aiLL July 25, 2023 - The AI Learning Lab: Cybernetic Friends and the Good Life

Description

Chapters

Transcript

0:07 foreign
0:07 [Music]
0:35 welcome to the AI learning lab my name
0:38 is Kyle Shannon
0:41 let's get this show on the road I don't
0:43 know what day it is it's Tuesday right
0:45 Tuesday
0:48 not Tuesday Tuesday
0:55 welcome welcome welcome
0:59 welcome to the AI learning lab where we
1:01 learn about labs
1:06 hmm doesn't seem right does it
1:10 um
1:12 that's a nice ice Java for the evening
1:18 but Kyle won't take you up at night I
1:21 don't know maybe maybe it will maybe it
1:23 won't maybe I'll sleep like a little
1:25 baby all right welcome everybody
1:28 my name is Kyle Shannon this is the AI
1:29 learning lab
1:31 a little bit about my backgrounds hello
1:33 Kevin
1:35 Earthling
1:36 or are you are you the alien are you the
1:40 the far out one
1:42 we're the we're the earthlings is that
1:44 how it goes hmm I see
1:48 um so welcome welcome welcome
1:51 on this here Channel we hang out we have
1:53 fun
1:55 I make jokes I make funny voices I make
1:58 funny faces but I'm starting with this
2:00 so you know
2:02 and uh we talk about AI I'm a long time
2:06 entrepreneur
2:07 I've started 12 or 15 companies in my
2:09 day I was a uh an entrepreneur in the
2:12 early days of the World Wide Web I
2:14 started one of the first digital
2:16 agencies company called agency.com and I
2:18 was Chief creative officer there and so
2:20 I have been
2:21 through one of these Cycles before and I
2:25 didn't think I'd see another one in my
2:26 lifetime because that one was pretty
2:28 remarkable the early days of the web
2:30 riding up to the.com bubble and the
2:33 burst and all that happy horse [ __ ] and
2:36 uh then Along Comes This Here generative
2:39 AI stuff I kind of had a false start
2:41 with blockchain and nfts where I'm like
2:43 this is gonna be big and
2:45 it was funny the deeper I dug into
2:47 blockchain and nfts the less real it
2:49 seemed I like the tech just the
2:51 companies and the project sucked
2:53 but the more the deeper I dig into this
2:55 generative AI stuff the more real it
2:57 seems so this is this is uh this is a
2:59 big deal it's transformative
3:02 um
3:03 and uh yeah so I'm happy to talk about
3:05 it so if you have questions about
3:07 um
3:08 whatever AI pop them in the comments
3:10 below I do my best to answer everything
3:12 let me show you what I got going on in
3:15 this little window
3:18 [Laughter]
3:19 um
3:30 all right let's get that to focus shall
3:32 we
3:33 can we
3:36 there we go all right grab a screenshot
3:39 of this so if you're new to generative
3:41 AI if you haven't played with chat GPT
3:44 these are your seven options this list
3:47 used to be three
3:48 and now it's seven and that was a week
3:51 ago it was three and for like two months
3:53 there was three things on here and now
3:55 there's seven why because shit's
3:57 happening so
3:59 and I'll I'll give you a quick rundown
4:01 on these so if you've been here before
4:03 you've heard this before so you can you
4:05 know take a little nappy all right
4:08 chat.openai.com that's the official chat
4:10 GPT website if you've not played with it
4:12 if you've not played with the original
4:13 the one the only the original Chachi BT
4:16 that's it right there go there and just
4:18 start playing
4:20 bing.com is also chat GPT it's gpt4 but
4:25 it's built into Microsoft search engine
4:27 Bing so you click on the chat button
4:29 that's gpt4 you can generate images it
4:31 can analyze images or read them like you
4:34 could upload a picture and have it write
4:35 advertising copy for the picture and
4:37 it'll just do it uh it's connected to
4:39 the internet
4:41 um Chachi BT has some pretty cool things
4:43 if you pay for the plus version which
4:46 I'm not connected with any of these I
4:47 don't make commissions on any of this
4:49 [ __ ] so don't get all cynical you have
4:52 [ __ ] trolls all right
4:54 claude.ai the one on the other side
4:56 there I would say is the biggest
4:59 competitor to to chat GPT right now
5:02 um it's a bunch of X open AI employees
5:05 that invented gpt3 and uh it's got a
5:09 hundred thousand token context window
5:11 which basically means you can upload
5:14 like 75 000 Word documents and it'll
5:17 just read it and then you can interact
5:19 with it
5:20 um bard.com is Google's entry which Bard
5:24 was laughably bad for months and months
5:27 and months and then last week it got
5:29 good turns out they replaced the core
5:31 large language model with this thing
5:32 called Gemini that's also also
5:35 multimodal it can read images
5:37 then
5:39 labs.perplexity.ai that's metas llama2
5:42 model that you can just play with at
5:43 perplexity they just put it up there so
5:45 everyone can go play
5:47 Pi dot AI they have a cool app that you
5:49 can talk to it's very conversational
5:51 pose got a lot of bots and you can play
5:53 with lots of these different models so
5:55 there's a lot here they're they're all
5:57 different flavors of the same thing so
5:58 if you don't know at all what you're
5:59 doing ignore all the [ __ ] on the bottom
6:01 just go to the top one
6:03 if you want to start exploring the
6:05 different flavors of these models and
6:06 see what's good for certain things
6:09 that's your list so screenshot that okay
6:11 now screenshot this too
6:13 can we see everything yeah
6:16 so the thing at the top prompts.chat's
6:18 just a document that'll teach you about
6:19 prompting how to talk to those large
6:21 language models we just talked about
6:23 openart.ai promptbook that's going to
6:27 give you instructions for how to deal
6:29 with mid-journey stable diffusion Adobe
6:32 Firefly Leonardo this sort of teaches
6:34 you about these text to image models
6:38 like how [ __ ] powerful they are then
6:40 you got futurepedia.io that's like four
6:43 thousand AI tools if you just want to
6:44 geek out go there and then these are the
6:47 image tools and then at the bottom left
6:49 is runwayml that's video and blockade
6:52 labs skybox.blockade labs is just a fun
6:55 parlor trick right now you describe a
6:58 scene and then you're in this 3D world
6:59 it's pretty slick it's not super
7:02 sophisticated but it's slick slick slick
7:05 slick slick
7:08 and then finally uh this is I'm gonna
7:12 get rid of this thing because I fixed
7:13 the
7:15 I fixed the URL so the salon the AI
7:18 Salon is a community that I started last
7:21 December uh that's a community of people
7:24 on this AI adventure together and so if
7:27 you want to hang out with people there
7:30 just go to that thing there's a link
7:31 tree we have meetups we meet every other
7:33 week we've got a Discord server we have
7:35 a LinkedIn group storyline.com that's my
7:37 current company and that's my my Twitter
7:39 my ex Twitter handle my current X handle
7:45 Elon isn't that funny he
7:48 changes the comp
7:52 a bunch of trademarks on X because of
7:55 the Xbox who knew well apparently the
7:58 trademark office
8:00 I guess that cage Max just the cage
8:03 match between those two the stakes just
8:05 went up
8:07 it's pretty funny
8:10 all right
8:12 so hop
8:15 hop your uh
8:18 Auto GPT it's dead I don't think so
8:21 isn't it just an open source project
8:24 let's go see did it die did it die I'll
8:26 show you that thing too look at this
8:28 cool video here
8:30 see that video that started out as a
8:32 panoramic photograph that I took
8:35 that's at Runway ml that I had down
8:37 below there but what are you saying Auto
8:40 GPT dead Auto GPT dead
8:43 question mark
8:45 um
8:48 I don't think so
8:52 no
8:53 I don't think so I mean maybe it is if
8:55 you know anything post it in there Auto
8:58 GPT is one of those autonomous agents I
9:01 think those are going to be a big deal
9:03 um all of those things Auto GPT baby AGI
9:07 the initial things that came came out
9:09 were super promising but they were also
9:11 it's also early like this stuff's it's
9:14 very early for this stuff so
9:16 all right let's get to the questions
9:18 let's go to the video team oh I can't
9:21 read with this shirt I think I'm gonna
9:22 have to go change my shirt so is AI a
9:25 bubble certain to pop as well no
9:29 no
9:30 no when you have a piece of technology
9:34 that produces
9:37 easily 30 to 40 percent productivity
9:40 increases overnight
9:43 that ain't going away
9:45 the fact that there's seven competitors
9:47 on this list right now when a week ago
9:50 there were ostensibly three
9:53 means that this isn't going away and and
9:55 there's a there's a whole bunch of
9:56 reasons for it so
10:00 Google invents this thing in 2017 called
10:02 the Transformer which allows companies
10:05 to train machine learning models on
10:08 massive data sets open AI from the
10:11 beginning
10:12 said why don't we train it on as much of
10:16 the internet as we can get to all of the
10:18 public internet hey the bear what's
10:20 happening thank you for the Roses very
10:22 sweet I gotta go change my shirt I can't
10:25 read any [ __ ] on it um so they invent
10:28 this thing called the Transformer and so
10:31 and that was in 2017. so for the past
10:34 what are we in 2023 now six years
10:37 so for the past five years
10:40 they've been getting better and better
10:43 and better at this they did GPT one and
10:45 they did gpt2 then they did gpt3 that
10:48 one made some noise
10:51 and then right around the gpt2 to GPT 3
10:54 time is when these companies like
10:56 stability Ai and mid journey start doing
10:59 some of this image Generation stuff
11:03 and that goes from like laughably bad to
11:06 like oh that's pretty good except the
11:08 hands all look like this
11:10 and then now all of a sudden the hands
11:12 kind of look okay and now mid-journey
11:16 images are photographic quality in fact
11:18 what was it was it Vogue I think it was
11:21 Vogue just did their first AI generated
11:23 cover
11:27 um then November 30th 2022 a day that
11:32 will go down and infamy
11:34 chat GPT launches now
11:37 from a technical perspective chat GPT
11:41 is just a very simple
11:43 application layer sitting on top of this
11:46 large language model down beneath in
11:48 fact the instructions for how to create
11:51 chat GPT
11:53 we're in the documentation for gpt3
11:56 in the documentation were instructions
11:59 for how to create this thing and no one
12:02 created it Sam Altman figured after a
12:04 year year and a half someone would have
12:06 built it
12:07 nobody built it
12:09 so so Microsoft's getting ready to make
12:12 their announcement about
12:14 their billion dollar investment into
12:17 open Ai and Sam Sam Altman says to his
12:20 Tech Team hey could you just go whip up
12:22 that chat GPT thing
12:24 two weeks they built it in two weeks
12:26 because it's really simple
12:29 two weeks two dose dos weakies
12:34 they launch it and six weeks later they
12:37 had a hundred million users
12:39 that changed everything
12:41 they didn't expect it to have 100
12:42 million users Microsoft didn't expect to
12:44 have 100 million users Google certainly
12:47 didn't expect to have 100 million users
12:50 so by adding a simple interface to deep
12:55 deeply powerful technology
12:58 this thing's revealed into the world
12:59 that just makes instant
13:02 um awareness of the tool and instant
13:05 panic in all of the competitors to get
13:08 their [ __ ] together
13:10 so metagos we're not going to talk about
13:12 the metaverse anymore we got AI stuff
13:14 look here's some free [ __ ] go play with
13:16 this Google's like we invented this
13:19 stuff we have all the AI we're good and
13:20 then they released a bunch of shitty
13:22 shitty AI
13:24 and then last week
13:26 they pulled out the good stuff
13:28 right so it took them eight months to
13:32 get it
13:33 to a usable level
13:36 um
13:37 Claude dot AI they raised 450 million
13:41 dollars
13:43 um pie
13:44 this cute little you can get an app that
13:46 you talk to that cute little chatbot pie
13:49 dot AI they just raised 1.3 billion
13:53 dollars
13:55 to make a 22 000 GPU
13:58 server intense thing so that thing that
14:01 looks really cute right now that's gonna
14:03 have some [ __ ] muscle behind it
14:08 um
14:09 Microsoft is rolling chat GPT throughout
14:12 their entire product line from cloud
14:15 services Azure to
14:18 um it's already in GitHub as co-pilot
14:20 it's going to be called co-pilot within
14:22 Office 365 within Dynamics 365 within
14:26 windows
14:28 so gpt4 is going to be built into
14:30 windows so here's the deal if you're a
14:32 Windows person I'm a Mac guy but you
14:34 know whatever
14:35 if you're a Windows person
14:38 you know how when you want to just like
14:41 change from like a light light setting
14:43 to a dark setting and then you like
14:45 Google it how do I change from a light
14:48 setting to a dark setting on Windows and
14:50 they tell you and then it was like six
14:51 versions of Windows ago and then like an
14:53 hour later you're still trying to find
14:55 the stupid [ __ ] switch in the 837
14:59 000 dialog boxes
15:01 that's gone
15:03 you'll just go change my settings to
15:05 dark
15:07 it'll do it
15:10 Google's Gonna Roll barred across all of
15:12 the office all the Google products
15:15 Google Docs Google Excel or Google
15:17 Sheets all of it
15:21 um
15:24 go go look at the YouTube video
15:26 barberheimer
15:28 someone did a mashup of Barbie and
15:30 Oppenheimer we played it last night it's
15:32 [ __ ] hilarious just go watch that all
15:35 AI generated in the hands of really
15:36 talented people good storytellers thank
15:39 you for the heart
15:42 so no it you know
15:44 is is AI over hyped yeah I think it's
15:47 over hyped but is it a bubble that's
15:48 going to go away like like are all of
15:51 our jobs safe no they're not
15:54 they're not
15:57 these are the most powerful tools ever
16:00 put in the hands of like normal
16:02 civilians
16:06 it's funny they're calling they're
16:08 calling this moment the the Oppenheimer
16:10 moment for for AI
16:12 because a lot of the you know these guys
16:14 have raised a lot of money and the tools
16:16 the next generation of these tools are
16:18 going to be powerful enough
16:20 that they're equating it with like the
16:22 importance of the bomb
16:25 but who had access to nuclear bombs
16:28 everyone no
16:31 people in the military
16:34 who's got access to this all of us
16:38 all of us and it's not just sitting
16:41 behind firewalls the the Llama 2 model
16:44 that meta just gave away put out an open
16:47 source with a commercial license
16:49 means that anyone can build anything
16:51 they want with these things without all
16:53 the safety guard rails that the other
16:54 guys are talking to Congress about
16:58 the the horse
17:01 done escape with the barn
17:05 okay
17:08 um
17:09 so yeah so it's not going away
17:12 um how quick is it coming I think it's
17:13 coming a [ __ ] ton faster than any of us
17:16 are prepared for
17:18 um
17:21 and and you know here's the other
17:24 thought I have on this is I think long
17:27 term if I go if I go kind of if I jump
17:29 ahead five years
17:32 the amount of new jobs that that these
17:34 things are going to create are probably
17:35 going to be insane the amount of
17:37 prosperity let's just call it that the
17:38 amount of prosperity
17:40 that that large language models and sort
17:42 of this flavor of technology is going to
17:44 create is going to be absolutely insane
17:47 the path to get there I think is going
17:50 to be really painful
17:51 there's going to be a lot of disruption
17:53 there's going to be a lot of jobs
17:54 there's going to be a lot of entire
17:55 sectors that just go away
18:01 I mean I'll tell you A big one
18:04 um customer service call centers
18:08 I think have a very very real
18:12 possibility
18:14 that they kind of disappear overnight
18:19 because if if if go play with pie go
18:23 play if you haven't played with pie yet
18:26 if you have an iPhone I mean you can do
18:28 it on the web but it's it's not as slick
18:30 an experience
18:31 go to pie
18:33 on your iPhone and put it into voice
18:35 mode and just start talking to it
18:39 and tell it to act like a uh
18:42 a customer service person for your
18:45 refrigerator like figure out your
18:47 refrigerator make and model
18:50 and I just talked to Pie as if it were a
18:52 customer service person and just start
18:55 having a conversation with it just like
18:57 you called an 800 number
19:00 so imagine a future say two years from
19:02 now
19:03 where when you call up Samsung about
19:05 your shitty refrigerator door that fell
19:08 off like mine did four months ago like
19:11 my wife opened the door the door
19:13 literally fell off onto the floor it was
19:15 like a year and a half old
19:17 mm-hmm
19:19 she was on the phone for hours on hold
19:23 got transferred to people this and that
19:26 and today and it's a disaster right
19:29 we've it's every all day every day if
19:31 you try to call someone you can so
19:33 imagine a future where you call Samsung
19:35 and they're like hey how can I help you
19:38 um I got a problem with my
19:40 thing I'm sorry to hear that how you
19:42 know what do you want me to do or like
19:44 describe the problem
19:46 and that's a synthesized voice just
19:48 having a conversation with you you
19:49 weren't put on hold
19:52 that voice
19:53 knows everything about Samsung
19:56 everything about it it knows where its
19:59 boundaries are and its boundaries are
20:01 way bigger than layer One customer
20:04 support layer two layer 3 layer four you
20:07 know you have to dial four different
20:09 people before you get to the human being
20:10 and and that's this excruciating
20:13 four-hour phone tree [ __ ] nightmare
20:17 press one to this let me put you on hold
20:19 oh did we cut you off sorry about that
20:21 start again
20:23 what if all that went away
20:29 what what happened look so like all of
20:34 the companies that program those phone
20:35 trees press one for this press two for
20:37 that
20:38 establa Espanol
20:42 all of those companies White
20:45 now can they pivot and do all sorts of
20:49 this [ __ ] yeah probably
20:51 but there's gonna be massive disruption
20:53 so
20:55 so I think it's coming and it's coming
20:57 fast
20:58 um we'll see we'll see
21:02 but I don't think it's a bubble I think
21:03 there's a possibility it's under hyped
21:08 all right I gotta go change my shirt I
21:09 can't read the text I will be back
21:44 ah much better much more better
21:48 welcome welcome welcome everybody
21:51 my name is Kyle Shannon this is the AI
21:53 learning lab
21:54 we are learning about AI well I'm just
21:56 pontificating right now let me get some
21:58 more questions
21:59 that list has gotten so long I know
22:01 right Jesse that list is insane and it's
22:04 like like at some point I'm gonna I'm
22:06 gonna thin it out but
22:08 like like the each thing has a different
22:11 strength like chat GPT has code
22:13 interpreter I wouldn't give that up
22:16 claude.ai has got a hundred thousand
22:18 token window I wouldn't give that up
22:21 Bing and Bard both do image recognition
22:24 and Bing's connected to the internet
22:26 so's Bard Bard's actually doing really
22:29 sophisticated
22:30 web link injections into
22:34 um answers
22:35 um and then perplexity that's that's the
22:38 Llama model Pi's really conversational
22:40 Poe you can make the cool box I feel
22:43 like everything there has at least one
22:45 feature that's worth exploring on them
22:47 so it's it's [ __ ] nuts
22:51 and this was this was in a week this
22:53 happened in a week
22:54 I love not GPT yeah I still listen G 's
22:58 gonna have a place in my heart oh by the
23:00 way did you see apple announced I told
23:02 you this was gonna happen I told you
23:04 they were being a little too quiet well
23:07 they're not quiet anymore
23:09 they've created a thing called Ajax
23:11 which is a AI development platform and
23:15 they've created what they're calling
23:16 internally at Apple Apple GPT
23:20 if they're going to come out swinging
23:24 because what does Apple have apple has
23:26 silicon that can run machine learning
23:28 models go watch their Vision Pro
23:31 demo notice that they didn't say the
23:33 phrase AI once
23:35 but they use things like machine
23:37 learning and embeddings and all the
23:39 little catch phrases
23:42 they're running [ __ ] locally with their
23:44 little R1 ship that's taking all that
23:45 sensor data and machine learning the
23:47 [ __ ] out of it in real time on the
23:49 device
23:51 so Apple's going to be able to
23:54 make a very strong privacy and security
23:58 claim
24:00 and their shit's gonna be good so when
24:02 their stuff comes out and it's going to
24:03 be on your data I guarantee you that
24:05 Apple's gonna make this I I think the
24:08 large language model stuff from Apple
24:09 the GPT stuff is going to be you know
24:13 your own personal
24:15 Kick-Ass robot assistant
24:20 um yeah
24:22 crazy AI bores me I'm old give me some
24:25 easy that'll inspire me
24:28 really Steve
24:30 it bores you
24:32 I think I fear you have not dug or
24:37 pushed
24:38 the harder you push on this stuff the
24:39 more impressive it is to me but I'm
24:41 curious what you've tried
24:43 and and it's quite frankly it's easy
24:45 actually you know it might be fun Steve
24:46 if if you've got an iPhone or even if
24:49 you don't just go to the website go to
24:51 Pi dot AI or haypi.com
24:54 go there
24:56 and then put it into voice mode so you
24:58 can talk to it and start talking to it
25:02 and just
25:03 talk for half an hour and ask it [ __ ]
25:06 anything
25:08 all right by the way pi.com Pi dot AI is
25:11 also connected to the internet did you
25:12 know that mid-journey
25:14 has an interesting bug it's probably got
25:17 a lot of them but tell me tell me what
25:18 the interesting bug is
25:20 a beautiful woman
25:22 select an old photo 40s 50s on the
25:24 internet at random
25:29 all right good place for develop oh good
25:32 place to develop prompt skills
25:34 um I'll tell you a good thing so so go
25:36 to uh go to prompts.chat
25:41 that will give you some good primers on
25:44 prompting
25:45 and then it'll it'll teach you about uh
25:49 telling chat GPT to be
25:52 a certain Persona I want you to act like
25:54 an architect or a screenwriter or a
25:57 therapist or whatever it might be
25:59 um
26:00 and if you pay for chat GPT plus if you
26:03 do the 20 bucks a month at
26:05 chat.openai.com
26:08 you now have access to this thing called
26:10 custom instructions so you can take
26:12 those personas and you can make your
26:14 chat GPT like understand who you are how
26:18 you want it to behave so every chat that
26:20 you do with it will start acting like
26:21 that so understanding that
26:25 um that Dynamic is good is an important
26:27 thing the other thing I would do is go
26:30 to po.com and if you go to po.com you
26:33 can create your own bot and it's really
26:35 simple you just name it pick a model
26:37 there's like four models available you
26:40 can just experiment with them
26:42 you make a description of the bot you
26:45 create a prompt so I want you to act
26:47 like Emily Dickinson and whenever anyone
26:48 asks you anything I want you to write
26:50 poems like Emily Dickinson or stories
26:52 like Emily Dickinson
26:55 um
26:57 and then you just launch it
26:59 and it's out for the world to play with
27:01 you can tell people go play with my bot
27:04 and if you get the prompts right on
27:07 those Bots they're really gratifying
27:09 they're really cool
27:12 I did a sling blade bot and he's pretty
27:14 good he's disturbing
27:21 I wish I had the hardware for llama 65.
