aiLL July 8th, 2023 - Exploring Generative AI Trends, Ethics, and Possibilities

Description

Chapters

Transcript

0:22 welcome welcome welcome
0:25 to the AI learning lab
0:28 happy Saturday night
0:31 I was just watching a little Tom Segura
0:33 his new special
0:36 what's it called I forget
0:39 he's a funny funny mofo I like Tom
0:42 Segura
0:43 uh but I'm Pom-Pom thanks for your time
0:45 on these lives or for these lives you're
0:48 very very welcome I appreciate the kind
0:50 words
0:52 good day good day good day
0:55 welcome everybody we'll wait for some
0:58 other folks to come on in feel free to
1:00 share the live
1:02 greeting Mr Vasquez
1:06 greetings greetings greetings
1:12 yeah I think hitting the hitting the
1:14 heart button lets people know we're live
1:16 too so
1:18 if you haven't talked about it recently
1:20 would you mind talking about API keys
1:23 and their use I can talk about that sure
1:27 I assume you mean do you mean just
1:30 generally or or using open ai's API keys
1:34 let me know let me know because it's a
1:37 it's a broad topic and I don't go very
1:39 deep in it but I can give you some I can
1:42 give you some of the basics
1:43 will they ever bring back Bing on chat
1:45 GPT I don't know apparently when a
1:48 lawyer gets out of their ass
1:51 oh man there's a lot of people here
1:52 welcome everybody my name is Kyle
1:54 Shannon this is the AI learning lab a
1:56 little bit about me I am an entrepreneur
1:59 I've started I don't know 12 or 15
2:01 companies in my day I should probably
2:03 get that number that's something I
2:04 should probably know uh my current gig
2:07 is I am the co-founder and CEO of a
2:09 company called storyvine which is an
2:11 automated video storytelling platform
2:13 and this channel is all about exploring
2:16 generative Ai and dealing with uh what
2:19 it is and what it's going to do and and
2:21 sort of talking about you know now that
2:23 we've hit this inflection point where
2:25 chat GPT has happened hey Emilio's wife
2:28 class is in session yeah class with the
2:32 disturbing Professor
2:36 um uh you know now now that now that you
2:40 know the the cat is out of the
2:41 proverbial bag with with all these large
2:43 language models and chat gbt
2:46 um what the hell do we do with this and
2:47 how do we make sense of it how do we
2:49 learn about it and what do we do and
2:51 is it going to take your job and is it
2:53 gonna you know are the robots gonna kill
2:54 us all that sort of stuff
2:56 um I tend to lean on the more optimistic
2:59 side of the the conversation I tend to
3:01 tend on the less technical side of the
3:04 conversation I'm more about what do we
3:06 do with it now that it's here rather
3:07 than you know how do we build the models
3:10 and do the engineering so that's why I
3:13 have a few questions about AI pop them
3:15 in the comments below I'll do my best to
3:16 answer them and uh you know other things
3:20 like I I swear sometimes so if you don't
3:23 like that I would encourage you to find
3:25 another AI learning lab
3:28 that has better values and then uh what
3:32 else uh I'm kind of a smart ass if
3:35 you're gonna troll be creative about it
3:36 don't just be a jerk and don't be a jerk
3:38 to other people in here please and don't
3:40 bring in politics and let's not have
3:42 Division and yeah yeah [ __ ] like that
3:44 all right let's go answer some questions
3:46 he's legit adorable no he's so cute what
3:52 are we gonna do oh do I learn about it
3:55 yeah
3:57 all right I don't know why I did that
4:00 because I'm an idiot
4:01 [Laughter]
4:04 all right let's see
4:08 old as in 1994 well I don't know what
4:11 that's in reference GTS wait GT's G
4:16 as old as in 1994.
4:18 I assume you mean you've been here
4:20 before and you've heard me talk about uh
4:23 starting uh a company in 1994 which was
4:26 one of the early web companies which is
4:28 one of the reasons I think this this uh
4:32 what's it called generative AI stuff is
4:34 such a big deal as I was there for the
4:36 early days of the web and started one of
4:37 the early digital agencies and this is
4:39 very very similar to that so anyway if
4:42 you haven't played uh is that top one
4:45 there
4:46 chat.openai.com that is
4:49 um that's the official chat EPT website
4:51 so if you haven't gone if you haven't
4:53 played with it just go there right now
4:54 and play with it and if you don't know
4:56 what to do go to that document
4:58 prompts.chat and that'll teach you what
5:00 prompting is and how to talk to these
5:02 things and what they really are and
5:03 it'll give you a bunch of ideas about
5:05 sort of personas that you can give chat
5:07 GPT to act like like act like an
5:10 accountant act like a
5:12 strategist act like a marketer act like
5:14 a screenwriter whatever it is and and
5:17 it's it's a really good document and
5:19 then Bing and Poe are also chat GPT
5:22 variants that you can play with being in
5:24 particular is gpt4 connected to the
5:28 internet and you can make images there
5:30 for free you just type in make me an
5:31 image of a I don't know 1969 GTO Judge
5:36 in a field and it will it won't look
5:38 like a GTO judge but it'll be close
5:40 it'll look like a 60s 70s muscle car
5:42 and then pro.com gives you access to a
5:45 bunch of different models and then
5:46 future pedia the one down below is uh
5:49 just a big directory of AI sites and
5:51 services and [ __ ] like that most of them
5:54 are crap but there's some good stuff
5:55 there all right
5:58 let's get into the questions
6:00 okay
6:06 one to add to this oh want to add this
6:09 competency on my resume what's a good
6:11 way to learn what I need to claim
6:14 confidently
6:15 um can you spell AI you got to get those
6:19 letters right like if you're dyslexic
6:20 it's a [ __ ] nightmare
6:22 so I'm an expert at IA
6:29 um let's see well I I you know I I would
6:32 I would say you know the reason I have
6:34 these URLs on screen for
6:39 these lives is that's where I think you
6:42 should start
6:43 um
6:45 just start to understand what you can do
6:47 with chat GPT
6:49 start solving problems with it
