:48 where it's going to
3:51:50 be trivial to completely copy a
3:51:56 business from the positioning to the
3:52:00 code to the marketing strategy all of
3:52:04 it and
3:52:06 so I don't know how
3:52:10 you maintain a competitive hold on your
3:52:13 customers in that
3:52:15 environment shy of you [ __ ] Delight
3:52:18 your customers right so so but it's
3:52:21 going to be really easy to knock off
3:52:22 things so so I would say just put it out
3:52:25 there start building it
3:52:28 and take the risk of someone stealing it
3:52:31 the idea of a business I've I've given
3:52:35 away a dozen of them this week the the
3:52:38 idea is if effectively worthless right
3:52:41 it's only in the execution of the idea
3:52:43 that the business gains any potential
3:52:46 value so if you're a good
3:52:49 executor people might knock you
3:52:52 off almost no one's a good executor
3:52:55 execution is really hard so i' I'd go
3:52:58 for it and the other thing
3:53:01 is by building it in public especially
3:53:04 if you're doing stuff with this AI stuff
3:53:06 everyone wants to know what the [ __ ] is
3:53:08 going on with this stuff so you'll get
3:53:12 opportunities for the business and
3:53:14 outside of the business just because of
3:53:16 building it in public so yeah the more I
3:53:18 talk about
3:53:19 it I would do that congrats Chief AI
3:53:23 generative officer just Chief generative
3:53:25 officer I took the AI out of it because
3:53:29 um generative could mean generating
3:53:32 Revenue generating
3:53:35 engagement generative AI so Chief
3:53:38 generative
3:53:40 officer doesn't really matter what I'm
3:53:43 generating um but it's tied enough to it
3:53:46 that I can say well yeah of course
3:53:47 that's AI so I like that I like it I
3:53:50 like it oh chat jpt I mean Tik Tok
3:53:53 jumping me to the bottom
3:53:58 again intellectual property laws govern
3:54:02 is this ex esoteric or normal can't tell
3:54:05 if the smmes here or ransoms what Louis
3:54:10 Scott I don't understand your words
3:54:12 randoms are you random maybe I don't
3:54:16 know what that means am I a subject
3:54:18 matter expert or
3:54:20 random I'm a generalist for
3:54:23 sure I I may be a subject matter expert
3:54:26 well I'm certainly a subject matter
3:54:28 expert in the intersection of
3:54:30 Storytelling and Technology that's been
3:54:31 my whole
3:54:33 career
3:54:36 um I think I'm a subject matter expert
3:54:39 in in the early
3:54:40 adoption of
3:54:42 technology in service of
3:54:46 business so other than that [ __ ]
3:54:51 random but oh you're talking to each
3:54:53 other I'm a random Oracle for blockchain
3:54:55 application subject matter expert like
3:54:57 analyst Etc Louis Scott I don't know if
3:55:00 you're talking to me or someone
3:55:02 else um comments get gray as they age oh
3:55:06 I
3:55:07 see yeah I'm only reading gray
3:55:10 [Laughter]
3:55:13 comments is it 1:00 a.m. where you are
3:55:15 I'm in Denver so it's 11: where I
3:55:17 am okay you guys are just talking to
3:55:20 each other okay this just got serious oh
3:55:22 my God he's
3:55:23 back I don't know what that means I
3:55:25 haven't left I need to hear about
3:55:28 building apps I'm paranoid about
3:55:30 intellectual
3:55:32 property um again I I think people are
3:55:36 over precious about their ideas
3:55:39 personally unless you you know if you
3:55:41 got something patentable [ __ ] patent
3:55:43 it um everything else I don't know just
3:55:47 start building it start [ __ ] building
3:55:51 it I mean here's the thing nobody's got
3:55:54 the [ __ ] time everyone's trying to
3:55:56 figure this [ __ ] out everyone's
3:55:57 panicking
3:55:59 yay everyone's trying to figure this
3:56:01 stuff out so it's like the amount of
3:56:02 people that have the time and energy to
3:56:05 seek out other people's ideas to steal
3:56:07 them is very very low
3:56:10 if you generate something that's good
3:56:12 and starts to get a lot of Buzz yeah
3:56:15 then people are going to steal it but in
3:56:17 the early stages I don't know start
3:56:22 building what was that 10 minutes I know
3:56:24 I said I was going to leave 10 minutes
3:56:25 ago and that was 50 minutes ago was that
3:56:27 10 minutes of Frozen the AI told you to
3:56:31 do it all right I'm going to go now I'm
3:56:32 telling you in the [ __ ] chat I'm not
3:56:34 even
3:56:35 listening ug please share what you would
3:56:39 do to make sure a sixth grader knew what
3:56:41 they need
3:56:42 to
3:56:44 um they will steal from you who's they
3:56:47 dirt bag
3:56:48 biker who's going to steal from you if
3:56:51 you put an idea are you talking about
3:56:54 putting an idea out there who's going to
3:56:56 steal from you who's
3:56:58 they I'm joking about the agent
3:57:01 comments he's leaving because of your
3:57:04 comments
3:57:05 exactly um what would I what would I
3:57:08 teach a six grader I
3:57:11 would I would just get them