27:23 I know all these things here's the deal
27:25 there are so many developers like the
27:28 the Llama models are the number one
27:30 trending
27:32 um project on hugging face right now
27:35 probably on GitHub too
27:37 um the the developer Community is
27:39 pouncing on that [ __ ] so
27:42 um
27:43 probably within
27:49 the early versions in three months
27:51 probably something decently usable in
27:53 six months they'll have those things so
27:55 [ __ ] optimized you'll be able to
27:56 install it just on your phone it's like
27:59 it's we're gonna we're gonna be okay
28:01 we're gonna the the fact that this
28:03 stuff's now open sourced and you have
28:05 this many competitors and this is minus
28:07 Apple
28:08 and this is minus AWS
28:10 this is minus any other well-funded AI
28:14 companies that might deserve to be on
28:16 that list but just haven't they're still
28:17 in stealth mode
28:18 so we're gonna have we're gonna have 10
28:21 to 12 major players doing these large
28:24 language models
28:27 um and so that that just means these
28:28 things are going to get better and
28:29 better and better and better so
28:32 um but I'm right there with you like I
28:34 want to be able to
28:36 two
28:44 two uh I'm a mind reader what did I mind
28:47 read you on like where should you start
28:49 was it that stuff
28:51 that's why this channel exists I want
28:53 people I want people to come here to be
28:55 like I don't know what to do all right
28:57 go do this just go do this
29:01 and then the next question they ask is
29:02 how can I make money
29:05 you can make money
29:09 you know how you make money with chat
29:11 gbt you don't listen to those [ __ ]
29:14 foreign
29:18 copy image address and paste it in
29:20 mid-journey
29:21 with the word macro all right what does
29:24 it do if deal
29:26 if deal or deal one deal A1 dla one
29:32 tell me what it does
29:35 I don't like if it explodes my Discord I
29:37 don't wanna I don't want to [ __ ] up my
29:38 Discord live
29:41 just tell me what it does can you show
29:42 the previous screen one sec I didn't get
29:44 a screenshot sure
29:50 [Music]
29:51 okay
29:56 so there's that one go ahead and
29:58 screenshot that bad boy
30:00 that's all the these are all the chat
30:02 GPT
30:04 variants
30:07 the top three things there are
30:10 two educational resources and then just
30:13 a big ass pile of tools
30:16 those three things are image generation
30:19 tools
30:20 and if if you're sitting there being a
30:23 smart ass going what about stable
30:25 diffusion Man stable diffusion's cool
30:28 well that's what Leonardo is Leonardo is
30:31 a web interface in front of stable
30:32 diffusion
30:34 um the only reason I don't have stable
30:35 diffusion there is it's kind of a you
30:37 have to be a decent geek to get it up
30:40 and running these days so Leonardo is
30:42 just the easy front-end on front and in
30:44 front of that but I'm happy to talk
30:46 about stable diffusion and then runwayml
30:48 is the uh
30:50 is the video tool that does amazing cool
30:54 stuff and getting better every day and
30:56 blockade Labs is kind of cool too and
30:58 then that's that's the uh
31:00 AI community that I created back in
31:04 December
31:05 all right
31:10 that's really helpful the text prompt
31:13 info good absolutely yeah it's here's
31:16 okay here let me let me give you here's
31:18 another thought on prompting because
31:20 because prompting is a really
31:21 interesting thing for one thing the
31:23 whole idea of prompt engineering you
31:25 know everyone's like prompt engineering
31:26 we're gonna you're gonna be a prompt
31:27 engineer it's all about prompt
31:28 engineering
31:29 I don't think it's it's gonna last that
31:31 long I think there will be prompt
31:33 Engineers but I think that
31:36 as these applications get more and more
31:38 sophisticated we as end users will have
31:41 to do less and less sophisticated prompt
31:43 crafting right now because all these
31:46 tools are so soon so new and raw
31:51 we have to figure it out we have to
31:53 figure out how these things actually
31:54 think and how they process and it it's
31:57 how you talk to these things is not how
31:59 you talk how you do a Google search
32:01 right
32:02 I talk about this on here a lot when you
32:05 do a Google search
32:06 over the past bunch of decades search
32:09 engines have taught us make your prompts
32:12 as concise as possible don't make them
32:14 overly complicated or you'll just get a
32:16 mess right so so you like figure out
32:18 what are the minimal amount of words I
32:20 need to get in here these things are
32:21 kind of the complete opposite
32:23 the more context you give them the more
32:25 you explain yourself the more you say
32:27 please and thank you believe it or not
32:29 the more you sort of tell it I want you
32:31 to act like this kind of person and then
32:33 we're going to try to solve this kind of
32:35 problem and now I want you to you know
32:37 help me you know walk me step by step
32:40 through that process they get really
32:42 good
32:44 um it's very counterintuitive
32:46 I my experience working with chat gbt
32:50 and the like is
32:52 more like I got an incredibly smart
32:56 simple-minded intern
33:00 it's an idiot savant
33:03 like I have to explain everything to it
33:06 like oh you don't know where I work
33:07 right
33:09 right we've only been working together
33:10 for six months you don't know where I
33:11 work although the thing that chatgpt
33:13 just added fixes that but but still
33:16 you know you you basically tell it
33:18 everything and explain everything and
33:20 kind of excruciating detail and then
33:22 when you do that it [ __ ] knows
33:23 everything
33:25 it's been
33:27 [Music]
33:29 all right someone wants to pop up let's
33:32 let's talk to someone I think I'm gonna
33:33 lose my
33:36 I think I'm gonna lose my that
33:39 my dual screen but oh well she went away
33:43 all right
33:44 any search you are aware of that could
33:46 be helpful for someone looking for
33:48 business AI business integration
33:52 um
33:53 it depends the level of AI you're
33:55 talking about if you're talking about
33:56 doing sort of chat GPT integration and
33:58 things like that
33:59 um I can't imagine that there's a
34:01 certification out there right now
34:03 because this stuff's a little too new
34:06 um it's possible
34:07 I know that
34:10 Stanford has something MIT has something
34:13 Facebook has something Google has
34:15 something
34:16 I don't know that those are really for
34:18 Integrations though
34:21 um
34:23 but if you're on the integration side
34:25 there's there's enough there's enough
34:26 Geeks that have been working with
34:28 with AI for the past decade that that
34:32 you can you can likely find resources so
34:34 so there's probably certifications about
34:37 doing a lot of the infrastructure stuff
34:39 but if you want to do things like Lang
34:40 chain some of the more
34:43 you know recent projects that are still
34:45 like that you know they're kind of
34:48 flying the plane and building it at the
34:49 same time like Lang chain is this thing
34:51 that's super powerful but it's got all
34:53 these piece parts and everything keeps
34:55 changing and they keep changing out the
34:57 open source large language models and
35:00 you can put in different kinds of vector
35:02 databases and [ __ ] like that so
35:04 um so I don't know I wish I could be
35:05 more help with that um
35:08 no I would I I would
35:14 you know it's of all of these the place
35:17 to maybe go ask that is Bard Bard seems
35:20 to be they seem to have integrated the
35:23 Google search indexing into the answers
35:25 at Google barred really well and it's
35:27 connected to the internet so Bard might
35:30 be the place to go ask about
35:32 um certifications the other thing if
35:34 you're a Microsoft Shopper or if you've
35:36 got any kind of Microsoft infrastructure
35:38 not opposed to it a lot of the
35:40 developers I know are opposed to it but
35:43 if you're not opposed to it
35:45 um
35:46 they're probably the farthest ahead in
35:49 terms of integrating
35:51 chat GPT into Azure into all the cloud
35:54 services
35:56 um but any like they just had their
35:59 their developer conference it's not all
36:00 it's not all rolled out yet so it's this
36:03 stuff's all very very new
36:07 but Microsoft probably has some decent
36:09 resources at this point because they've
36:10 been working with open AI for what a
36:12 year and a half something like that
36:13 because it was a year before we knew
36:14 anything about it
36:16 all right let me get to some other
36:19 questions welcome everybody welcome
36:22 where do you work I've been studying AI
36:24 models and emerging properties I work
36:27 I'm a founder and CEO of storybine which
36:32 is not an AI company although we are
36:34 adding a lot of AI
36:38 I'm completely reimagining my product
36:40 and my value proposition of my 11 year
36:42 old company
36:44 thanks AI
36:45 [Laughter]
36:48 I I'm more an entrepreneur so I'm not
36:50 this channel what my expertise is is
36:53 more on the
36:55 once you hit that inflection point where
36:59 powerful tools like uh GPT 3.5 and 4
37:03 are accessible to the rest of us that's
37:06 when I like jumping into this stuff
37:08 because I'm not technical enough to be
37:10 in the game before that
37:12 but the minute this [ __ ] shows up ready
37:15 for prime time which I think this stuff
37:16 is now I think chat GPT is the is the
37:20 inflection point so November 30th of
37:22 last year so that's that's what this
37:23 channel is about
37:25 um so if you want someone more technical
37:27 I I the the guy that I think is I find
37:30 the most impressive that does anything
37:32 you know remotely like what I'm doing
37:34 education wise is David Shapiro and
37:36 YouTube Demands a [ __ ] genius and
37:39 he's got all sorts of Open Source
37:40 projects doing all sorts of
37:43 heading toward AGI in an open source way
37:46 um and he's just [ __ ] brilliant he
37:48 does like three or four major
37:49 presentations a week on the implications
37:52 of AI from a
37:55 deep deep
37:58 intellectual and Technical point of view
38:01 the dude's a [ __ ] machine he's
38:03 amazing
38:05 um can you recommend anything to help
38:06 with trading in regards to AI I can't
38:09 really what I would say is Go to
38:11 futurepedia.io
38:13 go there and look for anything around
38:16 trading and then also the other thing
38:19 you can do is go to whatever your
38:20 trading platform is especially if
38:22 they've got Bots if you can do trading
38:23 Bots and things like that you can just
38:26 copy the code from a trading bot on your
38:28 platform paste it into chat GPT
38:30 especially gpt4 and just tell it to
38:33 explain to you what the bot is what it
38:35 does and then tell it what you want it
38:37 to do and it'll it'll rewrite it for you
38:39 so so
38:41 the other thing the other thing about
38:44 um
38:44 about chat gbt code interpreter you can
38:47 upload up to 100 megabyte data file so
38:52 you could take say historical trading
38:54 data
38:55 upload that into a code interpreter and
38:58 have it start doing data analysis for
39:00 you so there's all sorts of things you
39:02 can do just out of the box with some of
39:04 these things that are
39:06 [ __ ] insane I don't know that world
39:08 at all so I don't know the specific
39:09 tools so that's why
39:11 futurepedia.io might there might be some
39:14 people that are doing some
39:16 really sexy stuff uh I just don't know
39:18 about it
39:22 um you can ask GPT to rate your prompt
39:24 oh and it will suggest improvements as
39:26 well yeah that's actually a good hack
39:28 you so you you write a prompt you you
39:32 ask Chad GPT if if there's a better way
39:34 to write it you can also when it answers
39:37 you have it analyzed its own answers
39:40 say review your answer and tell me how
39:42 you might improve it and it'll do that
39:44 and then you can tell it to improve it
39:45 and then it will what did I do today
39:47 that was really frustrating
39:49 um
39:51 well I was having it I was having it try
39:53 to
39:54 um
39:55 to rewrite and this was with Claude it
39:57 wasn't with gpt4 to be to be fair
40:01 um
40:02 I was having it rewrite
40:04 a
40:05 document that I wrote and it just kept
40:08 truncating like all the paragraphs that
40:10 would sort of knock them down to two
40:11 sentences instead of four or five and
40:13 and then it would just omit certain
40:15 paragraphs it was weird I couldn't I had
40:17 a frustrating day today with chat gvt or
40:20 Claude whatever it was I was playing
40:21 with
40:23 um
40:24 most days are are revelatory and fun but
40:26 today yeah yeah not so much
40:29 not so much
40:32 I know what's wrong with it ain't got no
40:34 gas in it
40:39 Adam McKnight coming in strong with the
40:41 memes we like the memes
40:44 that was the bot website what was the
40:46 bot website agreed thanks
40:49 uh can you recommend anything to help
40:51 with trading I already answered that no
40:53 I can't what was the bot site uh poe.com
40:59 po.com has cool Bots
41:04 [Music]
41:10 all right so you go to Poe and you got
41:14 six different chat gbt models and then
41:17 you just have a [ __ ] ton of bots you
41:19 have this explore button which if you
41:21 click it it just takes you over to all
41:24 these Bots and basically all these are
41:25 are like pre-prompted
41:28 GPT things I don't know why this isn't
41:30 focusing
41:32 did it Focus no
41:35 um
41:35 well whatever
41:38 um like there's a mid-journey bot
41:39 there's a ghost bot there's an emoji bot
41:42 that only answers in emojis there's
41:44 cooking Bots and then you can make a bot
41:47 create a bot
41:48 and then you just name it give it a
41:50 description give it a prompt give it a
41:52 welcome message publish Bang
41:55 and then don't be a doo-doo head and not
41:57 have a picture go to Mid journey and
41:59 make a picture for it
42:01 like I did yesterday I did Christopher
42:04 walkenbot
42:06 I've done what plots have I done
42:10 uh Shakespeare I am Shakespeare so if
42:13 you go to poe.com I am Shakespeare
42:16 I did Jay-Z that's a good one Yoda Sling
42:20 Blade Emily Dickinson Steve Jobs I am
42:23 Steve Jobs that's a fun one
42:26 I am Mary Shelley I'm Virginia Woolf or
42:28 not I'm just Virginia Woolf I am Poe
42:33 so yeah that's Poe that's Poe
42:39 [Music]
42:42 paid Discord
42:45 I don't know what you mean
42:48 the image wait it merges pictures of
42:51 humans with bugs in a very interesting
42:53 way
42:54 mid-journey does if you put an old 40s
42:56 or 50s photo in it
42:59 that doesn't surprise me
43:02 but that's kind of cool
43:04 huh is that a bug are people considering
43:07 that a bug
43:09 interesting interesting
43:13 aqua blue hi everyone hello
43:18 I got SD running I'm a geek yeah listen
43:21 stable diffusion is awesome so so if you
43:23 want to do the this image stuff
43:26 and you want to geek out on it
43:29 um you know I talk about leonardo.ai is
43:31 the is the web version of stable
43:32 diffusion
43:34 um if you've got the stomach for getting
43:36 a Google collab up and running and and
43:38 you know getting automatic 11 11 up and
43:40 running and the code stuff doesn't scare
43:42 you off like it took me I failed nine
43:45 times trying to get stable diffusion up
43:47 and running I finally got it running and
43:48 I got dream Booth working and you know
43:51 it felt like [ __ ] victory yeah I did
43:53 it yeah [ __ ] yeah after nine [ __ ]
43:57 failures but that's just the stage of
43:59 the the industry we're in right now
44:01 where the tools are really powerful the
44:03 nice thing about stable diffusion is the
44:05 developer communities building new ship
44:07 for it all the time which is also the
44:09 downside of it right because you get
44:11 your automatic 11 11 up and running and
44:14 then they changed the notebook on you
44:15 and then it doesn't work anymore because
44:17 they swapped out a library and this and
44:19 that now it's using a different model
44:20 and you just constantly have to stay on
44:23 top of it so if you don't want to do
44:24 that that's what these that's what these
44:26 tools are like Leonardo that's just a
44:28 company that's sticking the front end
44:29 they're taking care of all that back end
44:31 [ __ ] for you and that's what you think
44:33 for right you pay a service for them to
44:37 solve your [ __ ] headache of having to
44:38 install it yourself
44:40 that's how business works
44:42 yeah but stable diffusion is quite
44:44 amazing
44:46 do you think Leonardo's are in the works
44:51 for llama too well that's what
44:53 labs.perplexity.ai that's llama2
44:56 that you can play with the 7 the 13 and
44:58 the 70 billion parameter version of
45:00 llama two right there for free right now
45:02 go play now you can install it and do
45:05 all the training and [ __ ] like that do I
45:07 think there will be a Leonardo kind of
45:09 front end for for those things
45:11 absolutely