6:52 um I don't know quite what you are if
6:54 you're a project manager or if you're an
6:56 account person or if you're a
6:58 salesperson
6:59 start to understand what you can do with
7:02 these tools for your job just start to
7:04 get AI literate about what's possible
7:06 here and then what you can do for what
7:08 you already do there
7:10 and that's a place to start and then the
7:13 the thing that I would add to it is
7:15 start talking about it like from day one
7:17 like tonight
7:18 go there if you haven't
7:21 and just start [ __ ] around have it
7:24 write a song have it write a poem have
7:25 it write a project plan have it write a
7:28 recipe have it write a meal plan like
7:30 doesn't [ __ ] matter just go there and
7:31 start playing
7:32 and then share that share it on Facebook
7:35 share it in your community group share
7:37 it with your family even if they think
7:39 you're [ __ ] weird and I I heard ai's
7:41 gonna kill us all why are you doing that
7:43 that's stupid you're dumb dumb Jenna you
7:46 know Jenna I don't know why you do all
7:48 this computer stuff it just takes you
7:50 away from your family
7:52 just ignore all that [ __ ]
7:54 I assume your family doesn't talk like
7:56 that but if they do I'm psychic uh
8:02 um and start telling people about it
8:04 because
8:07 um that may lead to a job for one thing
8:09 or a different job or a different
8:10 opportunity in your current job but
8:12 what's gonna start to happen and it's
8:13 already starting to happen is that
8:16 um people are increasingly going to want
8:18 to figure this out and they're going to
8:19 start asking who do I know that knows
8:21 something about this
8:22 and initially you might be crazy Jenna
8:25 who just talks about AI
8:27 um but at some point they're gonna go oh
8:29 I think I gotta deal with this I just
8:31 got laid off because of this stuff what
8:33 is this stuff can you help me
8:36 um be that person be that educated
8:38 person
8:39 um
8:40 my hair looks good so do you thank you I
8:43 don't know that hair looks good that
8:46 hair looks like it needs an intervention
8:50 ah
8:55 all right how do you get rid of business
9:00 when it's Tanked
9:03 how do you get rid of business when it's
9:04 tanked like if you started a business
9:07 I don't know you gracefully shut it down
9:09 you thank people for their hard work you
9:11 thank your customers and you do what you
9:13 can to get out of contracts I don't
9:14 quite know what you mean there
9:16 um if you need like legal advice I'd get
9:19 a lawyer
9:24 but I don't you can clarify that
9:27 can AI give lawyer advice it can but you
9:31 need to check it and double check it if
9:33 you're talking lost stuff but you can
9:35 have it like what it can do really well
9:36 is it can read
9:38 um it can read documents and and there's
9:41 this new thing called code interpreter
9:43 which now is available to all chat GPT
9:45 Plus members
9:47 where you can upload like it seems like
9:50 any kind of file and it it sort of
9:52 figures [ __ ] out about them
9:54 um but you could upload a big lawyerly
9:57 doc document
9:59 and have it analyze it and help you
10:02 understand it and potentially even come
10:05 up with strategies for how to respond to
10:07 it
10:08 um
10:09 and then you know unless it's kind of a
10:12 low stakes thing like like
10:14 um
10:15 a few months back we started we started
10:18 this newsletter called everyday Ai and
10:20 then some
10:21 partner from Andresen Horowitz also
10:23 started a thing called everyday AI
10:26 um and that was a case where like we
10:28 didn't want them to do it but it didn't
10:29 really matter so we're actually thinking
10:31 about you know
10:32 having an AI lawyer maybe we could get a
10:35 like a a reporter to follow the case
10:38 where we were just going to use a
10:40 full-on AI lawyer it ended up not going
10:43 anywhere we ended up dealing with them
10:44 you know without needing to do law stuff
10:46 but if it's something that is actually
10:49 really important then you should get a
10:51 lawyer but you can absolutely use chat
10:53 CPT to understand documents to come up
10:56 with ideas for strategies things like
10:58 that and then again
11:00 um go to prompts.chat and you know look
11:03 at the the structure of those prompts
11:05 where it tells the the
11:08 um chat GPT to act like a certain
11:09 Persona and have yours act like a lawyer
11:16 bow all right
11:19 let's see okay this is this one's
11:22 actually important
11:23 best singer Mariah Celine or Whitney
11:29 Celine bugs me
11:32 I think you gotta go with me I think you
11:34 gotta go Whitney I mean Mariah's
11:37 remarkable celine's remarkable
11:40 but there's just something about her
11:42 that I don't know
11:44 and and you know Whitney's just you know
11:47 she's just on a different level it's
11:49 Whitney it's it's Whitney uh there's a
11:52 fun there's a fun Christmas we had one
11:55 year where my wife's family she's she's
11:58 from the Bronx and
12:01 they they came up we had a country place
12:03 in Pennsylvania and I heard all this
12:06 screaming like after dinner like we
12:08 finished dinner and I went up to watch
12:09 TV with some of the guys and
12:11 and there's just you know loud loud
12:13 chatter and it's going on for like an
12:15 hour and it's and it's getting escalated
12:18 and it's turning into a fight
12:20 and so I went downstairs and I said what
12:23 the [ __ ] is going on like what what do
12:26 you and they were they had it was like
12:28 an a 90-minute full-blown argument about
12:31 who was more famous Barbara Streisand or
12:33 Michael Jackson
12:36 those are the best arguments
12:40 what operating system are you I I'm an
12:44 apple Fanboy so I am OS X through and
12:47 through
12:48 don't give me that Microsoft [ __ ]
12:50 except in the last six months I've
12:52 become a Microsoft fan
12:55 non-swearing classes down the hall
12:57 exactly teach me something okay I'm six
13:00 two
13:02 um
13:03 taller people sometimes have to duck
13:05 their heads through doorways
13:09 hope that's new information for you
13:12 um are you a robot no but I am chat GPT
13:14 5.