3:57:14 curious what are they interested
3:57:18 in if they're interested like you can
3:57:22 make them pictures of [ __ ] they're
3:57:23 interested in or teach them to make
3:57:25 pictures of stuff they're interested in
3:57:27 you can teach them to use these tools to
3:57:29 write stories to answer questions sixth
3:57:33 grader they're starting to be curious
3:57:34 about how the world Works how stuff
3:57:36 works if they're curious like I would
3:57:38 just feed their imagination I would use
3:57:41 these
3:57:44 tools I would get into a dialogue with
3:57:47 them about what are they most interested
3:57:49 in what do they want to learn what do
3:57:51 they like learning what do they like
3:57:53 doing I mean you probably already know a
3:57:55 lot of this when they play when they're
3:57:57 not doing their homework what are they
3:57:59 doing is it video
3:58:01 games have them did did you see the
3:58:04 thing that I did before with
3:58:08 uh
3:58:09 um Skybox what's it
3:58:15 called blockade
3:58:18 Labs go to Skybox blad labs.com
3:58:24 and let's see
3:58:29 um world of Hot Wheels
3:58:35 tracks
3:58:38 generate this is going to generate a 3D
3:58:40 world like
3:58:41 like like tap into what they're
3:58:44 interested in tap
3:58:46 into the things that turn them on and
3:58:50 then just reveal the magic of whichever
3:58:52 of these AI
3:58:56 tools can feed that
3:59:03 imagination let's see so like here's
3:59:06 this Hot Wheel track like I just
3:59:07 invented this
3:59:10 so like if they're into Hot Wheels
3:59:12 here's a Hot Wheels
3:59:13 World Skybox blad labs.com show them
3:59:17 stuff that's Visual and interesting
3:59:19 first but then get get their imagin
3:59:23 imagination engaged in whatever their
3:59:26 whatever they love where's their passion
3:59:29 tap into that don't judge it just feed
3:59:33 it and and just kind of get them
3:59:36 interested in this stuff and then turn
3:59:37 them loose
3:59:40 I don't know I think that's what I'd do
3:59:42 because the most
3:59:45 important the most important skill when
3:59:49 these tools do all the Tactical
3:59:50 execution is creative
3:59:52 thinking so how do you feed their or
3:59:55 critical thinking how do you feed their
3:59:57 critical thinking and their
4:00:00 creativity because these tools will do
4:00:02 all the
4:00:06 rest I wonder how this would help us
4:00:08 autistic kids I think it's going to be
4:00:10 huge for neuro divergent kids um I think
4:00:13 it's going to be huge for it um because
4:00:16 it can communicate in incredibly
4:00:17 empathetic ways incredibly patient
4:00:20 ways worst AI business Concepts Concepts
4:00:24 please give me some you can make money
4:00:26 with
4:00:29 Chach I don't know I think starting a
4:00:32 prompt library is going to be a losing
4:00:37 business um
4:00:42 um even though even though I'm building
4:00:45 a Content generation engine is a big
4:00:47 part of storyvine my company um I'm
4:00:50 putting a little twist on it I think
4:00:52 content and generation
4:00:54 tools are going to be bad businesses
4:00:57 just because there's going to be so many
4:00:59 of them they're so easy to
4:01:01 build um I have seen
4:01:05 672 children's book maker apps so
4:01:10 automatically generating children's book
4:01:12 complete with illustrations I think
4:01:14 there's going to be just an infinite
4:01:15 number of those I think that's going to
4:01:17 be a shitty
4:01:19 business
4:01:23 um I don't
4:01:25 know I think any business
4:01:29 that puts generative AI as as
4:01:33 a significant part of the value
4:01:36 proposition is a bad business
4:01:40 right if you say you
4:01:43 know this is this magic AI tool if you
4:01:47 make it all about the AI That's a losing
4:01:48 business make it about the problem
4:01:50 you're solving and it just happens to
4:01:52 solve the problem in a magical way
4:01:54 because of
4:01:56 AI so I think anything that leans too
4:01:58 heavy on the AI thing is a bad business
4:02:01 anyway all right wait did you see
4:02:03 someone was being sued
4:02:05 about the way they were portrayed on
4:02:08 chat gbt maybe I saw Sarah Silverman is
4:02:12 suing open AI cuz it gave one of her
4:02:16 jokes verbatim from her book yeah I saw
4:02:19 that chat was being sued I think I think
4:02:22 there are many many conversations with
4:02:25 uh lawyers going on at open AI right now
4:02:28 they just cut a deal with Shutterstock
4:02:31 they just cut a deal with Associated
4:02:33 Press um I think they're going to start
4:02:36 cutting a lot of deals we'll see anyway
4:02:39 my voice is horse my ass is numb this
4:02:42 has been fantastic thank you for hanging
4:02:43 out on a Saturday night you really
4:02:45 shouldn't do this but thanks I'm glad
4:02:48 you did hope you had fun I had fun all
4:02:51 right always remember you can make money
4:02:57 with peace out
4:03:01 everybody that's awesome