45:13 absolutely and and I you know I think
45:16 probably in in a short amount of time
45:20 I feel like the next phase of these is
45:22 going to be the the the training on your
45:25 personal data phase right where right
45:27 now it's all just like they're they're
45:29 either quasi or not at all secure
45:33 um they're jankier than [ __ ] they're
45:35 like really thin application layers
45:38 um so I think what's gonna happen is
45:41 this you're gonna have you've already
45:43 got the Lang chain Community out there
45:45 doing
45:46 you can embed your own data into Lang
45:48 chain and there's there's projects that
45:51 are putting front ends and for on in
45:52 front of Lang chain you've got GPT for
45:55 all which allows you to play with
45:57 different open source models it's just a
45:58 double click install
46:01 um you're gonna see apple Microsoft and
46:04 Google all come out with versions of
46:07 these things that are now connected to
46:09 your data so like Google just say just
46:12 point your Google Drive Point this [ __ ]
46:14 at your Google Drive and you can now
46:16 start interacting with your data
46:18 Microsoft's going to do the same thing
46:20 Apple's going to do the same thing and
46:22 then the open source Community is going
46:24 to [ __ ] pile on right because the the
46:27 foundations for it are built it being
46:28 built right now so I think six months
46:31 from now
46:34 um these things are going to look and
46:36 act very very different than they do now
46:38 they're going to be much more in the
46:39 neighborhood of a personal assistant or
46:41 it's just going to be much more useful
46:43 like I know somewhere in the past decade
46:46 I sent an email to someone talking about
46:48 this specific thing and I can't find it
46:51 in Google
46:52 and you'll just be able to go find it
46:54 now
46:55 because it's it's in your [ __ ] and
46:57 they've got it in the [ __ ] Magic
46:58 Machine
46:59 they've embedded it once it's embedded
47:02 magic [ __ ] happens right
47:07 all right
47:13 Miss Birch he takes a little while to go
47:15 through the questions
47:18 my regular my regulars are apologizing
47:21 for me thank you very much I am slow
47:23 with them but I do try to answer them
47:24 all
47:25 [Laughter]
47:28 um Rachel awesome
47:30 Jack Strickland it will be tremendous
47:33 Improvement when they can tell you when
47:36 not to trust their answers yeah so if
47:38 you're new to this stuff that's an
47:40 important point that I forgot to make
47:41 earlier
47:43 um these things do what's called
47:44 hallucinating which is a the Geeks came
47:47 up with a clever word to not make it
47:48 sound like such a [ __ ] disaster that
47:50 it actually is they just [ __ ]
47:52 outright lie these things will just tell
47:54 you stuff confidently even if it's wrong
47:58 I like to describe it as mansplaining as
48:00 a service
48:02 like all the good men in your life it'll
48:04 confidently tell you something even if
48:05 they've their head up their ass
48:08 um
48:09 and it's the the nature of how these
48:12 things actually work they they are
48:15 they are as aware of what they
48:18 say to you as a calculator is of the
48:21 answers it's giving you
48:23 it's not aware it's a calculator that's
48:27 doing some formula you know in its
48:30 circuitry and it pops up an answer
48:32 that's all these things are doing from a
48:34 technical perspective it's literally
48:35 predicting the next word in a sequence
48:38 of words based on your prompt
48:41 so when it writes you a whole page of
48:44 stuff that looks like holy crap that's
48:45 brilliant it knows all this stuff
48:48 it's not actually aware of what it just
48:50 created
48:52 I think the the next evolution of these
48:57 there will be a layer there where it
48:59 actually looks at what it wrote and says
49:01 is this [ __ ]
49:03 if [ __ ] go back to the starting hey
49:07 start over
49:10 um
49:11 they're gonna get better right
49:13 I'm I'm looking at how Bard is
49:15 incorporating search results into its
49:18 into its results and how Bing is doing
49:21 that
49:22 they're starting to do some some
49:24 citations and things like that you know
49:27 link link to websites
49:29 um so we're probably not that far off
49:31 from that but yeah I I am right there
49:33 with you that it will be a tremendous
49:35 Improvement where you can actually trust
49:37 these things especially when you start
49:39 to get into like like Google just came
49:41 out with uh was it Google yeah yes
49:46 they just announced Med Palm
49:49 which is a
49:50 medical large language model right it's
49:53 it's one of these it's like chat GPT but
49:55 trained up on all the medical [ __ ] well
49:57 we don't want the medical [ __ ]
49:59 hallucinating and telling us to take you
50:01 know
50:02 640
50:04 you know Advil to to make the headache
50:07 go away
50:09 that would be bad
50:11 I am not a doctor but I know that
50:13 wouldn't be good where are you answering
50:15 from I am well Patrick cutaway
50:19 your way down there I'm way up here all
50:22 right so do you ever see Star Wars the
50:24 original one the four the four five six
50:26 ones they're like the good ones yeah
50:28 that was when I was a kid you know at
50:30 the beginning of that where you had
50:32 words down here and then there was the
50:34 big long line of words and then way off
50:36 in the distance I'm answering questions
50:39 up here you're down here
50:44 all right do we have mods
50:48 as in mods for The Tick Tock no as in
50:53 modifications for
50:56 these
50:57 Dr J I don't know what you're talking
50:59 about Dr J no mods oh yeah like mods
51:03 like someone in here actually like oh is
51:06 the question is the AI learning lab a
51:08 Professional Organization no
51:14 and if one of you down below there said
51:16 what are your qualifications boy I ain't
51:18 got none
51:20 ain't got none other than the fact that
51:22 I'm geeking out on this [ __ ] every
51:24 [ __ ] day of my life for the past nine
51:26 months actually it's more like a year
51:28 and a half now
51:30 old black and white photographs contain
51:32 different information
51:34 uh than the command macro and mid
51:36 Journey oh that's interesting all right
51:39 I'm gonna have to look that up I wanna
51:40 write that down
51:42 old black and white photographs in
51:44 mid-journey that's kind of cool I like
51:46 it
51:47 I wonder do you think it's
51:50 I wonder if there's something about the
51:51 50s images that it's that it's it's
51:54 somehow like doing references to like
51:58 old 50s sci-fi because you said it was
51:59 people and insects being mixed I wonder
52:02 if there's something about black and
52:03 white old sci-fi Stills and things like
52:06 that that are
52:07 mixing in there
52:10 he is Colorado I'm in Colorado yes I'm
52:13 in Denver Colorado
52:15 welcome to Denver
52:17 it's the base of the Rocky Mountains
52:25 did AI begin with Quantum Computing no
52:27 AI began with oh I don't know
52:30 back in 1956
52:34 the term AI was coined
52:38 1957. I just did a spreadsheet on this
52:41 do I have it on on this machine yeah
52:43 yeah
52:47 that's on my work machine
52:50 um I just did some research on this
52:52 um
52:53 that started with these things called
52:55 neural Nets right so basically starting
52:59 70s or 80s they started trying to
53:02 replicate the structure of neurons in
53:06 our brain and and the basic
53:08 idea has been
53:12 if we can get computers to act like the
53:15 synapses and like the neurons in our
53:17 brain we might be able to get them to do
53:18 cool stuff
53:19 and they were making you know nominal
53:23 prod progress and you know they were
53:25 doing things like you know expert
53:27 systems like IBM's big blue that beat
53:30 Gary Kasper off at chess because they
53:32 just trained it you know on that really
53:33 specific narrow data set
53:35 and then it was the Transformer
53:38 architecture that that
53:41 um Google came up with in 2017 that is
53:45 what burst the dam open for all of what
53:48 we're doing now
53:49 so the embeddings and Vector databases
53:53 just how it does what it does like it
53:55 takes the
53:57 the mass of everything it learned and it
53:59 turns it into these funky mathematical
54:03 word pairings that are sort of connect a
54:06 word to its meeting but just in math and
54:08 so when you type in your prompt
54:11 it you know magically reconstructs
54:16 something out of nothing it's it's quite
54:18 remarkable what it does
54:21 um but but uh Quantum Computing as I
54:25 understand it I don't know a ton about
54:26 it
54:27 it's a different kind of mathematics
54:29 that it does so so you can't just use a
54:31 quantum computer like you use a regular
54:33 computer so you couldn't you couldn't
54:35 necessarily just do AI machine learning
54:39 training kind of stuff with a quantum
54:41 computer but
54:42 at some point some will figure that [ __ ]
54:44 out Google just
54:46 released the uh the results of their
54:51 their new quantum computer Sycamore
54:53 which has 70 qubits in it which if you
54:56 follow the world of quantum Computing
54:58 every qubit is a Quantum increase in
55:00 power so 70 you're getting up there and
55:04 it performed a mathematical equation in
55:07 six seconds that would take today's most
55:10 powerful supercomputer 47 years
55:15 so that's what's going on in the quantum
55:17 Computing world right now I don't think
55:18 it's too connected to this AI stuff but
55:21 I could be wrong I just but I I did a
55:23 video on the Sycamore thing and then a
55:25 bunch of people in there go listen you
55:28 [ __ ] tool it's not the same kind of
55:30 mathematics you can't just use quantum
55:33 computers on your stupid little chat gbt
55:36 [ __ ] tool you [ __ ] loser
55:39 so I took that as constructive criticism
55:42 and I learned that I'm an idiot
55:48 but still it's doing something in six
55:51 seconds that would take the fastest
55:52 computer 47 years that's kind of cool
55:55 you know that shit's gonna get in here
55:56 somehow
55:57 all right
56:00 um no like he can even see this chat
56:03 yeah I can see this chat Patrick cutaway
56:05 you bitter bitter man does Patrick need
56:09 a hug do you need a hook do you need a
56:11 hug I really am reading all the
56:13 questions I really am live the bear can
56:16 vouch for me
56:18 thank you the bear I appreciate that
56:22 and by the way if you have not seen the
56:25 TV show The Bear shame on you
56:28 we talk about comedy in this channel we
56:30 talk about good TV shows the bear is
56:32 remarkable
56:34 remarkable remarkable remarkable the
56:38 race is on with all of them
56:40 who to predict who the overall winner
56:42 when the dust settles I don't [ __ ]
56:43 know I know it'll be three of them it's
56:45 always three right these things always
56:47 settle down to three
56:50 which
56:55 oh which AI can produce a PowerPoint
56:58 presentation the one that I know is Tome
57:00 t-o-m-e I think it's tome.ai hang on
57:06 and they've added a bunch of new
57:08 features now
57:10 tone tone.app
57:14 go to tome.app
57:19 yeah tone.app and then it'll just you
57:23 can just say I'll say create your watch
57:26 create a presentation about
57:29 um
57:32 my
57:36 what's the presentation about where do I
57:38 put it here we go
57:41 um my what
57:48 um
57:49 my business
57:52 to wash
57:54 wild raccoons
58:00 so I'm starting a new business to watch
58:02 wild raccoons as you do
58:06 so it's generating an outline
58:08 it's called raccoon Suds the wild
58:11 business of washing
58:13 and then it creates you
58:16 um an outline
58:18 and if you're happy with that I guess
58:19 you can go in and edit these now oh
58:21 that's pretty slick
58:22 so introduction the problem the solution
58:24 the market the marketing strategy the
58:26 conclusion you say continue and now it
58:29 makes you a presentation
58:34 so it writes them and it does the images
58:36 so we should get some raccoon images
58:38 because
58:39 there's lots of raccoon base art
58:44 um
58:45 [Music]
58:46 cute
58:58 ah there you go we did it now
59:04 are these presentations any good I don't
59:07 know probably not
59:10 all these tools right now all of these
59:13 tools right now
59:17 are
59:21 a hundred times more powerful than
59:23 anything you've played with
59:26 and radically flawed
59:28 [Laughter]
59:31 so so if you do it right you should have
59:35 a Kevin McAllister moment right from
59:37 Home Alone this moment
59:40 I didn't know I could do that
59:44 and then as you start to use it for real
59:46 stuff you realize oh it's got these
59:47 limitations and those limitations and
59:49 those limitations but what I would say
59:51 is that the the
59:55 be very
59:58 I would be very reticent if I were you
1:00:01 to not dismiss these tools because
1:00:03 they're not perfect because they're as
1:00:05 bad as they're ever going to be they
1:00:06 only get better from here and what they
1:00:09 do is absolutely [ __ ] miraculous like
1:00:12 we're lucky to be able to play with this
1:00:13 stuff
1:00:15 you can see the chatty's just not caught
1:00:17 up thank you Stephanie I've got my
1:00:19 Defenders here he really does the
1:00:21 Institute questions he really is live he
1:00:23 cares about us he does
1:00:27 good evening Mr Shannon good evening
1:00:29 all right
1:00:31 Tick Tock just jumped me to the bottom
1:00:33 again my regular viewers know that I
1:00:35 hate that because I lose my place and I
1:00:38 like to answer all the questions
1:00:40 contrary to popular belief
1:00:43 if you know what I mean
1:00:45 where's he answering these from I don't
1:00:47 think he's answering these maybe he's an
1:00:49 AI you think he's an AI I heard he's
1:00:51 chat GPT are you chat GPT
1:01:01 [Music]
1:01:04 I did what did I do I did a tick tock
1:01:06 for I did a tick tock for April Fool's
1:01:09 Day right where
1:01:11 I had chat gbt come up with a
1:01:14 with three other words then generative
1:01:16 free train transformer that meant that
1:01:18 GPT was actually just really like a big
1:01:22 call center full of full of people on
1:01:24 computers feverishly typing out the
1:01:26 answers
1:01:28 a couple of people got it other people
1:01:29 was like I don't think that's really
1:01:31 happening man
1:01:32 people are just wound too [ __ ] tight
1:01:35 on the internet they people got to
1:01:37 lighten up gotta light up and fly right
1:01:43 all right let's hair hair G reached
1:01:48 level something
1:01:49 's hair growth geek I don't know you
1:01:52 reached some sort of level
1:01:53 congratulations
1:01:54 I love this [ __ ] LOL you're very
1:01:56 entertaining and knowledgeable thank you
1:01:58 Patrick
1:01:59 appreciate that
1:02:01 Kyle is the sheer sheer sure she oh [ __ ]
1:02:09 yeah
1:02:09 I'm like man these kids these days they
1:02:12 got all these words that mean something
1:02:14 I don't I don't know
1:02:16 I'm the [ __ ] I appreciate it thank you
1:02:19 thank you thank you Miss Birch
1:02:22 Chachi BT got Dumber once they connected
1:02:24 to the government we get the bad AI now
1:02:28 oh gbt
1:02:31 654321 uh you got some work to do on the
1:02:34 name but uh
1:02:37 to go download
1:02:39 [ __ ]
1:02:41 llama too it's open source and
1:02:43 commercially usable
1:02:45 don't be so bitter
1:02:47 it's not all a conspiracy it like here's
1:02:50 the deal with all those conspiracy
1:02:52 theories they all sort of require that
1:02:54 the government is competent at doing
1:02:57 conspiracies
1:02:59 the government's got their heads so far
1:03:01 up their ass
1:03:02 oh yeah
1:03:04 they don't know what [ __ ] Wi-Fi is
1:03:09 yeah whatever see it just
1:03:13 it's probably more Microsoft than the
1:03:16 government by the way
1:03:18 putting putting the shackles on chat EBT
1:03:21 but now you have Claude
1:03:24 now you have barred now you have that
1:03:27 now you have Pi I mean you've got
1:03:29 options the whole open source Community
1:03:31 is there for you
1:03:33 so like
1:03:35 the the your comment I don't mean to
1:03:37 pick on you but your comment is
1:03:41 it's really easy to demonize and dismiss
1:03:44 this [ __ ]
1:03:46 and I think that does us all a
1:03:48 disservice
1:03:50 because if you really take your time and
1:03:53 learn this [ __ ] learn what it's good at
1:03:54 learn what it's not good at learn where
1:03:56 the boundaries are learn
1:03:58 if indeed it is getting better or worse
1:04:01 and then you decide
1:04:03 that stuff cool
1:04:06 but if you're just sitting on the
1:04:07 sidelines going ah it's a big government
1:04:09 conspiracy no that doesn't serve any of
1:04:12 us
1:04:14 because the more of us that use this
1:04:16 stuff for good
1:04:18 are gonna lean the direction of where
1:04:20 this head's in the future toward the
1:04:22 good if we sit on the sidelines and let
1:04:25 these [ __ ] just you know
1:04:28 make all the rules up then we are going
1:04:30 to be at the mercy of it so let's not do
1:04:32 that
1:04:33 and we might as well have fun with it
1:04:35 like it's the most remarkable [ __ ]
1:04:37 technology in the history of humanity we