13:17 don't I look realistic
13:23 yellow little smart ass eat tonight I
13:27 like it I like it I haven't I have a
13:29 sense there's some there's some wine
13:31 drinking going on there might be a
13:33 whiskey or two
13:35 you know it might be might be some uh
13:39 some people playing the ganja what
13:42 company is doing the most with AI well
13:44 Microsoft is [ __ ] with the most
13:46 competitors
13:47 and and doing it joyfully open AI is the
13:53 thorn in the side of the most companies
13:55 and those two are like you know they're
13:56 little pals right now they're little
13:58 corporate Buddies
13:59 [ __ ] with the world
14:01 um
14:02 don't discount Google they invented the
14:05 Transformer which is the T and GPT
14:07 that's Transformer that Google invented
14:10 now their Bard model sucks but it'll get
14:13 better
14:14 so they're doing a lot in AI
14:18 um but they're definitely playing
14:19 catch-up
14:21 um meta is trying desperately to
14:25 detangle itself from its name
14:28 but it's not a meta's not really about
14:31 the metaverse per se I mean you know
14:34 meta has an A in it and then AI like if
14:38 we just added a an i at the med AI we
14:41 could we're an AI company right so but
14:44 they're actually doing some interesting
14:45 [ __ ] where they're they're releasing
14:48 um some of their technology into the
14:50 into the wild as open source which which
14:53 is interesting
14:55 um so they're doing interesting stuff
14:56 mid-journey
14:58 um is for the image generation tools in
15:02 my opinion by far the the uh the best of
15:05 the image generation tools but uh
15:08 stability AI that does stable diffusion
15:11 they keep
15:12 um releasing their um all of their
15:15 technology as open source which I think
15:17 is another interesting move
15:21 um then there's anthropic and then
15:24 there's
15:25 um inflection which just got funded for
15:28 1.3
15:30 billion dollars
15:33 um
15:34 but that's the dude that was the founder
15:36 of um Deep Mind so
15:39 he's he's a literal [ __ ] Brainiac
15:43 um so there's some some good pedigree
15:44 there
15:45 um those are the biggies Nvidia seems to
15:48 have a a relative Monopoly on gpus for
15:52 for AI processing which I find baffling
15:56 but it seems to be the case AMD is
15:58 trying to come up and do something with
16:00 them but like the the like all the
16:03 software stuff is so entrenched with
16:04 Nvidia
16:06 that I don't know it's that that's going
16:08 to be a tough one to topple
16:10 um
16:11 apple is
16:14 Apple to me right now feels like a shark
16:17 circling the boat
16:19 [Laughter]
16:21 because they've got their own silicon
16:24 that processes
16:26 um machine learning locally and they've
16:28 got a real good reputation for privacy
16:30 and they're [ __ ] Apple when they when
16:32 they get into this stuff it's gonna be
16:33 really good
16:34 and they're gonna like they're gonna
16:35 kind of reinvent it like right now ai is
16:38 really geeky and you gotta go to chat
16:40 GPT and you gotta learn about prompt
16:42 engineering and now there's this thing
16:44 called code interpreter and you got to
16:45 figure out how to what it is and how to
16:47 upload a file and how to query a file
16:49 and then you watch it do its thing and
16:50 it doesn't do it and you kind of it's
16:52 like everything right now is just this
16:54 janky spit and duct tape don't get me
16:57 wrong it's remarkable profound gonna
16:59 change the world forever technology but
17:02 right now it's janky and duct tapey
17:06 um those are the biggies for me those
17:07 are the ones I'm paying attention to
17:10 I'm looking for a good free AI to help
17:13 with coding well go
17:15 there
17:18 go there
17:21 go there
17:23 go there
17:25 that's why I have it here and when
17:27 there's a better one I'll put a better
17:28 one here so so if I I don't know if
17:33 um if co-pilot on GitHub is free but
17:37 co-pilot on GitHub is a a an a GPT
17:43 coding assistant so you do pair
17:45 programming where it does kind of code
17:47 completion for you but gpt4 does code
17:50 really well and they just launched code
17:52 interpreter which I'll show you here in
17:54 a second it's mind-blowing it
17:57 you upload a file and then you start
17:59 asking about it and it writes its own
18:01 python code and executes it what it
18:04 essentially does I'm starting to
18:06 understand it now
18:07 when you when you launch
18:09 an instance of code interpreter like a
18:13 code interpreter chat what it's
18:15 essentially doing is spinning up like a
18:17 virtual computer
18:19 and then you ask it to do something and
18:22 it's like sure and then it writes python
18:25 code it installs libraries if it needs
18:28 to it runs the code if stuff times out
18:32 it's like oh well let me try another
18:33 thing it it basically keeps running code
18:37 until it succeeds and then it gives you
18:39 what you want and it's doing this all in
18:41 this little kind of virtual computer
18:43 it's like you have this little this
18:45 little coding assistant this little
18:47 technology data analyst assistant just
18:50 sitting there waiting
18:52 um so if you haven't played with code
18:54 interpreter you should it's not free but
18:56 it's 20 bucks a month and for what it
18:59 does it's [ __ ] remarkable but if you
19:01 just want it to do code like GPT 3.5
19:04 does code really well
19:06 um bing.com is gpt4 so I don't know how
19:10 well that version of gpt4 is for code
19:13 but it's probably pretty good because
19:15 they're the ones that put together
19:17 um copilot for GitHub
19:19 and then there's a bunch of if you go to
19:21 the Bots at poe.com they're all free and
19:23 there's a bunch of coding Bots if you go
19:26 into the technology into the explore
19:28 button at poe.com here I'll I'll show
19:30 you a couple of those
19:32 [Laughter]
19:33 um
19:39 po
19:41 so when you go to po.com it's just a
19:43 normal chat chat bot kind of Chat gbt
19:46 thing there's all these different models
19:48 you can play with but this explore
19:49 button here you can go explore Bots that
19:53 other people have made
19:55 and then click on the programming one
19:58 and these are all programming bots so
20:01 python machine run python code HTML
20:04 machine JS machine
20:07 Pie golf Geo Json it
20:11 was just a ton of them and you know you
20:16 put in some stuff and it'll write code
20:18 for you so lots and lots of things to
20:20 play with but but honestly the just chat
20:22 GPT is is the deal so so here's check
20:26 this thing out so what you see on screen
20:28 right there this is code interpreter
20:31 this is just [ __ ] insanity
20:33 so what you're looking at is you're
20:35 looking at a pie chart of different
20:37 colors I know that's obvious it's like
20:39 duh Kyle we can see that uh but let me
20:42 let me read some of the color names
20:44 midnight mystery Chestnut chestnut horse
20:47 muted mustard Rusty Nail creamy Pearl
20:50 Golden Dawn Autumn Blaze cocoa Darkness
20:53 Lagoon Ripple these are all really good
20:56 color names and then there's these
20:57 different colored slices
20:59 so what this is is this is code