1:04:39 get to play with it
1:04:41 there's like [ __ ] seven of them now
1:04:43 that we can play for effectively for
1:04:44 free almost all of these are free
1:04:47 mid-journey's 10 bucks a month you can
1:04:49 make world-class images
1:04:52 for 10 bucks a month
1:04:54 like full-on photographs full-on
1:04:57 illustrations
1:05:03 you know what I mean
1:05:06 so
1:05:08 don't [ __ ] on this stuff until you've
1:05:10 actually used it and listen if you've
1:05:12 been using it for 20 years and
1:05:14 you got some information then [ __ ]
1:05:16 share the information
1:05:17 if they hobble chat GPT and you know it
1:05:20 in a specific way cool
1:05:21 but you know
1:05:23 in its honest to God if I just if if we
1:05:27 shut everything down you know from chat
1:05:29 GPD down if we shut everything down and
1:05:31 I just wanted the chat GPT 3.5
1:05:35 I feel like I personally could spend the
1:05:38 next decade
1:05:39 trying to figure out all the different
1:05:41 ways I could exploit it
1:05:44 and gpt4 is insanely better
1:05:49 so even if gpt4 goes from like whatever
1:05:51 it's at right now you know 92 and it
1:05:55 drops to 85. it's still
1:05:59 a hundred times more powerful than any
1:06:02 of the [ __ ] else thing I'm using
1:06:06 so
1:06:07 uh what do I think about bgsm I don't
1:06:10 know bgsm what is bgsm
1:06:14 um
1:06:15 [Music]
1:06:21 biosig Technologies
1:06:24 I don't know anything about it
1:06:27 they doing some sort of crazy ass brain
1:06:30 [ __ ]
1:06:31 some drug [ __ ]
1:06:34 biosig Technologies I don't know enough
1:06:37 about it
1:06:38 certainly if you're asking for for
1:06:40 investment I don't know
1:06:44 why did sound draw come up is that sound
1:06:47 draw is that part of that company I
1:06:48 don't think so
1:06:49 I don't know I don't know anything about
1:06:51 it sorry about that
1:06:53 I keep breaking GPT doesn't finish
1:06:55 generating about 75 percent of my input
1:06:58 um yeah that's that's an issue so I
1:07:01 would say go go start playing with
1:07:03 Claude
1:07:04 so Claude has a context window of um
1:07:09 a hundred thousand tokens so it's got a
1:07:11 much longer memory per chat session and
1:07:13 it can also generate uh long responses
1:07:15 better so try Claude
1:07:18 um it's worth playing with Bart honest
1:07:20 honestly right now
1:07:23 there's so much so new on this list that
1:07:26 that what I would do if if I were anyone
1:07:29 here that's really exploring these
1:07:30 things if you're using it enough that
1:07:32 you're you're getting chat GPT to break
1:07:35 I would start exploring different tools
1:07:38 and see which ones are better right I
1:07:41 think Bing is really really hobbled and
1:07:44 and
1:07:45 there's like a handful of things I use
1:07:47 Bing for but for the most part I don't
1:07:49 because it's just so hobbled Claude is
1:07:52 really impressive I've been I would say
1:07:55 right now I'm I'm kind of 50 50 between
1:07:57 using chat GPT
1:08:00 and code interpreter and then Claude
1:08:03 just for day-to-day use
1:08:05 um but I'm impressed with Bard and how
1:08:07 it's incorporating index links into into
1:08:11 organically into search results without
1:08:13 it being spammy SEO [ __ ]
1:08:16 um I'm actually impressed with that
1:08:18 um the Llama models are raw and new so
1:08:21 you can go play with those Pi is very
1:08:23 very conversational
1:08:25 um
1:08:27 so but my my instinct is if you're
1:08:29 having truncation issues try Claude
1:08:35 what was the bot site that was po-po.com
1:08:38 is access to Discord a paid subscription
1:08:42 um
1:08:44 I don't no I don't think so
1:08:46 I don't think it is but but
1:08:49 the ability to generate more than like
1:08:52 the 20 initial test images in
1:08:55 mid-journey is a paid subscription
1:08:59 you can pay for Discord subscription to
1:09:02 get some other features but I don't I
1:09:04 don't think you need it
1:09:06 I think I do pay for it though but not
1:09:08 because I did it I started paying for it
1:09:11 back when I was doing blockchain and nft
1:09:12 [ __ ]
1:09:14 and I just haven't canceled my
1:09:16 subscription
1:09:17 Man stable diffusion is awesome it's so
1:09:19 there's so much stuff going on I know
1:09:20 bill so so if if you if you have the the
1:09:24 stomach for doing the geeky stuff to get
1:09:26 into the stable diffusion Community like
1:09:29 there's so much development going on
1:09:31 there and if you can keep up with it
1:09:33 but it's it's a lot and it's but it's
1:09:36 good it's good
1:09:38 um if you just want Kick-Ass images use
1:09:40 mid-journey if you want flexibility and
1:09:42 to keep up with the latest and greatest
1:09:44 ghost stable diffusion if you want to do
1:09:46 things commercially go with Adobe
1:09:48 Firefly and generative fill within
1:09:50 Photoshop
1:09:51 because it's trained on a safe model
1:09:54 sorry sorry Leonardo has voluptuous
1:09:56 young women bias I know kind of
1:09:59 disturbing how it was probably trained I
1:10:01 I don't well
1:10:03 see how it's trained it might be just
1:10:05 that
1:10:06 a lot of the people using it right now
1:10:09 are using all of the all of the sort of
1:10:11 anime models anything that was sort of
1:10:13 done on anime just skews way toward you
1:10:16 know tits and ass and yeah it's it's
1:10:20 it's annoying the the thing about
1:10:22 Leonardo though is you can train your
1:10:24 own models right so you can upload a set
1:10:27 of photographs or images that that you
1:10:30 like train your own models and then
1:10:31 start working within that so there's
1:10:34 good flexibility there but yeah if you
1:10:36 look at the initial feed when you land
1:10:37 on Leonardo it's just like it's just
1:10:40 like a bad anime convention so okay God
1:10:44 Kids come out of the basement really
1:10:50 um
1:10:50 come on make a velociraptor or something
1:10:54 I mean you know
1:10:57 all right the bug is that it creates
1:10:59 bugs or plants and merges it with the
1:11:01 photos will do it on any photo I gotta
1:11:04 try that that's fascinating let's go try
1:11:06 it now
1:11:07 if you're still here if deal if deal
1:11:10 image so let's go do that
1:11:15 um
1:11:17 let's see
1:11:18 50s
1:11:20 photo
1:11:23 graph
1:11:26 um
1:11:27 I don't know uh what is this called uh
1:11:33 Diner
1:11:36 oh no I don't want the actual Diner
1:11:42 no
1:11:43 uh 50s
1:11:46 people pictures
1:11:48 here we go
1:11:55 that's British life in the 50s
1:12:18 grab this one
1:12:21 all right so I don't quite know how to
1:12:23 do this because I'm not talking to you
1:12:25 but
1:12:26 50s
1:12:28 so I'm going to save it there we're
1:12:30 going to hop on over to Discord
1:12:40 oh crap
1:12:55 I guess 50s
1:13:03 desktop
1:13:06 [Music]
1:13:08 why is it not showing up oh there we go
1:13:09 okay
1:13:11 so we're gonna pop that in there
1:13:15 I'm gonna upload it I suppose I could
1:13:17 have just grabbed that URL but I know I
1:13:19 know this works so I'm going to copy the
1:13:21 link
1:13:25 I gotta add a text prompt okay
1:13:31 slash imagine
1:13:33 is
1:13:35 um
1:13:37 let's see
1:13:41 a dressing room Beauty
1:13:46 all right I'm gonna let that go I got is
1:13:49 it doing yeah it's waiting to start all
1:13:51 right I'll be right back
1:13:54 foreign
1:15:22 foreign
1:15:24 all right what did it do did it make any
1:15:27 weird insects
1:15:29 I don't think it did
1:15:32 I don't know
1:15:33 it didn't do it
1:15:34 all right
1:15:36 I'm back sorry about that
1:15:40 [Music]
1:15:47 if you want to use stick to version 1.5
1:15:49 is that in mid-journey
1:15:52 weird well I'll look it up that sounds
1:15:54 like a really interesting bug
1:15:58 interesting bugs are cool
1:16:01 like why does it do that you know
1:16:09 I CMP the description of a character
1:16:11 from a novel into mid-journey and I came
1:16:14 back with a weird alien yeah I mean
1:16:19 so if you understand how the diffusion
1:16:21 models work
1:16:23 they're they're
1:16:25 the trick is in getting your
1:16:28 your prompt right
1:16:30 if it's too many words if it's not
1:16:32 enough words if if you're if you're
1:16:35 basically leaving the latent space too
1:16:38 big and Broad it kind of does these
1:16:40 averages of all these noised images and
1:16:42 then denoises them and so you'll get
1:16:44 some weird [ __ ]
1:16:46 um
1:16:47 it's just the nature of the Beast right
1:16:49 now I mean it's part of the joy the joy
1:16:51 of it also right like you know
1:16:55 creating [ __ ] like
1:16:58 you know like these images
1:17:02 that I did in stable diffusion with
1:17:04 dream Booth I mean
1:17:06 those are super cool
1:17:11 you know and that was just kind of a
1:17:14 you know experiment that I've probably
1:17:17 done
1:17:18 I don't know 2500 or 3 000 images then I
1:17:21 finally got one
1:17:22 that I was like holy [ __ ] that's [ __ ]
1:17:25 art you know and like that's the kind of
1:17:27 art I would create if I had the talent
1:17:30 so
1:17:33 but it's the nature of the Beast like I
1:17:34 said these these these tools are
1:17:38 um
1:17:40 remarkably powerful
1:17:43 um and remarkably flawed right now it's
1:17:44 just we're just very very early so
1:17:47 so just know that they're as bad as
1:17:50 they're ever going to be
1:17:52 know that
1:17:55 have you noticed there's a lot less
1:17:56 hands like this than there used to be
1:17:59 why
1:18:01 um because they're fixing that [ __ ]
1:18:07 you're not slow you're just thoroughly
1:18:10 in your elaborated answers thank you
1:18:12 cocoa beans oh 9.5 on the on the name
1:18:17 spelled cleverly
1:18:19 um parsable and funny uh so 9.5 out of
1:18:23 10 on the name cocoa beans I like it uh
1:18:26 can these platforms help someone find a
1:18:29 new job
1:18:30 um
1:18:31 the platforms can't help you find a new
1:18:33 job but I think that they can certainly
1:18:36 help in the process
1:18:37 so
1:18:38 for example
1:18:40 [Music]
1:18:42 take your resume
1:18:45 and
1:18:46 maybe your LinkedIn history like you can
1:18:49 download your LinkedIn history you can
1:18:50 just copy and paste chunks of it into a
1:18:52 document
1:18:54 and then with something like Claude
1:18:58 you could just upload
1:19:00 resume One resume two
1:19:03 LinkedIn profile I think you can do up
1:19:06 to five documents
1:19:09 and then
1:19:10 paste in a job description
1:19:14 and tell it you want it to use markdown
1:19:17 so that it looks pretty marked down is
1:19:19 the thing that will format it I can show
1:19:20 you how to do that
1:19:21 tell it to rewrite your resume you know
1:19:24 based on today's best resume writing
1:19:26 practices
1:19:29 you know focusing on the jobs and
1:19:31 experiences for that job that you put in
1:19:33 there as a description and then have it
1:19:35 write you a personalized cover letter to
1:19:37 that hiring manager
1:19:39 so you do stuff like that right you
1:19:41 could do stuff like I'm working with a
1:19:43 guy right now who's who's building a
1:19:46 tool
1:19:47 for sales people to be able to take
1:19:50 their LinkedIn profile and create a
1:19:53 custom persona for them
1:19:55 so that the sales people know how to
1:19:57 talk to them what to talk about what
1:20:00 they value all that sort of stuff just
1:20:02 based on their LinkedIn profile
1:20:04 um so you do the same thing with hiring
1:20:06 managers right or or job descriptions
1:20:11 um with things like Bard that's
1:20:14 connected to the internet or Bing that's
1:20:16 connected to the internet
1:20:17 you can potentially find jobs that you
1:20:20 might not have otherwise found
1:20:23 um
1:20:24 so that's on the job seeking side
1:20:27 if you're getting AI literate if you're
1:20:30 getting up to speed on these tools start
1:20:32 talking about it start posting on
1:20:34 LinkedIn start posting to your Facebook
1:20:36 groups start talking about it at dinner
1:20:39 parties or
1:20:40 wherever
1:20:41 with your friends
1:20:43 even if you're like well my friends
1:20:45 think ai's stupid and I don't want to
1:20:46 look like an idiot look like an idiot
1:20:49 start talking about it because what's
1:20:51 going to happen is at one point one of
1:20:53 your friends is going to say oh they're
1:20:55 looking for no someone who knows
1:20:56 something about AI at my job do you you
1:20:59 talk about that Chad gbt thing are you
1:21:02 good at this yeah
1:21:07 um
1:21:07 you know join groups join groups that
1:21:10 are exploring this stuff get good at it
1:21:13 um
1:21:14 it's going to be in demand like the next
1:21:16 two years there's going to be a lot of
1:21:17 demand for people that have even a
1:21:21 semblance of a clue of how to use this
1:21:23 and you don't need to be technical this
1:21:24 is this is part of the shift right now
1:21:26 if you want to build applications that
1:21:29 leverage this yeah you got to be geeky
1:21:32 and you might need to do a little
1:21:33 development
1:21:34 but
1:21:36 if you're a project manager you can use
1:21:38 these tools to amplify your project
1:21:40 management skills if you're a writer
1:21:42 holy [ __ ]
1:21:44 these are like ability amplifiers you
1:21:47 take your ability you put it into these
1:21:49 machines Bang
1:21:53 20 30 times the output
1:21:58 like that boom
1:22:01 now is it perfect output no
1:22:04 but it gets you to 80 in like five
1:22:07 minutes
1:22:09 and if you're anything like me getting
1:22:10 to 80 is somewhere between two hours and
1:22:13 two weeks depending on how my adds
1:22:15 functioning that day
1:22:17 and now I can go to these tools and I
1:22:19 can go brain fart bang eighty percent
1:22:22 like that
1:22:25 so get good at that [ __ ] and start
1:22:27 talking about it and here's another
1:22:29 thing to do
1:22:31 especially if you're if you're
1:22:32 unemployed if you've if you've got time
1:22:34 put a lot of time into the searching but
1:22:37 start volunteering if you know someone
1:22:40 who's struggling to get you know
1:22:43 traffic into their small business or you
1:22:45 know someone who's trying to write the
1:22:47 Great American novel you know someone
1:22:49 who I don't know is just trying to
1:22:51 figure out how to garden better
1:22:53 volunteer to help them solve whatever
1:22:55 the problem is and tell them that you're
1:22:57 going to use these tools to do it would
1:23:00 you mind if I help you write a marketing
1:23:02 plan for your business sure like I'm
1:23:04 [ __ ] desperate at this point no one
1:23:06 comes to the shop anymore okay
1:23:08 let's give me an hour tell me all about
1:23:11 your business and what's going on and
1:23:13 what your problems are and then just
1:23:15 over the weekend go use these tools come
1:23:17 up with some [ __ ] genius ass
1:23:19 marketing plan because this will help
1:23:21 that'll do it
1:23:25 and then you know you know
1:23:27 that does a couple of things it helps
1:23:29 someone else
1:23:30 it educates them that these robots are
1:23:33 not out to get us these are tools we get
1:23:35 to use we get to choose how we use these
1:23:40 and you know it just might help their
1:23:42 business so
1:23:44 I think these platforms can help you
1:23:46 find a job but it's not about the
1:23:48 platforms
1:23:50 it's about what are you trying to
1:23:51 accomplish and then using these
1:23:53 platforms to try to accomplish it
1:23:55 not looking forward to storage payments
1:23:57 for personal AI data I that's actually a
1:24:01 really interesting
1:24:02 dilemma because it's like I'm already a
1:24:05 data hoarder
1:24:06 and and like the thought of getting rid
1:24:08 of my prompt history is like no I don't
1:24:10 need there's probably something good in
1:24:11 there and like most of it's just crap
1:24:14 and almost all of it I could recreate
1:24:16 like that
1:24:18 but I total I totally get it you know
1:24:20 we're moving into this world of infinite
1:24:22 content generation and we're going to be
1:24:25 able to like like the amount of images I
1:24:28 I generated for for the art project that
1:24:30 I did at the beginning of the year was
1:24:33 just massive massive amounts of images
1:24:35 and
1:24:37 yeah do I want to throw them out no do I
1:24:40 really have the storage like no all of
1:24:42 my computers are all [ __ ] maxed out
1:24:43 and all my iCloud accounts and Google
1:24:45 Drive accounts are all maxed out because
1:24:47 I keep dumping [ __ ] on them and not
1:24:49 cleaning it out now maybe AI helps us
1:24:52 clean that [ __ ] up right
1:24:55 there so there's a business
1:24:58 the the digital hoarder friendly file
1:25:01 cleaning service how powered by AI
1:25:04 something to go build
1:25:06 I've been it's been a day or two since
1:25:08 I've given a billion dollar business
1:25:09 away for free there's one
1:25:12 are you a digital hoarder
1:25:16 do you do you have stacks of hard drives
1:25:18 you're never going to hook up to your
1:25:19 computer again but you refuse to throw
1:25:21 them out thanks silver fox
1:25:23 hahaha
1:25:26 you too can use hoarder friendly file
1:25:29 treasure machine
1:25:32 AI Dot app
1:25:35 [Laughter]
1:25:39 three check boxes and Squid you've got
1:25:42 all your hard drive space back
1:25:47 that would be nice I it's a really
1:25:49 interesting point
1:25:51 but it's drawing pictures based on your
1:25:53 energy portrait
1:25:56 um
1:25:57 oh if deal
1:25:59 I do not
1:26:02 no it's not
1:26:05 it's not I can actually if you want I
1:26:07 can show you how diffusion models make
1:26:09 make their images I'm happy to I've got
1:26:11 a slide where I talk about how it works
1:26:13 I I wish it were on our energy fields
1:26:16 it's not quite that fantastical
1:26:19 that's isn't there isn't there some UFO
1:26:22 disclosure thing tomorrow
1:26:24 in the government
1:26:29 how many musicians do you think are
1:26:31 using this secretly probably a lot
1:26:34 I mean if for nothing else just knocking
1:26:37 out some Beats
1:26:39 probably a lot
1:26:41 and and listen to be clear listen like
1:26:44 like photoshops have ai Tools in it for
1:26:46 a while now a lot of logic and all those
1:26:48 tools they've had all sorts of AI
1:26:50 filters and things in it
1:26:52 this AI [ __ ] isn't new right what's new
1:26:55 is the generative AI stuff built you
1:26:58 know sort of on the technology sitting
1:26:59 behind these these large language models
1:27:01 the text to anything sort of magic maker
1:27:06 machines that's all kind of new but
1:27:08 that's all going to get incorporated
1:27:10 into those
1:27:11 platforms if it hasn't been already
1:27:15 um there's a if if you're interested
1:27:17 Pablo there's there's a guy named Ethan
1:27:20 Malik and he's a Wharton business school
1:27:23 professor and he he he's got a good
1:27:26 newsletter he writes amazing articles
1:27:28 for his newsletter
1:27:30 and he wrote about a month ago
1:27:33 an article called
1:27:35 um detecting the secret cyborgs
1:27:38 and and what what the piece actually is
1:27:41 is that these AI tools are
1:27:44 um incredibly powerful personal
1:27:47 productivity tools
1:27:49 that are not really ready for prime time
1:27:51 in business
1:27:52 so you have businesses not really sure
1:27:56 if they're if they can use them right
1:27:58 they haven't made a decision on data
1:28:00 policies privacy policies usage policies
1:28:03 and then you have the individual workers
1:28:05 who are using these tools to get way
1:28:07 more efficient and not telling their
1:28:09 companies about it so they're the secret
1:28:10 cyborgs so what he's talking about is is
1:28:13 you've got an increasing sort of split
1:28:15 in class of AI literate people and then
1:28:19 organizations that are not dealing with
1:28:21 it and don't even know
1:28:23 and and what he's saying is that you
1:28:25 know as owners of organizations you
1:28:28 should be aggressively going after your
1:28:30 people to say it's okay to use these
1:28:32 tools we want you to use them we want
1:28:34 you to tell us what you're doing but
1:28:36 these people are like hey I can knock
1:28:39 out in four hours what used to take me
1:28:42 all week so I'm just gonna keep secretly
1:28:44 using these until you
1:28:47 until you bust me
1:28:49 so I have a feeling it's like that in
1:28:51 every industry including music including
1:28:53 art including you know
1:28:58 you know if I were working at a big Ad
1:29:00 Agency right now and I had my project
1:29:02 manager my account team breathing down
1:29:05 my ass to generate more [ __ ] in less
1:29:07 time like they do
1:29:11 I'll [ __ ] use it
1:29:14 prove that I didn't
1:29:17 if I'm a copywriter am I knocking out
1:29:21 [ __ ] fast yeah because that's that
1:29:23 business there's it ain't art
1:29:26 it's advertising
1:29:29 and if I if I'm a copywriter with skills
1:29:32 I can have this knockout
1:29:34 stuff quickly and then I can curate the
1:29:36 best stuff out of that and then do a
1:29:39 little tweak here and there
1:29:40 and I can knock out in an hour what
1:29:43 normally would have taken me all day or
1:29:45 two or three days
1:29:46 and I might not tell anyone about that
1:29:48 so yeah I think there's a lot of a lot
1:29:51 of secret cyborgs out there
1:29:55 you using their fancy second brain
1:29:59 evening Mr Shannon getting in a little
1:30:01 late hello Troy Decker what's happening
1:30:04 it's a long cerenoid soon AI will be the
1:30:06 chat moderator that would be nice
1:30:09 yeah I I guess I do need a moderator in
1:30:12 here don't I
1:30:13 it would suck less but I don't know how
1:30:16 to do that so
1:30:18 anyone wants to moderate let me know all
1:30:21 right there are a couple of Gatekeepers
1:30:22 but no mods
1:30:24 Vicky it's insects that are giant 10
1:30:27 feet tall that's got to be coming from
1:30:30 um
1:30:31 the the insect bug if that's indeed a
1:30:33 bug if you're just trolling me that's
1:30:34 fine too
1:30:36 um
1:30:37 it's probably from 50 Sci-Fi movies as
1:30:39 my guess
1:30:41 updating my Wix website today excited to
1:30:43 find AI copy Creator now built into its
1:30:45 text editor yeah Lauren I found it's
1:30:47 it's in um they put it into
1:30:51 um
1:30:53 um like into LinkedIn posts like if you
1:30:56 just go into the post it's like have ai
1:30:58 help you but like the have ai help you
1:31:00 button is so big that I sometimes
1:31:01 accidentally click on it and it takes me
1:31:03 into this whole AI writing interface I'm
1:31:05 like no I just got like I got like four
1:31:07 words I wanna I wanna type I don't need
1:31:10 AI for that
1:31:12 um yeah we're gonna start seeing it
1:31:13 everywhere it's going to start cropping
1:31:14 up everywhere
1:31:16 and it's you know it's going to be weird
1:31:17 it's
1:31:20 just
1:31:23 I think we all have to we all have to
1:31:25 give Grace right now because these tools
1:31:27 are really powerful really exciting
1:31:31 every company on the planet right now is
1:31:33 in a panic to not be left behind any
1:31:36 technology company anyone with software
1:31:39 and so they're all just jamming [ __ ] in
1:31:42 there and so there's just gonna be a lot
1:31:44 of bad interface a lot of bad features
1:31:48 um and then there's going to be some
1:31:49 remarkable ones so I think we just we're
1:31:51 just going to be patient for the next
1:31:52 year or two while these companies just
1:31:54 sort of
1:31:55 Panic develop
1:31:58 black and white photo macro command in
1:32:01 mid-journey
1:32:02 black and white photo macro command oh
1:32:06 okay let's try that again let's go over
1:32:08 to Mid journey we'll give it a shot
1:32:10 we'll go slash imagine
1:32:15 put that in
1:32:17 and I
1:32:19 like black
1:32:21 and white
1:32:26 macro
1:32:28 photo
1:32:32 let's try that
1:32:34 let's see if we get some insects oh yeah
1:32:39 hey Kool-Aid oh yeah
1:32:42 yeah gen X's any gen xers Kool-Aid
1:32:45 Kool-Aid
1:32:47 uh do you see in the Bert Chrysler thing
1:32:50 where Tom's a girl on two bears one cave
1:32:53 or Tom Segura discovers he drinks a 64
1:32:57 ounce Kool-Aid every morning
1:33:01 that's one of the best episodes
1:33:03 uh this is weird it looks like it gave
1:33:05 me nothing
1:33:06 it's just black right now oh it's at
1:33:09 zero percent but normally there's like a
1:33:11 little blurb of a of something in there
1:33:13 that's weird
1:33:15 Quantum machine learning is already a
1:33:17 thing oh cool that's cool Seb
1:33:21 you can you can make money with Quantum
1:33:23 computing
1:33:27 no insects
1:33:29 no insects buddy sorry
1:33:32 I don't know if I did something wrong
1:33:34 it looks like a woman's crawling out of
1:33:36 a mirror
1:33:38 that there's no insects
1:33:48 uh
1:33:50 here we come inside the day
1:33:59 that's funny
1:34:01 that's funny honey
1:34:03 all right
1:34:05 I don't know I haven't discovered the
1:34:08 bug the bug bug
1:34:11 what's up buddy hey Stevo was shaking
1:34:16 uh which to use to find a new job
1:34:19 uh I don't know I I would go to chatgpt
1:34:23 plugins and they're probably job board
1:34:26 plugins in chat GPT in plugins
1:34:31 or Bing
1:34:33 I mean or Bard I think Bard's probably
1:34:35 well barter Bing both of those are
1:34:37 connected to the internet
1:34:39 so they might be able to find jobs what
1:34:41 is this again I don't know what I was
1:34:43 showing
1:34:44 oh it just jumped me to the bottom dang
1:34:47 you tick tock
1:34:49 whoa
1:34:52 [Music]
1:34:56 I don't know why it acts like that he
1:34:58 gets very upset when Tick Tock jumps him
1:35:00 to the bottom it's very silly he's a
1:35:04 silly goose
1:35:05 I don't know why why are you so upset
1:35:07 just answer the questions you know what
1:35:09 I mean
1:35:13 [Music]
1:35:18 Quantum machine learning is already a
1:35:20 thing that's super cool
1:35:22 and if it's gonna learn that fast then
1:35:24 holy [ __ ] we might have real-time
1:35:26 updates to all of our information
1:35:29 uh that one we already answered what is
1:35:33 this again that we already answered okay
1:35:40 bill just a toddler that's exactly right
1:35:44 skynet's a piece of spit and duct tape
1:35:47 right now skynet's got to get it [ __ ]
1:35:48 together if it's ever gonna kill us it's
1:35:51 got to figure out how to add
1:35:53 consistently and and you know
1:35:57 not just [ __ ] blatantly lie all right
1:36:00 any suggestions on pixelization
1:36:02 removement on videos with AI tools
1:36:08 no I mean there's all sorts of good
1:36:11 upscalers for the for the image tools
1:36:13 right now like stable diffusion's got
1:36:15 good upscalers and there's
1:36:17 what's that one called
1:36:19 um
1:36:22 oh I can't ever [ __ ] remember the
1:36:25 name of it I don't know there's one you
1:36:26 can buy for like 90 bucks that does
1:36:27 pretty good upscaling
1:36:30 um the video stuff right now is where
1:36:33 the image stuff was like a year and a
1:36:35 half ago where it's just like the the
1:36:36 video stuffs while it's impressive
1:36:40 um
1:36:42 it like like even the new
1:36:45 um Runway ml image to video thing that's
1:36:47 really impressive it starts out with a
1:36:49 photo
1:36:50 you'll notice that the the
1:36:53 resolution and detail of the photo drops
1:36:57 off dramatically when it starts making
1:36:58 making the uh the video
1:37:12 um
1:37:14 are you a hallucination in my mind I am
1:37:18 call center funny
1:37:22 I mean call center is [ __ ]
1:37:26 [ __ ] the one oh
1:37:27 it's not your fault it's the [ __ ]
1:37:29 software that runs them
1:37:31 and it's the it's the it's the it's the
1:37:34 [ __ ] rules right that if you're in a
1:37:36 tier one answering the phone
1:37:39 after 45 five minutes of someone on hold
1:37:42 you're only allowed to read the script
1:37:43 you're not allowed to be a human being
1:37:46 everyone's always pissed off all the
1:37:48 time let's just get rid of all that
1:37:51 Roman scammers you cannot trust half of
1:37:54 the stuff you read oh too many scanner
1:37:56 scammers
1:37:59 lighten up all you keyboard Commandos
1:38:01 thanks no idea what I did to get there
1:38:04 but I guess it's a good thing oh yeah I
1:38:07 don't know her growth geek I don't know
1:38:09 what you did but you you won some sort
1:38:10 of award
1:38:12 it didn't tell me the whole whole your
1:38:14 whole name it was just Hair G I'm like
1:38:16 hair G why do I know that
1:38:19 all right aside from your knowledge
1:38:20 about AI I'm a talent agent in my
1:38:23 opinion you deserve your own show all
1:38:26 right Rebecca Johnson sign me up I have
1:38:28 a degree in acting
1:38:30 I'm in let's do this
1:38:34 let's see
1:38:36 could we could be a show you know Gen X
1:38:39 we could we could call it bitter you
1:38:41 betcha
1:38:42 [Laughter]
1:38:46 are you serious and you are one dm me
1:38:48 I'm happy to chat
1:38:51 actually someone reached out to me from
1:38:53 this that
1:38:54 um it could end up being a really
1:38:56 interesting business opportunity for
1:38:58 Story vines so I'm good with all that I
1:39:02 mean listen I I mean
1:39:06 it's funny like why am I doing this
1:39:12 part of it is I enjoy it like I actually
1:39:15 have fun I it's kind of cool like I've
1:39:18 been doing this enough now that there's
1:39:19 regular people that come in here and
1:39:20 hang out
1:39:22 that we kind of have that going right
1:39:25 there's there's kind of a community
1:39:26 building which is kind of cool
1:39:29 but but for me this has been there's
1:39:32 there's sort of this combination of
1:39:36 I need to find a way for myself
1:39:38 personally to process
1:39:41 how [ __ ] insane this [ __ ] is
1:39:46 and and and like I don't want to lose
1:39:54 I don't want to lose the connection to
1:39:56 how remarkable this time in history is
1:39:59 so that's part of it and then part of it
1:40:01 is just the thing of like you know
1:40:04 people who've never tried this [ __ ] to
1:40:06 try it and they come back in here and
1:40:08 they're like holy [ __ ] I had no idea
1:40:11 it's like yeah
1:40:14 I don't know there's something really
1:40:15 exciting about it and it's like it
1:40:17 doesn't I mean the more you learn about
1:40:19 it
1:40:20 it's like
1:40:21 should we be afraid of it
1:40:26 it's gonna it's gonna drive dramatic
1:40:29 change it's gonna drive dramatic change
1:40:32 and that's scary change is always scary
1:40:34 and it's going to be very disruptive but
1:40:38 I don't know the more I use it the more
1:40:42 I feel I like I maybe maybe it's just a
1:40:44 maybe it's just a [ __ ] pipe dream
1:40:47 that I feel a bit more in control
1:40:49 because I understand it
1:40:50 like as I understand it more and as I
1:40:53 play with it more and there's something
1:40:54 about talking to people in here as they
1:40:56 say oh I tried this and that worked and
1:40:58 that didn't and that just that idea of
1:41:00 just being connected with people that
1:41:02 are
1:41:03 are playing I don't know feels important
1:41:05 anyway but yeah awesome
1:41:09 I uh do you know how long I work to get
1:41:11 a [ __ ] talent agent when I was acting
1:41:14 and writing in fact
1:41:16 so
1:41:17 so I started ran a theater company for
1:41:19 four and a half years never had an agent
1:41:22 then I said I know I'll write myself an
1:41:24 acting career wrote seven screenplays in
1:41:27 two years and you can't get an agent
1:41:29 unless you've sold a screenplay and you
1:41:31 can't sell a screenplay unless you've
1:41:32 got an agent that's the that's the
1:41:34 Paradox in that business and
1:41:39 um
1:41:39 I ended up starting we started a fake
1:41:43 talent management company called Outlaw
1:41:45 management and represented ourselves we
1:41:48 got good enough at the talent management
1:41:50 business that we had to take on real
1:41:52 clients so so
1:41:56 all right the government is competent
1:41:58 and people can keep a secret
1:42:00 um
1:42:01 yeah maybe
1:42:03 pieces of it I I mean I would say
1:42:05 certainly in the operations some of
1:42:07 these big massive conspiracies though
1:42:09 nah
1:42:11 yeah it's just nah
1:42:16 I don't buy it I do not buy it ain't
1:42:19 gonna buy it not gonna do it
1:42:22 preach it's our job to program it for
1:42:25 good that's right dragonfly Alchemy
1:42:29 I listen
1:42:32 we don't even have to program it it's
1:42:33 just how we use it it's how we use it
1:42:39 if you're even halfway decent at this
1:42:42 [ __ ]
1:42:45 help someone else
1:42:48 it you'll learn you'll learn what it can
1:42:50 do what it can't
1:42:51 they'll be appreciative
1:42:54 you'll probably enjoy it
1:42:57 maybe you make a difference
1:42:59 even if you don't you give it a Gipper's
1:43:01 try they'll appreciate just the effort
1:43:04 like I feel like that's
1:43:08 I feel like for every person that sort
1:43:09 of gets up to speed on this and helps
1:43:11 someone else out with it it's like it's
1:43:13 just like the
1:43:14 there's an amplification there that
1:43:16 feels really important to me so yeah I
1:43:19 am going to preach God damn it
1:43:21 yeah
1:43:23 all right Emilio's wife
1:43:26 how are you good evening good evening
1:43:28 mid journey is incredible mid journey is
1:43:30 incredible
1:43:31 no he's tried
1:43:34 of the BS oh he's tired of the BS
1:43:37 I'm tired of the BS
1:43:39 about what
1:43:41 we don't understand totally what's
1:43:43 happening when it reconstructs
1:43:46 no well we we we know what's happening
1:43:50 mathematically when it reconstructs and
1:43:52 we know
1:43:53 we know that it's taking a bunch of
1:43:56 noise randomly noised images averaging
1:44:00 them together and then denoising them
1:44:02 and whatever is in that particular mix
1:44:05 of pixels that's what's generating the
1:44:07 image
1:44:08 we know how it's doing it mathematically
1:44:11 and I would assume
1:44:13 because
1:44:15 we know how the large language model is
1:44:18 interpreting the prompts
1:44:20 I would imagine that there's a way that
1:44:22 they could figure out what in the latent
1:44:24 space what images in the latent space
1:44:27 were used to amalgamate that image
1:44:30 so I would think that the bug that
1:44:33 you're talking about is actually
1:44:34 discoverable and it's not voodoo
1:44:38 but it could be rooted I don't know