21:03 interpreter so what I did was I uploaded
21:11 this image which was an image from the
21:15 AI Salon this is one of the women in the
21:18 AI Salon made this which is super cool
21:20 so I uploaded that image and
21:24 and then I said to chat GPT or I said
21:26 Dakota interpreter whatever the [ __ ] it
21:28 is chat GPT I said find the top 15
21:30 average colors in this image plot their
21:33 relative percentages on a pie chart make
21:35 each segment the color represented and
21:38 label each color with a creative Crown
21:40 color that is memorable and brings that
21:43 color to life
21:44 and and so then what it does is it came
21:47 up with its own instructions number one
21:49 load the image and flatten it down to a
21:51 list of RGB colors apply clustering
21:54 using the k-means algorithm for instance
21:57 to find the 15 most common colors
21:59 calculate the relative percentages
22:01 generate creative and memorable names
22:03 for each color plot a pie chart so it
22:05 came up with its own instructions I mean
22:07 I I did it in English but it sort of
22:09 wrote them so it could understand it
22:11 and then it said Okay first uh loading
22:14 the image and applying k-means clusters
22:16 blah blah blah blah and and that one
22:18 failed and then it says I'm sorry it
22:21 took too long because that image was too
22:22 big but what I can do is I can down
22:24 sample that image and I'll down sample
22:26 it now what's Behind These Little Boxes
22:28 if you open these up
22:30 it's python code it's loading libraries
22:34 it's writing code it's running code and
22:37 then if it times out it gives itself new
22:41 instructions and it does the next thing
22:43 down so this one down sampled the image
22:45 there it found the color clusters
22:49 um now it says now let's generate
22:50 creative and memorable names
22:53 and so it came up with the names and the
22:55 relative percentages and then I guess on
22:57 the back end it's got the the actual RGB
23:00 values or the hex values
23:03 then I asked it to create the chart and
23:05 it screwed up once and I asked it to
23:06 create it again and it screwed up again
23:08 and I asked it to create a third time
23:10 and it did it
23:11 and so
23:14 so that's it with an image you can also
23:17 just upload a zip file full of csvs full
23:20 of data and
23:22 it just knows what they are
23:24 um and and can analyze it and write
23:27 charts about it and so anyway that's
23:29 that's code interpreter
23:32 um
23:32 it's just it's just mad nuts it is
23:41 happy Saturday everybody hope you're all
23:43 doing well
23:45 Silver Fox I asked chat the exact
23:48 question and it gave me a learning path
23:49 and job subscriptions or job suggestions
23:52 which
23:53 which question
23:55 has chat the exact question
23:57 I don't know which question you're
23:59 talking about
24:06 we'll get back to it
24:08 I have a prompt engineering wait I have
24:10 prompt engineering on mine oh on your
24:13 LinkedIn
24:14 got a job data automation Tech at twenty
24:17 dollars an hour
24:18 keep talking about it keep putting it
24:20 out there twenty dollars an hour seems
24:22 low to me
24:23 um just but keep going that's good like
24:25 yes far out good for you keep going and
24:30 uh keep knocking on doors because
24:33 as more and more companies start to hit
24:36 the panic button and realize oh [ __ ] we
24:39 gotta figure this stuff out who knows
24:41 about this you know about that I'll pay
24:42 you double
24:43 that's gonna start happening that
24:45 happened in the in in the early days of
24:47 the web I had when when it when it got
24:49 really frothy like like a year or two
24:52 before the the.com bubble burst
24:55 like this like 90s late 97 early 98. I
24:59 had
25:00 three different employees walk into my
25:02 office and say hey I love working here
25:05 but um I just I I was out with drinks
25:10 with some people I didn't know and I I
25:12 talked about some hat idea I had about
25:14 the web and they funded me for 20
25:16 million dollars I had three people in
25:18 one month quit because they got funded
25:20 for like 10 or 20 million dollars
25:23 um so so keep talking about it keep
25:26 putting it out there and keep exploring
25:27 like even if you're even if you got the
25:29 job
25:31 um doing data automation Tech or
25:32 whatever that is
25:34 um just keep learning keep exploring
25:35 keep playing on the weekends just keep
25:37 keep expanding your mind learn what this
25:40 code interpreter can do I
25:43 I can't get my head around it yet it's
25:45 it's too insane what it's doing like
25:50 it's prompting itself writing code
25:52 executing code
25:55 correcting it if it gets it wrong and
25:57 giving you a result and then if the
25:59 result's not right you're like no I
26:00 don't like it it's like okay here I'll
26:02 go again it's like it's like it's like
26:04 it's like a golden retriever it's like a
26:07 coating golden retriever
26:09 thank you for answering my question I'm
26:12 a junior Cloud engineer been trying to
26:14 use it for all I can oh that's awesome
26:16 that's awesome Jenna yeah just I mean
26:18 keep going and the other thing I would
26:20 encourage anyone if if you're sort of
26:22 digging into it use it in your specialty
26:24 use it in the stuff you're doing but
26:26 also start to explore the other tools
26:28 right if you haven't done image
26:30 generation go play with mid-journey
26:33 um if you're geeky install
26:35 um stable diffusion and automatic 11 11.
26:39 um if you're a level geek geekier than
26:41 that start playing with Lang chain and
26:43 start doing your own embeddings and play
26:45 with auto GPT and baby AGI and some of
26:48 these autonomous agents things because
26:50 again the more the more you can get your
26:53 head conceptually around all of what's
26:55 happening I think the better off all of
26:58 us are going to be so it's a good time
27:00 to just geek out just get you know OCD
27:04 and
27:05 ADHD tunnel vision and whatever the [ __ ]
27:08 you do if you're really
27:10 um if you have executive function unlike
27:13 me make a list check it twice and just
27:15 you know go down through the list and
27:17 teach yourself a bunch of [ __ ] if you're
27:19 more on the add side of the thing just
27:21 [ __ ] dick around
27:23 dick around until you find a rabbit hole
27:26 that you want to run down for three days
27:28 and then do that
27:29 can I marry AI what AI stocks you like I
27:33 just talked about a bunch of companies
27:34 that I liked so I I would never
27:38 um take stock advice from me I have a
27:40 degree in acting and I'm shitty with
27:41 money so don't take stock advice from me
27:43 I um all the all the stock [ __ ] I do I
27:47 do it completely emotionally like data I
27:50 mean now that I have chat GPT and and
27:52 code interpreter in particular now what
27:56 I might do is get emotionally hooked on
27:58 some company and then I'll actually go
28:00 run the analysis have it run the
28:02 analysis and just tell me you know what
28:04 I should know about it
28:06 uh but yeah you can probably marry AI I
28:09 would be surprised if we don't see some
28:11 sort of headline in the next two years