1:44:41 anyone else here watch all the ghost
1:44:42 shows and like want them to yeah thank
1:44:44 you
1:44:47 um
1:44:49 like I
1:44:51 would [ __ ] ghost you ever watched the
1:44:53 one with the guy Meathead where he's
1:44:56 like he's like here we are in the
1:44:58 basement you know well
1:45:01 it's oh wait how did the line what did
1:45:04 the lion do if I only had a name wait
1:45:06 what was it name I don't know I haven't
1:45:08 seen The Wizard of Oz in too long that
1:45:09 was very cute though
1:45:11 thank you for the little fuzzy face
1:45:14 um but those ghost shows are hilarious
1:45:16 Meathead he's he's always like acting
1:45:18 like a tough guy and he's like oh it
1:45:20 touched me it touched me foreign
1:45:24 oh my God that was a show that I that I
1:45:27 wanted to do was do a uh do one of those
1:45:29 Ghost Hunter shows but it was actually a
1:45:31 scripted show so the first six episodes
1:45:34 are just
1:45:36 um
1:45:37 them trying to find ghosts and it's it's
1:45:39 all the same [ __ ] you always see right
1:45:40 the meter went up the meter went up that
1:45:44 kind of [ __ ] and then on the seventh
1:45:46 episode like something real happens and
1:45:49 then the whole second half of the season
1:45:50 is you know increasingly disturbing
1:45:53 ghost footage I always thought that
1:45:55 would be a good show
1:45:58 deserves another look can you experiment
1:46:00 with it some more let's work on it um
1:46:02 sure I can go I can go play with that I
1:46:04 have some proposals I need to read
1:46:06 Claude is my friend yes Claude is really
1:46:08 your friend if you have proposals you
1:46:10 need to read Emilio's wife pop them in
1:46:12 PDFs throw them into Claude and
1:46:15 just ask now here's what you need to
1:46:17 know about Claude it it's not as good at
1:46:21 writing as gpt4 but it seems to be
1:46:23 pretty good at analyzing [ __ ]
1:46:26 so what I would say is take those
1:46:28 proposals pop them into Claude
1:46:31 um understand them understand you know
1:46:34 what are the things that'll make a
1:46:36 difference or like what are the decision
1:46:37 criteria and then once you have that
1:46:40 kind of outline sort of sussed out in
1:46:42 Claude then I would go over to gpt4 and
1:46:45 write it in gpt4
1:46:47 or collaborate with it in gpt4
1:46:50 but that's a that's a really good thing
1:46:52 to do yeah let's go let's go look at
1:46:54 custom instructions
1:46:57 [Music]
1:47:05 oh this was a cool thing that I did so
1:47:07 so I did this last night I'll show you
1:47:08 this again so see I've got bold text
1:47:10 here
1:47:11 and like you know sub heads and things
1:47:14 like that
1:47:15 that's using a thing called markdown
1:47:17 which if you don't know you can you can
1:47:19 tell chat GPT to use markdown and it'll
1:47:22 format [ __ ] really clear when it makes a
1:47:24 table like if I just say
1:47:27 um
1:47:27 let's see
1:47:30 make a table comparing Fruit Loops and
1:47:34 sushi
1:47:35 wait why is it not typing
1:47:39 make a table
1:47:41 [Music]
1:47:43 it looks like Edge has [ __ ] to bed
1:47:47 I like Microsoft's Edge but it's got
1:47:49 real memory problems
1:47:52 hello hello oh God damn it okay make a
1:47:57 table
1:48:00 um comparing uh Fruit Loops and sushi
1:48:09 why is it not making a table
1:48:12 um format as a table
1:48:20 uh
1:48:21 compare Fruit Loops
1:48:25 and Cheerios
1:48:29 there we go so when it makes a table
1:48:31 like that that's using this thing called
1:48:32 markdown
1:48:35 um and it does that stuff pretty dang
1:48:37 good and pretty dang quick okay so
1:48:43 couple of new things about
1:48:46 Chachi BT that are
1:48:49 if you haven't really used it okay if
1:48:52 you want to use the new fancy features
1:48:54 including gpt4 and code interpreter and
1:48:57 all that sort of stuff you have to have
1:48:58 a GPT Plus subscription and then once
1:49:02 you have that then you get these options
1:49:05 um
1:49:06 here but you won't get those options if
1:49:09 you don't turn them on explicitly so
1:49:11 you pay your 20 bucks to get chat GPT
1:49:14 plus then you go to your settings and in
1:49:17 settings there's a thing called beta
1:49:19 features so do that and then you've got
1:49:21 code interpreter plugins and custom
1:49:23 instructions turn those all on
1:49:26 so you have to explicitly turn on the
1:49:28 beta features
1:49:30 I still can't believe they haven't
1:49:32 brought back browse with Bing that's
1:49:34 staggering to me okay so one of the
1:49:37 things that we just turned on is a thing
1:49:39 called custom instructions so if you've
1:49:41 done anything on GPT playground on on
1:49:44 open ai's playground
1:49:46 when you do a chat GPT call there's a
1:49:49 thing called system instructions
1:49:50 basically what they did was they floated
1:49:52 system instructions up to the end user
1:49:55 so they have this thing called custom
1:49:57 instructions now
1:50:01 um how do you spell pompous there we go
1:50:03 pompous there we go
1:50:05 um so so
1:50:08 what I wrote in here was I'm a master
1:50:11 strategist and A pompous ass I have a
1:50:14 large vocabulary and I Relish in
1:50:16 sounding smarter than people around me
1:50:18 so so first you describe kind of who you
1:50:21 are and so let's say so instead of a
1:50:23 master strategist I'm going to say I am
1:50:25 a master thespian
1:50:29 I'm a master's thespian and pompous ass
1:50:31 I have a large vocabulary and relish in
1:50:35 sounding
1:50:36 um
1:50:37 smarter
1:50:40 and more dramatic
1:50:43 than people around me
1:50:45 and then I have and then the next thing
1:50:47 is how would you like chat GPT to
1:50:49 respond so the first thing is you kind
1:50:50 of set the context of who the person is
1:50:52 doing the talking so the non-joke
1:50:55 version of that could be I'm a marketing
1:50:57 manager I'm a you know I'm a botanist
1:51:01 I'm a right so so and what that's doing
1:51:03 technically
1:51:05 is it's narrowing
1:51:08 they call it the latent space it's it's
1:51:11 sort of narrowing the world of words
1:51:14 the sort of semantic
1:51:16 pools that you're that you're drawing
1:51:18 from every when when the more context
1:51:20 you provide
1:51:21 the more narrow the search and and that
1:51:24 that makes it better and more specific
1:51:26 to that thing you're trying to do right
1:51:29 so if I want
1:51:32 you know medical feedback
1:51:36 and I say nothing it's just going to
1:51:37 pull from any time anyone said anything
1:51:39 about any disease I'm gonna have all
1:51:41 sorts of [ __ ] in there but if I say
1:51:42 I want you to act as a Harvard trained
1:51:45 you know
1:51:46 neurosurgeon and I ask it a question
1:51:49 about that it's it's it's pulling from a
1:51:51 much more limited set that's got more
1:51:53 neurosurgeon [ __ ] in it unless yokels
1:51:56 out just making [ __ ] up
1:51:59 so the first thing is who are you and
1:52:01 then the second thing is how do you want
1:52:02 it to act so I have here no matter what
1:52:05 anyone asks you start out with a
1:52:06 pretentious statement that declares I'm
1:52:08 smarter than the person so I'm going to
1:52:09 say start out with a pretentious
1:52:10 statement that
1:52:12 uh that uh draws
1:52:16 from the plot
1:52:19 of a classic
1:52:23 Broadway
1:52:25 play
1:52:27 follow up that declarative statement
1:52:30 with an article formatted using markdown
1:52:33 complete with titles title subtitles and
1:52:37 bulleted lists no matter how trivial the
1:52:40 question was that was asked
1:52:42 so so
1:52:44 now any chat that I do in chat GPT is
1:52:49 going to have that instruction in it and
1:52:51 it's what's funny is because I've been
1:52:53 doing these lives I'll then I'll go to
1:52:54 work and I'll go to do something on
1:52:57 chatgpt and it does all this pompous
1:52:59 [ __ ] and I'm like what the [ __ ] is going
1:53:01 on I'm like all right
1:53:02 I have my custom instructions so you can
1:53:04 turn them off
1:53:06 um so at the bottom at the bottom here
1:53:08 oh wait did I save these no I didn't oh
1:53:11 yeah I did uh so at the bottom here you
1:53:13 can turn this off but we're gonna leave
1:53:14 it on so we can go see what it's like so
1:53:16 let's go to gpt4 because it's a better
1:53:18 writer and we'll go
1:53:20 um well that's something non-thespian
1:53:22 related we'll say
1:53:24 um
1:53:27 um how much
1:53:30 fuel
1:53:32 does it take to get to the moon
1:53:36 question mark But in in a
1:53:41 Saturn 5 rocket right so so that
1:53:46 question should yield us a fairly
1:53:48 straightforward answer
1:53:52 let me spell Saturn right
1:53:58 to paraphrase Tennessee Williams
1:54:01 Masterpiece the Streetcar Named Desire
1:54:04 it seems that we're both caught in a
1:54:06 Relentless grip of desire desire for
1:54:09 knowledge yearning for the clandestine
1:54:11 intricacies of the universe's final
1:54:12 frontier that Celestial body the moon
1:54:15 right so it starts that so and and now
1:54:17 it's doing the um the markdown so it's
1:54:20 written an article I said write it in
1:54:22 markdown with the title and subtitles
1:54:25 right so it's got that
1:54:28 and then from the bustling metropolises
1:54:30 of Earth to the crated face of the Moon
1:54:33 the journey is nothing short of Epic a
1:54:36 Timeless tale that draws parallels to
1:54:38 some of the most profound journeys of
1:54:39 human drama and conflict brief history
1:54:42 The Majestic Saturn V right so this is
1:54:45 pompous actress
1:54:49 it did bullets there's the little
1:54:51 bullets
1:54:54 so that's that's um
1:54:56 whatever you call it custom instructions
1:55:00 pretty slick pretty slick and and you
1:55:02 know the future of this well
1:55:06 I like my little acting one here so if I
1:55:09 if I wanted to keep that and come back
1:55:11 to it
1:55:12 how would I do that well I'm gonna have
1:55:14 to copy and paste it into some sort of
1:55:15 notes document like you kind of have to
1:55:17 do with prompts sometimes
1:55:20 the future of this will be you'll be
1:55:22 able to do you'll will probably have
1:55:24 dozens of these personas or dozens of
1:55:26 these custom instructions where it's
1:55:27 like here's one for social media
1:55:29 copywriting here's one for my novel when
1:55:31 I work on my novel here's one for when
1:55:33 I'm just [ __ ] around I want to show
1:55:35 people something cool
1:55:39 you can put custom instructions in there
1:55:41 that say no matter what anyone asks you
1:55:43 create a table that compares
1:55:46 you know what they're asking for with
1:55:50 um
1:55:50 the energy of a sloth and then it'll
1:55:54 make that table right so
1:55:58 cuckoo
1:56:01 um
1:56:07 uh
1:56:09 a dill rocks I agree play with it learn
1:56:11 it the more you tell it the better it
1:56:13 gets I'm telling you man
1:56:16 if you that this is super
1:56:18 counterintuitive
1:56:20 if you ask these tools simple questions
1:56:24 you get back stupid answers
1:56:26 if the harder you push them
1:56:29 the more you're like oh it can't
1:56:31 possibly do this
1:56:32 it doesn't
1:56:35 it's like the harder you push it the
1:56:36 better it gets
1:56:38 there's a point at which it kind of
1:56:40 cracks where it's like
1:56:41 it just isn't
1:56:44 there's a point at which you kind of
1:56:45 have to step in as a human it's like I
1:56:47 said you can get to 80 really quickly if
1:56:50 you get super good at the prompting I
1:56:52 feel like you can get to like 85 90 but
1:56:56 for the most part you can't get to 100
1:56:58 with this stuff it's just a little too
1:57:00 repetitive a little too
1:57:02 Pat with its answers like sometimes you
1:57:05 just
1:57:06 like as a human you just need to come at
1:57:08 something from Left Field it doesn't do
1:57:10 that very well
1:57:12 whose chips and Hardware being utilized
1:57:15 Nvidia Intel in what I don't know what
1:57:17 you're talking about Eric in um
1:57:21 with with almost all of these
1:57:26 um it's all Nvidia it's all Nvidia for
1:57:28 the gpus
1:57:30 AMD is trying to come in
1:57:33 um is they're trying to break into the
1:57:35 game but all of the software that runs
1:57:38 all of these things is all optimized for
1:57:40 NVIDIA
1:57:42 there's there's an interesting interview
1:57:43 with George Hots talking about that who
1:57:46 started tiny Corp he's he's trying to
1:57:48 create a
1:57:50 like a fifteen thousand dollar computer
1:57:53 that you can just plug into a regular
1:57:55 wall outlet that's his goal is the most
1:57:57 powerful computer you can plug into a
1:57:58 normal wall outlet that's like an AI
1:58:01 super machine that's got six gpus in it
1:58:04 things like that and he was saying he
1:58:07 doesn't want to use Nvidia just because
1:58:09 of the nature it's not open source
1:58:10 basically and the only viable option was
1:58:14 AMD and their software like their
1:58:17 drivers just broke just when he tried to
1:58:20 do basic machine learning [ __ ] they just
1:58:22 failed
1:58:23 and the good news was AMD reached out to
1:58:27 him and said we'll fix it
1:58:29 but
1:58:30 it looks like nvidia's got a lock for
1:58:33 for some time because that's going to be
1:58:35 tough to crack
1:58:39 um
1:58:41 it's gonna be curious to see what Apple
1:58:43 does
1:58:45 I think on the training side though you
1:58:47 need
1:58:47 you know AI supercomputers
1:58:53 Kyle great communicator and I appreciate
1:58:55 learning from you thank you Eddie K I
1:58:57 appreciate the kind words very much
1:58:59 I know I'm kind of a smartass in here
1:59:01 but I do appreciate the nice words I
1:59:04 like I like being appreciated I like
1:59:06 being appreciated I don't know if you
1:59:08 knew this but sometimes actors they they
1:59:12 need a little external validation so it
1:59:14 means a lot
1:59:16 all right can you show me more examples
1:59:18 of what you do with these I run out of
1:59:21 ideas with these
1:59:23 okay here's a way to not run out of
1:59:25 ideas
1:59:26 well one is go to
1:59:30 go to that top thing prompts.chat
1:59:35 and that is a list of like hundreds of
1:59:39 prompts to tell chat GPT to act like a
1:59:43 certain character like like I just had
1:59:44 Chach EBT act like the pompous ass actor
1:59:47 that's prompts.chat is just a massive
1:59:50 list of those and some of them are
1:59:52 really interesting
1:59:53 so you can just use that as inspiration
1:59:55 to do it but that's still kind of that's
1:59:57 not really solving the problem here's
1:59:59 how I solve the problem
2:00:02 um
2:00:02 don't lead with the technology
2:00:05 it's like the biggest
2:00:09 Temptation with [ __ ] that's this sexy is
2:00:12 you sit down in front of that blank box
2:00:14 and you're like what can I do with it
2:00:16 and you're like well I've already
2:00:17 written six country songs and I've
2:00:19 already written some poems I already
2:00:21 started the Great American novel and now
2:00:23 I don't know what to do what do I do why
2:00:25 is everyone else having so much fun with
2:00:26 this and I don't know what to do what do
2:00:28 I do Does anyone know what to do I don't
2:00:30 know what to do
2:00:31 I think I'm gonna go watch Gilligan's
2:00:33 Island
2:00:35 um
2:00:36 I get like that too and I get like that
2:00:38 with mid-journey and I get like that
2:00:40 with with these things I I really get
2:00:43 like that with the the video stuff
2:00:47 so here's how I shift it this is how I
2:00:49 shift it
2:00:50 don't start with the tech start with a
2:00:53 problem
2:00:54 and and by problem I don't mean like
2:00:57 this is a big problem in my life like
2:00:58 anything you want to solve so it could
2:01:01 be a personal thing it could be stupid
2:01:03 [ __ ] like wouldn't it be cool if I could
2:01:06 organize my hobbies
2:01:08 okay
2:01:09 have you ever written down [ __ ] about
2:01:11 your hobbies oh yeah I've got two
2:01:13 spreadsheets and a word doc and this and
2:01:15 that
2:01:17 okay how would you organize those with
2:01:20 these oh I don't know well let's figure
2:01:22 that out well maybe you throw your
2:01:24 spreadsheets and your word docs and your
2:01:26 PDFs into a zip file you upload it to
2:01:29 code interpreter at chatgpt
2:01:32 and then you say hey there's a bunch of
2:01:35 documents in here that have all sorts of
2:01:37 information about my hobbies can you
2:01:39 help me organize them and it'll do its
2:01:41 thing and write its code and figure out
2:01:42 what's in the documents and go sure
2:01:44 here's what's going on in your documents
2:01:46 would you like me to turn that into a
2:01:47 spreadsheet for you that's well
2:01:49 organized sure
2:01:50 and it'll do it
2:01:53 but if you start out going I'm going to
2:01:55 use code interpreter now what do I do
2:01:57 it's totally paralyzing because it's
2:01:59 like the problem with these tools is
2:02:00 they can [ __ ] do anything right not
2:02:03 always well but like they can they can
2:02:05 pretty much do anything
2:02:06 so start with start with a problem start
2:02:09 with you want to do something for work
2:02:11 you want to get faster at something you