28:13 where you know
28:15 someone applies for marriage Hood to a
28:18 to a bot that's happening because it
28:23 quite frankly it probably goes
28:26 someone makes the news because they did
28:28 it first right and then a bunch of
28:30 people do it and then it becomes a
28:32 reality show
28:33 right isn't that America isn't that the
28:35 way we go
28:36 like the first person does it like
28:38 you're a [ __ ] freak you can't marry a
28:40 robot dumb dumb and then two years later
28:42 we're like watching you know
28:45 bot love
28:47 the bot Bachelor
28:52 oh I keep freezing ah [ __ ] should I
28:56 um restart I might need to restart my
28:58 router
29:00 um let me jump jump to the bottom here
29:02 fear and greed control the market AI is
29:04 still young has a long way to go very
29:06 young
29:07 are you concerned with the safety of
29:08 your data with when you download from
29:11 these AIS yes am I still freezing a lot
29:18 fine now okay not freezing for me great
29:21 Amelia's wife sounds like it you must
29:24 have Comcast Xfinity too it's the worst
29:28 but at least it's expensive okay so I
29:31 won't restart anything all right let me
29:33 go back and find where I was in the
29:34 questions if you haven't been here
29:35 before I'd like to answer all the
29:36 questions so I get a little OCD about
29:38 that but you know
29:41 um API key someone asked about API Keys
29:43 okay
29:45 so
29:51 so wait a minute so let me I'll start at
29:54 a I'll start at the I'm going to try to
29:57 keep this as as non-geeky as possible
30:01 um
30:03 at the the base definition of an API
30:05 it's called it's called the app uh
30:07 application programming interface so an
30:10 API is basically a set of instructions
30:12 that says if you want to send something
30:15 to my computer or want to pull something
30:17 from my computer
30:18 he here's the API and what the API is
30:23 um you can ask for
30:25 um let's say let's say that the computer
30:27 holds movie reviews right and so the API
30:32 might say you can ask for movie title
30:36 um number of reviews you know average
30:38 ratings and the text of the reviews or
30:41 whatever it might be and then you know
30:44 that that's if you're asking things here
30:46 and then there might be another API that
30:48 says you can submit a review and you
30:50 could submit a movie title you could
30:52 submit a review you could submit you
30:55 know the rating whatever it might be so
30:57 the API is just the set of instructions
30:59 to to say if you want computer a to talk
31:02 to computer B
31:03 that uses an API the API key is is is
31:08 basically your magic token and I'll show
31:12 you one in a second well actually I
31:14 won't because if I show it to you let's
31:17 see do I have any old ones
31:20 well I can you know oh you know what
31:21 I'll do here we'll go to we'll go to
31:23 chat gbt hang on hang on one second
31:34 all right
31:35 um
31:38 um create 10
31:41 fake API keys so I can see what they
31:47 look
31:49 like
31:52 okay so API Keys tend to look like you
31:56 know how sometimes when you when you go
31:58 to someone asks for a password and then
32:00 and then the your password software goes
32:03 do you want a strong password and it
32:05 makes some kind of random ass thing if
32:07 they're kind of like that so so it's
32:09 it's generally a hexadecimal string of
32:11 numbers and sometimes it's got a little
32:13 letter and dash at the beginning of it
32:15 like uh like uh open AIS I think have SK
32:18 or something like that and then it's
32:20 just a big long list of numbers this big
32:22 long list of numbers is essentially
32:25 the the key that Associates you your
32:31 account so so
32:32 if let's say
32:35 you want to do API calls to the open AI
32:38 thing you want to write in you want to
32:39 write a chat bot and you want that chat
32:41 bot to send [ __ ] to open Ai and get back
32:43 a response well you put in your API key
32:46 and what the API key does it says
32:49 securely okay
32:51 this person has permission to use the
32:53 open AI API
32:55 and then when one of their customers
32:58 uses their application we're going to
33:00 pass that key along and then we'll know
33:02 that we're going to charge your account
33:04 you know 0.2 cents every for every
33:07 thousand tokens that that customer uses
33:09 so it's literally your key to be able to
33:12 use the API
33:14 so how you get it for for open AI
33:18 is you go to this thing called The
33:21 Playground so you go to
33:23 platform.openai.com playground
33:26 and this is where you do your prompt
33:28 engineering you can you can choose
33:30 different models like you can choose I
33:33 now have access to gpt4 woohoo so does
33:36 everyone I'm not special I've got GPD
33:38 3.5 turbo 16k and 3.5 turbo they all
33:42 have different pricing and things like
33:44 that you can read it in the
33:45 documentation but then you go to your
33:48 account and you say view API keys and it
33:52 automatically doesn't display the full
33:54 thing here but basically you say create
33:56 a new key
33:57 and then it gives you the key and you
34:01 copy and paste that and then for example
34:03 where you might use that is
34:07 let's go to zapier
34:14 oops
34:18 it's like I'm going to zapier but not
34:23 come on
34:25 Internet is slow tonight
34:28 so so what zapier is is you've you've
34:31 got these little automations that you
34:32 make out of these kind of Lego blocks
34:34 and I'll show you what I did
34:36 this is what I did the other day as an
34:38 example
34:40 but so
34:43 um here's a Google form so basically
34:45 when someone inputs things into a Google
34:47 form it's going to take the input from
34:49 that Google form and then I'm going to
34:51 send it out to open AI
34:54 and I've got a prompt in here if I look
34:56 at the thing here's here's my prompt
34:58 right so
34:59 you know I want you to use the following
35:01 variables and then I got a variable
35:03 there and then I've got a a prompt write
35:05 a poem about a funny colored animal in
35:08 the style of Edgar Allan Poe so this is
35:11 going to get sent to open AI when I when
35:14 I first set this up with my account it
35:18 asks me for my open AI key or the API
35:22 key
35:23 so now that I've established it I don't
35:25 have to put it in anymore but but when
35:26 you first do this and connect your
35:28 account to it it's going to say where's
35:30 your API key and then you will have
35:33 copied that to your clipboard and then
35:35 you paste that into your thing and then
35:37 you're off to the races
35:38 so hope that helps
35:47 I mean here's the good news
35:50 the good news is we're early and
35:55 you know if you want to do things like
35:57 automations with zapier you kind of need
35:59 to know what an API key is but you don't
36:01 need to get much more geeky than that
36:02 code interpreter
36:04 I feel pretty [ __ ] stupid right now
36:06 with code interpreter because I don't
36:08 know [ __ ] about data analytics and it's
36:11 really good for that but there's going
36:14 to be all sorts of things it's good for
36:16 like like people are just now starting
36:17 to talk about you know examples other
36:21 than the examples we've heard about for
36:23 the last two months like all the
36:24 examples have been the same worst I'm
36:26 starting to see some new ones crop up
36:28 but I I feel like in the next month or
36:29 two
36:30 as I get my head around it like I'll be
36:33 able to share more about it because
36:36 because it
36:39 yeah
36:40 it's so
36:43 it's so powerful like I like it there's
36:47 probably so much it can do I can't even
36:50 comprehend right now I mean I know that
36:52 I know it is I know it is so anyway I
36:54 don't know that's what's going on so I'm
36:56 feeling so so here's the thing
36:58 the thing about these Technologies is
37:00 even though chat GPT is this simple
37:03 Google like thing
37:06 um
37:07 where it's like well it's actually
37:08 really simple well it's actually not
37:12 chat.openingi.com chatgpt is a simple
37:15 interface sitting on top of Decades of
37:19 you know deep computer science and deep
37:22 engineering and all of this stuff is is
37:27 um evolving in real time and and it's
37:30 incredibly powerful stuff so in in the
37:34 current state of the generative AI world
37:36 if you're not coming at it from the
37:38 technology side if you're coming at it
37:39 from like the end user side you know
37:41 your project manager your you know an
37:44 art director you're a copywriter you're
37:46 a pro whatever all that sort of stuff
37:49 you're gonna need a little get a little
37:51 geeky right now to get good with it like
37:53 if you really want to start to do things
37:54 like automations and things like that
37:57 um but if you wait for a year
38:00 um It's Gonna Get Easier and easier and
38:02 easier right these things are ultimately
38:04 going to turn into our agents they're
38:06 going to be like little personal
38:07 assistants we have we're just like hey
38:09 Sally why don't you go you know
38:12 fire up you know
38:14 you know these emails and then go order
38:17 my groceries and come up with a recipe
38:19 for tonight and
38:20 it'll run off and do that [ __ ]
38:24 um but right now we kind of have to all
38:27 build that stuff ourselves the Bills
38:30 should use AI
38:33 could they win Super Bowl
38:38 that's funny what do you feel are the
38:40 most impactful AI startups talked about
38:42 that
38:43 um Celine bugs me too Whitney gets my
38:45 vote as well yes Whitney's it all right
38:48 how to build an app no code experience
38:50 so I just kind of showed that
38:53 um
38:54 the way the way zapier works is there's
38:58 three things you need to know with
38:59 zapier
39:00 there's an input
39:02 and then there's you sort of do
39:04 something with the input and then
39:06 there's an output it can get way more
39:08 sophisticated than that but that's the
39:10 basic concept
39:11 and they call the input a trigger and so
39:14 the trigger is something like a new row
39:17 gets added to a spreadsheet or a new
39:21 um document is uploaded into a folder or
39:25 somebody fills out a Google form that
39:27 you created so the the content
39:29 generation engines that I've been doing
39:31 are I'm using Google forms so I'll
39:34 create a Google form that'll have a
39:36 couple of inputs
39:37 I then
39:39 go into zapier and I say okay I'm going
39:42 to connect it to that form and when
39:44 someone puts in new data in that form I
39:47 want you I want that to act as a trigger
39:48 you're going to take those inputs from
39:51 that form I incorporate them into a
39:54 prompt
39:54 that gets sent with the open AI little
39:58 Lego block from zapier that gets sent to
40:01 open AI I get back to response and then
40:03 I put it in an email or I send it to a
40:05 slack Channel or something like that so
40:07 just at its most basic there's an input
40:09 you process it and in this case send
40:12 that stuff off to open AI you can do
40:13 lots of things in there but that's what
40:15 I've been doing lately and then you do
40:17 something with it you save it to a drive
40:19 you add a row to a spreadsheet you put
40:23 it in an email something like that
40:26 um there's lots of no code tools here
40:28 here's the deal with no code tools
40:32 while there are no code they're not
40:35 um
40:36 no complexity they tend to be you know
40:42 they're they're conceptually as
40:44 complicated as coding is
40:47 because you still need to think about
40:48 the logic okay I need these inputs do I
40:51 have someone fill in a spreadsheet do I
40:54 have them post something to a slack
40:56 channel do I create a form
40:58 is it is it going to be an input on my
41:00 website is it something coming out of
41:03 Salesforce it could be all those right
41:05 so you gotta you gotta think logically
41:07 and critically and yeah and then you
41:09 think okay now that I've got the the
41:11 starting point what's the prompt that
41:13 I'm going to write and how how am I
41:14 going to incorporate that data into that
41:15 prompt that's a whole nother sort of
41:17 logical [ __ ] crazy straw you got to
41:20 go through and then like okay now that I
41:22 got open AI responding with some sort of
41:25 response and it does it I got the
41:27 response I want cool awesome
41:30 um now what am I doing with it how do I
41:31 get it back to the person or do I get it
41:33 back to the person all that sort of
41:34 stuff
41:36 if you look at what code interpreters
41:39 doing this new thing it's doing a lot of
41:42 that logical thinking you're just kind
41:43 of giving it a goal
41:45 and then it goes and
41:47 does some [ __ ] now zapier just added a
41:52 large language model input where you can
41:54 you can just describe the automation you
41:56 want and it will attempt to build you
41:58 that little Lego block thing
42:01 I find that I find that harder than just
42:04 doing it but that's probably just
42:06 because I started doing it
42:08 um
42:09 the the tools that I know of that are
42:11 worth looking at zapier's been around
42:13 for I think 10 years now they're pretty
42:16 good but it's it's it's janky
42:19 um there's one called make m-a-k-e
42:22 that's like zapier but it's kind of got
42:24 a Slicker interface and it seems like it
42:26 might be a bit more modern in terms of
42:28 its structure but I I don't know
42:30 and then if you want to do apps there's
42:33 one called Bubble and then there's a
42:35 couple of others people have mentioned
42:36 in the live here
42:38 um but but I would just you know go uh
42:40 go to Reddit or go to Twitter and look
42:43 up no code tools and just find one that
42:45 doesn't suck and doesn't make your head
42:47 explode
42:48 definitely not a robot
42:51 I'm here for the wine exactly I although
42:54 I'm I should go get a actually there's
42:56 there's a bottle of wine out my kids had
42:58 some wine tonight maybe I'll get a wine
43:02 why not
43:03 why not get it wine
43:06 why not
43:13 all right
43:15 it was like a dad joke but I have enough
43:19 self-awareness to know that it was a
43:20 stupid dad joke so I made fun of myself
43:23 making a dad joke get it
43:29 what AI can summarize audio files uh
43:33 well uh summarize so whisper so if you
43:37 go I was showing before let me switch my
43:41 um