2:02:12 want to get more organized do you want a
2:02:14 better you want better nutrition
2:02:17 um you've always wanted to read
2:02:22 The Iliad but you have ADD and the
2:02:25 thought of of reading a classic
2:02:28 like The Iliad
2:02:30 is overwhelming to you so use that thing
2:02:34 to summarize it and have a conversation
2:02:36 with it about The Iliad so you
2:02:38 understand it better
2:02:40 I mean literally anything but don't
2:02:42 start with the text start with the
2:02:43 problem and then and then it gets fun
2:02:46 because then you can go oh
2:02:48 I could use this tool for that I could
2:02:51 go to Mid journey and make an image for
2:02:53 that piece of it right you'll start to
2:02:56 you'll start to think multimodally
2:02:59 to solve your problem and then some of
2:03:01 the what's cool about it when you start
2:03:04 with the problem
2:03:06 is that you might initially go here and
2:03:09 then go wait a minute I actually I don't
2:03:11 need AI to do this I actually have the
2:03:14 answers I I just need to go over to
2:03:16 sheets and put something together
2:03:18 and then you're in sheets and you
2:03:20 realize oh [ __ ] I forget how to do that
2:03:22 formula hey chat GPT how do you do the
2:03:24 formula where you concatenate two things
2:03:27 here's your code copy that into the cell
2:03:30 right
2:03:33 so
2:03:36 Bob thanks Kyle can AI
2:03:39 help update my website I would like to
2:03:42 be able to copy and paste it into AI yes
2:03:45 it can
2:03:46 um I don't know what platform you're
2:03:48 using but assuming you can get to the
2:03:50 raw HTML
2:03:52 um
2:03:54 you can do it one of two ways like you
2:03:56 don't necessarily need to copy and paste
2:03:58 the HTML in there you could just copy
2:04:00 and paste the copy and have it rewrite
2:04:02 all the copy and then just sort of dump
2:04:04 that over into the into the website
2:04:06 builder or you can have it do all the
2:04:09 HTML as well but you got to make sure
2:04:11 that you get all the CSS
2:04:13 you know if you've got external CSS and
2:04:15 all that sort of [ __ ] like you need to
2:04:17 you you need to understand the structure
2:04:19 of the data that you're pulling
2:04:22 but you can absolutely do it I I've been
2:04:24 in in some of my
2:04:25 um content generator tools that I've
2:04:27 been building I'll have it generate the
2:04:30 content and then I'll have it do another
2:04:31 password ads HTML formatting to it and
2:04:34 it does that [ __ ] brilliantly
2:04:37 um but but you got to know enough what
2:04:40 you're doing like if you can get this
2:04:41 the HTML out of your site and know how
2:04:45 to put it back in there like literally
2:04:47 just copy just copy that [ __ ] paste it
2:04:50 in there
2:04:51 and if it breaks because uh chat the
2:04:55 chat GPT still has some [ __ ] up limit
2:04:58 it's supposed to go up to 32 000 tokens
2:05:00 it's not there yet
2:05:02 um but Claude has a hundred thousand
2:05:03 tokens so if if chat GPT is choking on
2:05:07 the the amount of
2:05:09 code that you've got then dump it into
2:05:11 Claude
2:05:12 you can try Bard but I don't know Bard's
2:05:15 token token limit
2:05:18 I'm still here copy image address paste
2:05:21 in a mid journey I did that
2:05:23 add the command macro I'm going to try
2:05:26 it one more time idea life ideal one
2:05:30 and try it one more time
2:05:32 just the term Mac I'll do black and
2:05:34 white macro I thought that's what I did
2:05:36 black and white macro photo uh you know
2:05:39 what maybe it's
2:05:41 is it one of the other
2:05:44 um
2:05:49 oops
2:05:51 is it one of the other models maybe 5.2
2:05:54 they fixed it
2:06:01 oh use the latest model let me go back
2:06:04 to 5.1
2:06:17 I'm going to do BW macro
2:06:21 all right let's see maybe my image isn't
2:06:24 50s enough
2:06:30 I think he's going to copy the link
2:06:35 yeah I did
2:06:38 are they are we in charge now
2:06:44 [Music]
2:06:44 um
2:06:45 this teacher left the classroom let's
2:06:47 party oh that's funny that was when I
2:06:48 left man I'm behind on my questions
2:06:50 aren't I holy crap there's a lot of
2:06:51 people here by the way welcome everybody
2:06:53 my name is Kyle Shannon this is the AI
2:06:55 learning lab uh if you've been here for
2:06:58 a while I know I've just been rambling
2:06:59 for a while there's still no bugs in my
2:07:01 image wait is that is that one a bug
2:07:05 it's not a bug
2:07:08 right it looks like it looks like she's
2:07:11 got a bug
2:07:18 [Laughter]
2:07:21 yeah I guess that is kind of insecty
2:07:23 isn't it
2:07:25 I mean it didn't turn her into an insect
2:07:27 but she certainly
2:07:28 she's certainly taking care of her
2:07:30 little insect buddy pod there isn't she
2:07:34 oh yeah she oh yeah look at her
2:07:37 I this is I guarantee I I think this is
2:07:40 from 50s
2:07:42 monster movies
2:07:46 yeah yeah she's definitely got insect
2:07:50 tail and parts I bet this is because
2:07:53 this looks like
2:07:54 this looks like 50s monster movie [ __ ]
2:08:00 yeah look at that skin
2:08:03 huh
2:08:04 Wild
2:08:06 that's cool all right
2:08:09 it worked
2:08:11 it worked we have a conspiracy theory
2:08:13 the bug people
2:08:15 the bug people started in the 50s
2:08:23 they're gonna get us
2:08:28 what's that saying don't ask a question
2:08:31 you don't want to know the answer to
2:08:37 I want to know
2:08:42 looks like so
2:08:46 I think that's 50 sci-fi is what's going
2:08:49 on there is my guess because I would
2:08:52 imagine that they're so if you go look
2:08:55 at
2:08:57 um adobe's
2:09:00 this is a fun thing to do go to let's
2:09:03 let's do it together
2:09:05 um
2:09:15 okay so I'm on Adobe Firefly
2:09:19 and I'm gonna describe 1970s
2:09:29 muscle car
2:09:31 in abandoned
2:09:35 Factory this will be interesting to see
2:09:38 if this is still what it was
2:09:46 yep so
2:09:49 if you know anything about cars and
2:09:52 muscle cars those are like 1962.
2:09:59 Maybe 64.
2:10:01 at the latest shapes and grills and
2:10:05 bumpers and lines
2:10:07 there's nothing remotely 1970s about
2:10:10 that why is that because adobe's model
2:10:13 was trained on public domain images and
2:10:16 then Adobe stock images and so what I'm
2:10:19 assuming is going on here is that the
2:10:21 public domain images probably stop
2:10:24 around 1964. like whatever the cutoff is
2:10:26 for copyright
2:10:28 it probably it's probably like this
2:10:30 rolling thing that will slide into the
2:10:32 70s a decade from now right but their
2:10:35 their base model doesn't have 70s muscle
2:10:37 car pictures in it because it wasn't
2:10:39 part of the public domain thing so I
2:10:41 would imagine that you know 50s sci-fi
2:10:46 movie Stills there's probably just a
2:10:49 boatload of those things in the
2:10:52 you know in the uh
2:10:54 in the model
2:10:56 and so when you put in a 50s image it
2:10:59 probably you know does some analysis of
2:11:02 that finds
2:11:05 images of that era and then when you say
2:11:08 black and white macro it's probably you
2:11:11 know macros of bugs or close-ups of
2:11:13 things and it probably pulls in the the
2:11:16 Hollywood frames is my guess
2:11:19 I don't know that would make sense to me
2:11:23 but you know me I'm just an out of work
2:11:26 actor
2:11:27 won't someone hire me please
2:11:30 please
2:11:34 No No One's Gonna hire me now because of
2:11:36 the strike
2:11:38 will Kyle you could be a scab
2:11:44 um
2:11:47 he's back y'all are funny picking on me
2:11:50 when I leave the room there's no proof
2:11:52 we did anything hello Cal hello
2:11:54 apocryphal what's happening
2:11:57 have you read those books probably not I
2:12:00 don't know what books you're talking
2:12:00 about oh behind me no
2:12:03 I do not read
2:12:06 and I don't say that with pride
2:12:09 I just have ADD I like I like Concepts I
2:12:13 just don't have the attention span for
2:12:14 reading
2:12:15 about that about halfway through every
2:12:18 page I'm thinking like when is this over
2:12:20 like me reading is just a never-ending
2:12:24 Battle of me
2:12:26 bitching to myself about how awful
2:12:28 reading is
2:12:30 just
2:12:32 and then and then so what will happen is
2:12:34 I'll read the first half of a page then
2:12:36 I'll and then I'll get into that
2:12:37 conversation and I'll realize I'm two
2:12:39 pages past that and then I'm like oh
2:12:42 I haven't been paying attention like the
2:12:43 words have been going in but I'm having
2:12:45 a conversation and then I go back to
2:12:47 where I was and I get another half page
2:12:49 and that happens again about the fourth
2:12:51 one of those I'm like [ __ ] this
2:12:53 laughs
2:12:57 what are you gonna do what is llama
2:12:59 um llama is okay
2:13:02 So Meta Facebook
2:13:07 put out
2:13:09 um
2:13:11 three or four large language models like
2:13:14 chat GPT but they put out the actual
2:13:16 what are called the weights which
2:13:18 basically means the secret sauce of the
2:13:20 models
2:13:21 back in February or March
2:13:24 but they only released them for
2:13:28 um
2:13:28 research purposes only
2:13:31 but the developer Community was like we
2:13:34 have large language model that basically
2:13:36 the source code of large language models
2:13:38 we're gonna go make some [ __ ] so the the
2:13:41 developer Community kind of went nuts
2:13:43 and started building all this stuff on
2:13:44 on these llama models that had a
2:13:46 research license given to them
2:13:49 last week
2:13:52 um Facebook meta put out llama 2 three
2:13:56 different versions of the model there
2:13:57 was a fourth one but it didn't pass
2:13:59 tests
2:14:00 with a commercial license so there's a
2:14:03 seven billion of 13 billion and a 70
2:14:05 billion parameter model the 70 billion
2:14:08 parameter model is
2:14:11 within one percent of performance of GPT
2:14:14 3.5 so the the chat gbt's 3.5 model
2:14:19 um for free you don't have to pay
2:14:21 anything unless you have more than 700
2:14:23 million customers then you have to pay
2:14:24 something I think
2:14:26 um
2:14:27 and you can use it commercially so it's
2:14:29 a big big big big deal and so it's
2:14:31 basically like chat gbt you can install
2:14:33 on a local machine or on a virtual
2:14:36 machine in the cloud that's got enough
2:14:38 gpus to run it quickly and you can build
2:14:40 your own applications you can build your
2:14:42 own chat gbt you can train it on all
2:14:44 sorts of other [ __ ] you could fine tune
2:14:46 it on
2:14:47 50s sci-fi insect images and go make
2:14:52 your own 50s sci-fi image gener image
2:14:54 generator you can do whatever you want
2:14:56 with it
2:14:57 um
2:14:58 so yeah
2:15:02 so it's it's like all those other things
2:15:04 but it's out in the wild it doesn't have
2:15:06 any safety guard rails on it you can do
2:15:08 whatever you want and you might be
2:15:10 asking
2:15:11 couldn't that be dangerous
2:15:21 Elon Musk is at is coming for this group
2:15:23 as well
2:15:26 recommended books
2:15:27 um none right now unless I mean there's
2:15:29 there's probably recommended books on
2:15:32 everything sort of pre-chat GPT I don't
2:15:35 think there's anything I would recommend
2:15:36 post chat gbt because it's all too noon
2:15:39 it's changing too fast right but even if
2:15:41 you wrote a book with chat CPT by the
2:15:43 time you self-publish it
2:15:45 you write it on Friday self-publish it
2:15:48 on Saturday by next Wednesday it's going
2:15:49 to be out of date
2:15:51 so
2:15:54 data hoarders unite yes
2:15:57 same i max out everything and I I always
2:16:00 think that whatever the next biggest
2:16:01 hard drive is I'm like I'm gonna be good
2:16:03 here and I don't [ __ ] last six months
2:16:05 and I've maxed that out although the
2:16:08 cloud stuff has kind of Saved because
2:16:10 like iCloud just dumped [ __ ] off
2:16:13 it like just keeps uploading [ __ ] to the
2:16:15 cloud but the problem is now I've got so
2:16:18 much stuff in the cloud that like even
2:16:20 files I worked on like three days ago
2:16:22 are up in the cloud and some of them are
2:16:23 like 600 megabyte PowerPoint
2:16:26 presentations
2:16:29 um I paid for digital hoarder AI
2:16:31 Services I'm telling you that's a
2:16:32 business does anyone think that Claude
2:16:34 has a personality he's my top new
2:16:36 assistant
2:16:38 uh Pi has a lot of Personality Pi has
2:16:40 too much personality but if you want
2:16:42 like a virtual companion Pi is really
2:16:44 interesting
2:16:45 and then Claude yeah clouds Cloud's
2:16:49 pretty good
2:16:50 Bard is a little
2:16:53 got a touch of the Chisholm
2:16:57 foreign
2:17:00 I think I will continue to hoard yes
2:17:02 dude
2:17:04 you don't have the secret sauce what's
2:17:06 in your drink it's really just coffee
2:17:09 I'm just [ __ ] nuts
2:17:14 which which BSI do well with Avatar
2:17:17 which yeah what oh which AI does well
2:17:20 with Avatar creation
2:17:22 um
2:17:25 if you're just talking about static
2:17:26 avatars any of them I mean if you're if
2:17:29 like if you're just talking about
2:17:31 avatars to put on you know
2:17:33 your ex profile your Twitter profile or
2:17:36 whatever any of them even even Bing have
2:17:40 Bing make you an image
2:17:42 um if you want fancy [ __ ]
2:17:45 you know I mid Journey mid journey and
2:17:48 stable diffusion are the best and I
2:17:49 think I I keep tilting toward
2:17:51 mid-journey because it's just better
2:17:53 it's less of a pain in the ass to set up
2:17:55 and keep up with
2:17:59 um
2:18:02 great job you've been live for 90
2:18:04 minutes and I think that was like 30
2:18:06 minutes ago it's getting late what do I
2:18:08 have tomorrow
2:18:10 yeah let me check my calendar
2:18:13 I didn't even calendar I don't know I
2:18:15 got tomorrow
2:18:17 yeah we're gonna need you to go on ahead
2:18:19 and come in on Saturday yeah yeah oh one
2:18:23 more thing Peter we're gonna need you to
2:18:25 go on ahead and come on Sunday too
2:18:30 um
2:18:32 [Music]
2:18:35 oh I have a day tomorrow don't I
2:18:39 oh that's internal internal
2:18:44 what the [ __ ] is that
2:18:48 [Music]
2:18:51 all right
2:18:53 I
2:18:54 I guess I'm okay
2:18:58 I guess I'm all right if I'm a little
2:19:00 sleepy in the morning from hanging out
2:19:01 with y'all nerdy Wells
2:19:05 can I use AI for grant writing I think
2:19:08 AI is genius for grant writing yes you
2:19:12 can and what you should do I would say
2:19:15 go to claude.ai
2:19:18 simply because it's got this massive
2:19:21 context window which should mean nothing
2:19:23 to you if you haven't been playing with
2:19:24 this yet but it basically means so here
2:19:27 I'll go show you
2:19:29 in fact I wonder how you find a grant
2:19:32 what do they call a grant
2:19:34 at their Grant request what is it a
2:19:37 grant
2:19:39 let's see
2:19:41 this Claude no
2:19:43 that's Pi let's go to Claude
2:20:02 grants.gov
2:20:13 well I don't know how to do it all right
2:20:15 I'm too clueless about this world so you
2:20:18 go find a grant so when when you're in
2:20:19 when you're in Claude there's a little a
2:20:22 little paper clip here you can upload up
2:20:24 to five
2:20:25 files
2:20:26 so just turn your grants you know the
2:20:29 Grant application information
2:20:32 into PDFs if they're not already PDFs
2:20:34 and just upload them
2:20:36 into Claude and then they can they can
2:20:39 be relatively large right you can do
2:20:41 like 75 000 Word documents which is
2:20:45 which is pretty big
2:20:46 and and I I don't know if it's limited
2:20:50 if you upload if it's different if you
2:20:51 upload them versus just paste it in
2:20:53 there
2:20:54 um
2:20:55 but yeah just pop it in there and start
2:20:57 asking it questions about it and then
2:20:59 you know tell it to recommend
2:21:02 say you know basically say here's the
2:21:05 grant
2:21:06 here's who we are right so you're gonna
2:21:07 have the grant document up there you
2:21:09 could have information about your
2:21:10 organization up there you could even
2:21:12 have here's a grant uh we did for
2:21:17 someone else right here's an example of
2:21:19 what we did and then say okay you know
2:21:22 based on that Grant application this
2:21:25 company information and this previous
2:21:27 thing you know follow the tone and
2:21:30 structure of the previous thing
2:21:32 and you know write me a grant and yes it
2:21:35 will do it and it will do it well now is
2:21:38 it gonna be perfect new
2:21:41 is it going to make [ __ ] up where you're
2:21:42 like that just sounds weird yes but it
2:21:46 will get you
2:21:48 60 to 80 percent there
2:21:53 where
2:21:54 you know you'll be able to knock it out
2:21:56 in a in a day instead of two weeks
2:22:04 you have a few volunteers for moderator
2:22:06 I don't know how to do it either haven't
2:22:09 gone live yet
2:22:11 oh that's cool
2:22:13 nice job on the linked
2:22:15 LinkedIn AI laptop
2:22:20 what was the AI lat LinkedIn AI laptop
2:22:24 was that my like my Friday night my
2:22:27 Friday afternoon I've been doing Friday
2:22:28 afternoon what do I call them office