43:43 good playground so if you're at open
43:46 ai's playground so again the the URL for
43:49 this is
43:50 platform.openai.com playground
43:53 and if you come here
43:57 and you go to the complete mode
44:02 in the upper right hand corner of this
44:04 window is a little microphone
44:06 and if you click on that
44:08 that's whisper
44:10 which is open ai's audio transcription
44:13 you can upload a video or an MP3 or a
44:16 web you know
44:17 web audio
44:19 and
44:21 um or you can just record stuff and it
44:23 will take that audio and transcribe it
44:26 for you you can also do that via the API
44:30 so you could you could probably do audio
44:33 transcription through zapier
44:36 um
44:36 there are also lots and lots of tools
44:40 um where they put it in a in a sort of
44:43 easier front end if you go to
44:44 futurepedia.io
44:46 just look up transcription or audio or
44:49 things like that there are probably I I
44:52 would be surprised if there were less
44:54 than a hundred Tools in futurepedia.io
44:58 that do that I tried uploading audio
45:01 file and I couldn't it couldn't
45:03 summarize it if it's too big you may
45:05 need to break it into chunks that's
45:08 again where I'd go to futurepedia.io and
45:10 and look for ones that say you can
45:12 upload up to two hours of a file right
45:17 um so so there's lots of things out
45:18 there that you can use the the open ai's
45:21 Whispers really really good but even in
45:24 the playground it's limited to a certain
45:26 length so you may need to chunk it up
45:30 but
45:31 that should help
45:34 all right
45:36 I don't know I tried to use it for C
45:39 sharp and it failed a lot
45:42 then it comes to wait file to file
45:45 reference
45:46 not quite sure what you're talking about
45:48 there
45:50 um
45:51 yeah depending on what you're trying to
45:53 do
45:54 um GPT can be a mess with programming
45:56 you know I would say if you're trying to
45:58 do serious coding stuff I'd go to GitHub
46:00 and use copilot
46:03 um because it's more it's it's specially
46:04 designed for coding
46:07 um
46:08 so
46:09 is code interpreter available through
46:11 the open AI API
46:14 I don't know where's the bottom I I
46:16 don't think so but I don't know like I I
46:18 didn't dig into that today that's that's
46:20 a question that I have as well
46:23 um
46:24 I I assume it will be but I don't know
46:30 so it should be
46:32 are all plug-ins accessible through the
46:35 API I don't think so I don't think they
46:37 are right now the plugins right now are
46:38 a little little janky you can make your
46:41 own plug-in
46:43 so wait if you can make your own plug-in
46:46 you're interacting with the API so they
46:49 probably are I don't know I don't
46:50 [ __ ] know
46:51 but but you know how you now have
46:54 function calling within the API right
46:57 within the openai API so you can kind of
46:59 write your own plug-in function calls
47:03 plug-in like function calls within the
47:05 code
47:06 um I haven't done I haven't I haven't
47:08 needed to and so because of that I
47:10 haven't dug into that part of the the
47:12 world
47:13 um but if you find out let me know that
47:15 would be good to know I don't have code
47:16 interpreter yet lasai oh that sucks
47:19 um I just tell you
47:22 um did you just cabrono I assume
47:27 um
47:28 I assume you've gone to your settings to
47:30 just to to your uh beta features and and
47:33 enabled it
47:35 if you haven't
47:37 um go do that
47:39 how I could work in AI
47:42 um
47:43 I don't know what you mean silver fox
47:46 how can you get a job
47:48 I don't know start learning it start
47:50 talking about it
47:51 would there be some time when AI helped
47:54 network security yeah
47:57 and and there's also a time when AI
48:00 gonna attack network security
48:03 there's going to be all sorts of new you
48:05 know white hat and black hat battles
48:07 going on where where people are using AI
48:10 increasingly to exploit vulnerabilities
48:12 and then other people are going to be
48:15 using AI to catch the bad guys and
48:17 they're going to be able to catch the
48:18 bad guys in increasingly sophisticated
48:21 ways so yeah that's coming
48:24 um do me a favor we're kind of dwindling
48:25 down here if you would just hit the uh I
48:27 know it's been a little boring here for
48:28 a section I've been geeking out a little
48:30 bit but um just hit the like tap the
48:32 screen get some likes going and if you
48:33 would share the live so we can just let
48:35 let some people know we're here that
48:37 would be great
48:38 uh
48:40 um are you concerned with safety of your
48:41 data with the download of these AIS um
48:44 yes so so I wouldn't put anything super
48:48 critical in there
48:50 um
48:51 I I'm not super worried about it you
48:54 know for me personally but I do a lot my
48:57 company does a lot of work with Pharma
48:58 clients and they're you know
49:00 appropriately anal about their safety
49:03 and privacy and and things like that so
49:05 anything we're doing for clients
49:08 um we're going to be creating you know
49:10 Standalone servers and firewalls and all
49:13 that sort of [ __ ] so
49:15 um
49:16 so if if you need to do
49:20 proprietary kind of stuff and it's
49:22 actually super critical then you need to
49:24 look into that you know go go look at
49:27 Microsoft Azure like how they've
49:29 Incorporated open AI calls and the
49:32 options they have for for like like
49:34 stability or I mean open AI announced
49:37 three months ago that they've got these
49:39 business servers that you can buy that I
49:42 think they start at like 25 Grand and go
49:44 up to a million where you can have your
49:46 own Standalone instances of open AI
49:48 servers
49:50 um so
49:51 thank you for all the hearts that's
49:53 appreciate that yeah AI is still young
49:55 we have a long way to go totally Ryan it
49:57 is we are we are in such early days if
50:00 you're feeling at all like I'm behind I
50:03 can't all these people know more than I
50:05 do my sister's I can't I can't right now
50:08 because I don't know anything nobody
50:10 [ __ ] knows anything
50:12 if you know the engineers the people on
50:14 the science side of this there's lots of
50:16 people that know [ __ ] there but on the
50:17 on the post chat GPT side of this where
50:20 people are now incorporating this into
50:22 their businesses this is brand new for
50:25 [ __ ] everyone it's like like even the
50:28 even the uh the people in the data
50:31 science side of this
50:32 um I I was watching more of the
50:34 interview with George hotz who's the
50:35 who's the comma AI guy and
50:38 you know super super smart programmer he
50:41 said he said the the the difference
50:43 between how he was using AI five years
50:46 ago and what's happened in the past six
50:48 months he said it's like it's not
50:50 anywhere in the same ballpark of of how
50:54 he thinks about Computing anymore
50:57 um it's a completely different thing so
50:59 all this generative AI stuff the Chachi
51:02 PT stuff and the the large language
51:04 models this idea of you know all the
51:06 stuff the the the image generation
51:09 things and people reading brain scans