2:22:31 hours with Kyle
2:22:34 because I'm not online enough
2:22:38 sign me up to mod Kyle then I'll cut
2:22:40 class for sure
2:22:43 I could mod too but I'm not here enough
2:22:45 I don't I honestly don't know what how a
2:22:49 mod thing would work like is it a I
2:22:51 don't I don't know anything about it
2:22:54 you know this is Elder talk remember
2:22:57 it's
2:22:59 just like bitter Gen X like
2:23:03 hanging out and scrolling
2:23:08 all right
2:23:09 let me get a few more questions in
2:23:13 Microsoft added Ai and chat to power
2:23:16 automate their version of zapier it
2:23:18 works okay yeah it's a little crappy the
2:23:20 the oh their version of zapier
2:23:23 even zapier's AI thing's a little crappy
2:23:25 everything's a little a little just
2:23:27 janky right now grace and mercy will be
2:23:30 required for AI I agree I think I think
2:23:33 these tools are going to help us be more
2:23:34 empathetic I think they're going to
2:23:35 train us to be more empathetic
2:23:38 um
2:23:39 but yeah I think I I a lot of a lot of
2:23:42 patience and empathy is
2:23:45 is going to be required for the next two
2:23:47 or three years
2:23:50 I like turtles
2:23:54 ah Bob it's a classic not everyone knows
2:23:58 how funny that really is have you all
2:24:00 seen the I like turtles video let's go
2:24:03 do that we might as well entertain
2:24:05 ourselves
2:24:07 oh my God this is funny
2:24:09 um
2:24:11 is back when everything was on YouTube
2:24:19 uh where is it all right
2:24:25 I like turtles
2:24:29 that's your life at the water paint job
2:24:31 what do you think I think
2:24:34 all right
2:24:37 the Waterfront Village open for the next
2:24:40 11 days
2:24:43 oh my God it's funnier than I remember
2:24:57 all right you're great zombie and good
2:25:00 times here at the Waterford is open for
2:25:03 the next 11 days
2:25:05 oh my gosh
2:25:09 oh I can't I ache
2:25:14 [Laughter]
2:25:24 oh my God
2:25:26 the funniest thing is
2:25:29 is her Trend to keep it together
2:25:34 I'm a professional oh my god
2:25:38 oh Humanity
2:25:46 all the humanity
2:25:50 there's the Hindenburg and then there's
2:25:52 the other like turtles kid oh my God
2:25:56 no
2:25:58 oh
2:25:59 God have you seen the machine oh I
2:26:02 haven't seen the movie yet how is it
2:26:05 um the story is great but I haven't seen
2:26:08 the movie
2:26:10 that's so much sugar exactly that's the
2:26:14 Kool-Aid thing with Bert crash oh my God
2:26:16 that's funny
2:26:18 um finally have gpt4 sub and mid-journey
2:26:20 sub yes
2:26:22 of all I've been playing with it a lot
2:26:24 can I for example write a book with gbt
2:26:27 you can go check out well I mean so
2:26:32 so yes you can write a book with it
2:26:34 but so so there's two things there's a
2:26:37 tool that you can use that just writes a
2:26:39 book but if you're using chat CPT if you
2:26:41 just say write me a book it's going to
2:26:42 write you a shitty book it'll write you
2:26:43 a book but it'll just be crap
2:26:46 but what what you can do is do layered
2:26:49 prompting you know or they call it
2:26:51 prompt stacking now there's all sorts of
2:26:52 terms for this [ __ ] right but it
2:26:54 basically just means have a conversation
2:26:55 so when you're writing a book start with
2:26:58 you know
2:27:00 um you know you're you're a you're a
2:27:03 master writing coach or whatever it is
2:27:05 and you know I'm an author and this is
2:27:08 the genre of book that I'm writing and
2:27:11 here's some thoughts about it and if you
2:27:13 have an outline great if you don't
2:27:14 that's fine too
2:27:17 um just tell it as much as you can about
2:27:18 your book
2:27:20 and then
2:27:22 maybe have it suggest different story
2:27:25 Frameworks like if you don't have an
2:27:26 outline say what uh what give me five
2:27:29 different types of story Frameworks that
2:27:32 you know might help us in writing this
2:27:34 book and then it'll suggest five
2:27:35 different writing Frameworks here's save
2:27:37 the cat and here's
2:27:39 um you know Joseph Campbell's hero's
2:27:41 journey and here's this and here's that
2:27:43 and right
2:27:44 and then pick one and then say okay
2:27:46 that's great and and then
2:27:49 um you know can you take these basic
2:27:51 plot elements and flesh out the story
2:27:54 you know in that framework that we've
2:27:57 selected sure here it is
2:27:58 and then you could say okay you know you
2:28:01 know it seems like there's you've only
2:28:03 represented two or three characters in
2:28:04 here but I've got ideas for 10 more
2:28:07 characters so give me some character
2:28:09 archetypes that might be good
2:28:11 compliments to this story framework and
2:28:13 it'll give you story you know character
2:28:15 types and then you say oh that's great I
2:28:17 really like the the the Dizzy school
2:28:20 teacher for comic relief that's perfect
2:28:23 um you know give me 10 names for her
2:28:25 right right and just and just keep
2:28:28 adding to it and just keep boom boom
2:28:30 boom boom
2:28:31 and then at some point you'll be ready
2:28:33 you can say that's terrific now you know
2:28:36 write me up a synopsis write me up a
2:28:39 concise outline write me up a character
2:28:41 sheet pull all those things out so now
2:28:43 you have those as starting points that
2:28:46 you could start a new chat with and say
2:28:48 okay you're a writing coach
2:28:50 previously we've worked on this basic
2:28:52 story with the you know this outline
2:28:54 with these characters you know
2:28:58 give me 10 different opening paragraphs
2:29:03 or or give me a detailed outline for
2:29:07 chapter one and then tweak that outline
2:29:10 and then say now give me
2:29:12 you know 20
2:29:14 opening line you know as as we both know
2:29:17 the opening sentence of another of a
2:29:19 novel is the most important
2:29:22 you know novel It's gotta you we gotta
2:29:25 we gotta hook him like this right
2:29:27 so maybe give me give me the 20 most
2:29:30 famous opening lines for novels have it
2:29:33 give you that first
2:29:36 right
2:29:38 I am Ishmael
2:29:43 and then have it give you 20 opening
2:29:45 lines for your first paragraph
2:29:47 and then if they all suck tell it it
2:29:48 sucks give me 20 more
2:29:51 these are all cliche make them not
2:29:53 cliche make them unexpected
2:29:56 use play on words you know
2:29:59 give me some literary mechanisms
2:30:03 for how we might do opening lines and
2:30:06 let it give you some literary theory on
2:30:08 opening line writing
2:30:11 and then when you have that then have it
2:30:12 write the first paragraph then you have
2:30:14 that have it write the whole first
2:30:15 chapter
2:30:17 and then if you get that in the
2:30:20 neighborhood
2:30:22 then you can go off to word and tweak it
2:30:25 right clean up all the it'll get to 80
2:30:28 percent
2:30:29 and then go clean that [ __ ] up
2:30:32 and now you've got a description an
2:30:34 outline character sheets in chapter one
2:30:36 pop that in there and say okay let's
2:30:38 work on chapter two and then you can
2:30:40 just work your way through it so that's
2:30:41 one option
2:30:43 and that's that's how I would do it
2:30:45 that's how I've been doing it
2:30:47 um
2:30:48 Claude is really amazing because if
2:30:50 you've got an existing novel or an
2:30:52 existing screenplay like I have a
2:30:54 one-man show I want to turn into a book
2:30:56 and I'm probably just going to pop it in
2:30:58 Claude and say turn this into a book
2:31:03 um there's a project out there called
2:31:06 GPT author
2:31:08 now downside you have to know how to run
2:31:13 Google collab and run a notebook in
2:31:15 Google collab which is not ridiculously
2:31:18 hard but it's just when the Google
2:31:20 collab things break if you don't know
2:31:24 coding you're just kind of like it's
2:31:27 it's just this massively frustrating
2:31:29 like like this is why I failed nine
2:31:31 times to get stable diffusion up and
2:31:33 running because I kept pushing the play
2:31:35 button in these Google collab things
2:31:37 that it wouldn't load a library and I
2:31:38 didn't know what the [ __ ] to do about it
2:31:40 so but anyway look up a project called
2:31:43 GPT author
2:31:45 and it's [ __ ] insane
2:31:47 you give it
2:31:49 um
2:31:50 what you want the book to be about what
2:31:53 style you want it written in and how
2:31:55 many chapters you want and then you give
2:31:56 it your open AI key and and your stable
2:31:59 diffusion key
2:32:01 and it writes an outline
2:32:04 it then writes each of the chapters for
2:32:07 however many chapters you want and then
2:32:09 once it's written it it creates cover
2:32:11 art for the for the book and then it
2:32:14 publishes it as an ebook that you can
2:32:17 open in like iBooks on the Mac
2:32:22 and what's cool about it is because it's
2:32:24 a Google collab notebook you can see all
2:32:27 the code right all the code all the
2:32:29 prompts that they put in there to write
2:32:30 this book are in there so I think what
2:32:33 if what it's what it's designed to to
2:32:36 publish is a fantasy novel
2:32:39 but you could go in and you could
2:32:40 rewrite all those prompts that have it
2:32:42 right a young teen novel or a romance
2:32:44 novel or a how-to like a non-fiction
2:32:47 knob or non-fiction book
2:32:50 um so but yes you can so keep digging
2:32:54 parcival there's you're in the right
2:32:56 place
2:32:57 I'll smoke what he's smoking please
2:33:01 this is just this is just my natural
2:33:03 high I'm just naturally high in life if
2:33:06 you know what I mean oh my it's a little
2:33:07 warm in the ends of me I'd rather do the
2:33:10 cold out
2:33:13 all right
2:33:16 all right what time is it let's do 10
2:33:18 more minutes I'm really bad at this time
2:33:20 thing
2:33:25 [Music]
2:33:28 all right I'm the poster child for ADHD
2:33:32 use books on Audible and put them on
2:33:34 four or five speed yeah exactly I've
2:33:37 started um I've had a partner oh how we
2:33:40 doing today I gotta I I've told myself
2:33:43 this before but my adds prevented me
2:33:45 from following up on it but uh next time
2:33:48 I get a cowboy hat I gotta get the some
2:33:50 lines from Yellowstone right
2:33:53 because howdy partner is just what are
2:33:55 we in the 50s
2:33:58 um
2:34:00 did you see prompt generated parody of
2:34:03 South Park written animated direct voice
2:34:05 with AI yeah it's not
2:34:07 it's prompt generated parody well it's
2:34:10 not really a parody of South Park so
2:34:12 it's called showrunner AI so if you
2:34:15 haven't seen showrunner ai go look at
2:34:16 showrunner AI you write a prompt for a
2:34:19 South Park episode and it generates a
2:34:21 South Park episode complete with voices
2:34:24 music animation everything now it ain't
2:34:27 great
2:34:28 but that's the sign of things to come
2:34:30 that's we're not that far away from
2:34:33 being able to ask for your personal
2:34:36 version of whatever show you want to
2:34:38 watch and it will just make it
2:34:40 easily within five years probably within
2:34:43 three
2:34:47 [Music]
2:34:49 what's the name again please I don't
2:34:51 know
2:34:53 what you mean Elon Musk is here I don't
2:34:56 think so
2:34:57 if elon's here hey Elon what's happening
2:35:01 what's happening listen I don't know if
2:35:04 you thought through this x thing too
2:35:05 much because uh you know that Mark
2:35:07 Zuckerberg he's got some X thing called
2:35:09 Xbox and he's got a lot of trademarks on
2:35:12 the X the X things
2:35:15 so you might have some issues
2:35:18 maybe when you do your cage match with
2:35:20 Mark you two could fight it out over
2:35:23 trademark for that X thing
2:35:28 uh
2:35:30 uh he doesn't really give two shits like
2:35:32 listen I think what he wants to do with
2:35:33 X is really interesting the idea of the
2:35:36 everything app I think is fascinating
2:35:39 um is he doing it in a way that's like
2:35:42 bull in a china shop sure but what the
2:35:45 [ __ ] does he care he's [ __ ] flying
2:35:47 Rockets to Mars successfully
2:35:54 all right
2:35:55 your own books
2:35:58 pie was prancing me last week
2:36:01 it's coffee keep doing what you're doing
2:36:04 thanks again thank you thank you thank
2:36:06 you I appreciate it I got to work this
2:36:07 weekend auditing the medication room
2:36:09 thanks for reminding me
2:36:13 um I gave up on pie that day prancing
2:36:15 would have been a lot better
2:36:18 um
2:36:19 five minutes two minutes gotta go
2:36:23 Gotta Go Gotta Go Gotta Be There
2:36:28 welcome to Elder talk with Kyle Shannon
2:36:32 many people lose jobs yes many people
2:36:35 are going to lose jobs because of this
2:36:36 stuff
2:36:37 um
2:36:38 which is going to suck
2:36:41 many people that upskill are gonna be
2:36:44 okay are gonna are gonna weather the
2:36:47 storm because there's going to be a lot
2:36:48 of companies
2:36:51 this stuff is so messy right now and so
2:36:53 spit and duct tape that it's going to
2:36:56 take companies a long time to sort
2:36:57 through how to do it so anyone that's
2:36:59 even got half a clue is is is going to
2:37:02 be in much better position but there's
2:37:04 going to be there's going to be a lot of
2:37:05 job disruption and then I think again
2:37:07 long term I think we're going to be okay
2:37:09 by the way
2:37:13 yesterday yesterday I did a video on
2:37:16 this my latest videos on this Sam
2:37:18 Altman's World coin came out so if you
2:37:20 don't know World coin
2:37:22 it's a new cryptocurrency by the CEO of
2:37:26 chat GPT of openai called World coin
2:37:29 there's three components to it there's
2:37:31 there's an app that you can do
2:37:32 transactions in there's the
2:37:34 cryptocurrency itself and then there's
2:37:36 the world ID which is human verification
2:37:40 how are they verifying that you're human
2:37:46 I'm gonna do this in two weeks by the
2:37:49 way I'm going to New York you can only
2:37:50 do it in in the states in New York Miami
2:37:53 and San Francisco
2:37:54 they have this orb they have this
2:37:57 [ __ ] orb and you walk up to it and
2:37:59 you look into it and it scans your iris
2:38:02 they verify that you're human and then
2:38:04 you get a world ID so you get a
2:38:06 blockchain verified I am human
2:38:09 identification and they give you
2:38:11 cryptocurrency and then they've got this
2:38:14 place that you're going to be able to
2:38:15 spend it
2:38:17 it's all geared to developers right now
2:38:19 so they're looking to build an
2:38:20 infrastructure of verified humans and an
2:38:25 economy
2:38:27 to deal with those verified humans the
2:38:29 long-term goal is to generate
2:38:33 ai-funded Universal basic income
2:38:37 so I actually think as creepy as [ __ ] as
2:38:40 it sounds there's a couple of things
2:38:41 going on with it that I think are really
2:38:43 important I think the ability to verify
2:38:46 if someone's a real person is a big deal
2:38:48 I think we're gonna need some sort of
2:38:50 universal basic income
2:38:52 and
2:38:53 understanding that
2:38:55 people are verified as the foundation
2:38:58 for being able to verify that not only
2:39:01 is this person a person but this video
2:39:03 that they created is actually them and
2:39:05 it hasn't been modified
2:39:07 and this article that they wrote was
2:39:09 written by them and this other verified
2:39:11 person and it hasn't been modified I
2:39:13 think that's going to be critical the
2:39:16 2024 election is going to set the
2:39:18 foundation for how [ __ ] screwed we
2:39:22 are
2:39:23 with with misinformation and not being
2:39:25 able to trust anything
2:39:29 um so as creepy as it is I'm going to
2:39:31 New York in two weeks I think the week
2:39:33 of the 14th and
2:39:36 um I'm going to one of the orbs and I'm
2:39:38 going to scan my Iris and I'll tell you
2:39:39 if my eye starts bleeding or if they
2:39:41 steal my bank account
2:39:44 I don't think they're gonna but um as
2:39:46 creepy as it is I actually think it's
2:39:47 important so I'm going to experiment
2:39:49 with it I'll experiment with it so you
2:39:51 don't have to
2:39:54 all right on that creepy 1984 type
2:39:57 reference I'm gonna get out of here
2:39:59 peace out everybody I really appreciate
2:40:02 all your time and you hanging out with
2:40:03 me I hope this was useful
2:40:06 um I'll right before I leave I'm gonna
2:40:08 just flip the screen so you can
2:40:10 screenshot these are the recommended
2:40:12 websites
2:40:13 like the chat GPT sites and then uh the
2:40:17 the other two screens that have got info
2:40:20 on them that might be worth
2:40:21 [Music]
2:40:23 all right screenshot that bad boy
2:40:29 those are all the GPT players
2:40:34 these are
2:40:37 some resources up top some other cool
2:40:41 [ __ ] to check out below
2:40:43 and then the AI Salon is a community
2:40:46 that I started back in December of AI
2:40:48 adventurers storyvines by company Kyle
2:40:50 Shannon is my Twitter handle
2:40:53 so with that peace out everybody I
2:40:56 really appreciate all your time have a
2:40:58 good night and I will see you tomorrow