51:11 and music being written and and songs
51:14 that sound like this artist and that all
51:16 this [ __ ] is new all of it's new and and
51:19 it's you know
51:21 the the the the popular press has heard
51:24 about it has been talking about it for
51:26 the last six months since November 30th
51:29 but it's really only been this stuff's
51:30 been kind of in The Ether for maybe a
51:33 year and a half maybe two at tops we're
51:36 very very new and this stuff is
51:39 transformative
51:41 fear and greed control the market I
51:43 would agree with that but uh this
51:46 channel exists to inject a little
51:48 optimism in the middle of that fear and
51:50 greed so [ __ ] those [ __ ]
51:52 don't be scared AI can be your friend
51:59 got a bot and his name is Charlie
52:01 Charlie makes me nice stories in pretty
52:04 pictures
52:05 um
52:06 about is probably more emotionally
52:08 intelligent than all the broken toys on
52:10 hand that's [ __ ] genius I'm telling
52:14 you man the the what's cool about the
52:17 Bots is that the the Chachi BT things is
52:20 um
52:23 you can ask them questions about really
52:26 sensitive topics you know especially in
52:29 today's environment
52:30 and it doesn't get triggered it just
52:34 responds with like a balanced answer
52:36 like just with with like nuance and like
52:39 well you know here's an opportunity and
52:41 or here's an option and here's an option
52:43 and
52:44 and it's just way more nuanced and it is
52:46 it's way more like like
52:49 these language models talk with just a
52:51 [ __ ] ton more
52:52 empathetic response because they don't
52:54 have empathy but they respond more
52:57 empathetically than almost all humans so
52:59 the botular that's good
53:02 um not freezing for me fine now okay
53:04 great
53:05 um not freezing for me this is the
53:06 freezing stuff
53:13 trying to find my thing
53:16 where was I oh there's the fear and
53:18 great fear and greed
53:20 Fear and Loathing in Las Vegas did y'all
53:23 see the the [ __ ] sphere in Las Vegas
53:25 that that new venue with with like the
53:28 LED outside of it holy [ __ ] is that
53:32 thing amazing
53:34 it's crazy like I saw one picture of
53:36 someone driving down the Las Vegas Strip
53:38 and and the sphere was like at the like
53:40 at the end of the road it looked like
53:42 the Moon Rising but it was like an eye
53:44 blinking I'm just like
53:47 [ __ ] trippy man
53:52 um
53:52 and tokens are okay tokens good question
53:57 so
54:02 this is painful it's like why do they
54:04 call it code interpreter because that's
54:06 the technical name of what it's doing is
54:08 code interpreting
54:11 it's a shitty marketing name and so are
54:13 tokens okay so tokens are thank you for
54:17 that
54:20 how do I describe it okay when you type
54:22 a prompt that prompt gets turned into
54:24 numbers and and then those numbers are
54:28 sort of put into this big fancy s
54:30 mathematical formula that then
54:32 determines the next most probable number
54:34 called a vector and then that Vector is
54:36 associated with a word so all of the
54:38 [ __ ] going on behind the scenes with
54:42 um
54:42 with chat GPT
54:44 are
54:46 Vector addresses they're basically these
54:49 long strings of numbers
54:51 and so tokens are you know those those
54:55 Vector calls basically and I think
54:58 spaces and words or parts of words all
55:01 end up being tokens
55:03 so how they charge someone before was
55:06 asking about apis the the open AI apis
55:09 how they charge for the API is for every
55:11 thousand tokens you get charged two
55:14 pennies or a tenth of a penny or
55:16 whatever the hell it is
55:18 um
55:19 that's just their definition of the
55:21 chunk that they're sending
55:23 but it actually translates into like for
55:26 us as humans
55:28 if you've got a thousand tokens that are
55:32 being sent that actually translates it's
55:34 about three quarters the amount of words
55:36 so 75 or 750 words is about a thousand
55:40 tokens so it's not a direct thing
55:43 and why don't they don't just call it
55:45 words well because they're engineers and
55:48 scientists and they're called tokens and
55:50 the token is the mathematical thing
55:52 right
55:53 so so that's what so tokens are
55:55 basically just kind of your word count
55:57 and so chat gbt right now has a um a
56:01 token limit of 4 000 tokens or about
56:04 three thousand words
56:05 three quarters of the amount
56:08 so what that means is you've kind of got
56:10 this window of 3000 words that as you're
56:12 typing into chat GPT
56:15 that's its context window so that amount
56:18 is is essentially what's being used to
56:21 to determine the answer and that's why
56:23 sometimes if you have a long
56:24 conversation with chat GPT it seems like
56:28 it like you you started out you told it
56:30 about your lovely wedding and then it
56:32 seems you know 10 minutes into the
56:34 conversation it doesn't seem to have any
56:35 [ __ ] clue about the wedding anymore
56:37 that's because the Wedding part like
56:39 drifted out of the context window
56:42 so that context window is growing
56:45 um the the some of the API
56:48 um uh things for for openai right now
56:51 are sixteen thousand tokens which is
56:53 twelve thousand words so that's starting
56:54 to be better and then ultimately gpt4
56:57 can do 32 000 tokens which is about 25
57:00 000 words and there's if you go to
57:02 poe.com the that one right there
57:06 um
57:08 there's one of the Claude models that
57:11 you can actually use has a hundred
57:13 thousand token limit and that's 75 000
57:16 words and it's it's the it's 75 000
57:18 words is the combination of your prompt
57:20 and the response
57:21 it's got that it's got a big token
57:24 window so you could you could literally
57:26 paste a whole screenplay or a whole
57:29 novel in there and start interacting
57:31 with it right but you can't with chat
57:33 GPT because it's got a much smaller
57:36 token context window so that's what
57:38 tokens are
57:41 tokens are what you're charged yes
57:43 tokens are what you're charged and they
57:45 are Loosely related to the amount of
57:46 words you can use
57:49 it's complicated as like like so so
57:52 here's the deal all of this stuff we
57:55 will not need to learn about
57:58 um
57:59 two years from now this stuff's gonna
58:01 all drift into the background the reason
58:03 it's all in the foreground right now is
58:05 because we're early right and so if you
58:08 want to play with this stuff you kind of
58:09 got to get a little geeky you gotta
58:10 understand what tokens are
58:12 you think when Apple launches their AI
58:14 agents they're going to mention anything
58:16 remotely like [ __ ] tokens nope the
58:19 developers might need to deal with terms
58:21 like that but none of the tools will
58:23 will be like this for long
58:26 um so no need for court reporters I I
58:30 Court I
58:31 I there no need for them or like like I
58:34 look at I think one of the one of the
58:36 areas
58:37 um
58:38 that where I think America has been hurt
58:41 the most in the past
58:44 I guess it's been happening for about a
58:46 decade
58:47 but the lack of local news local
58:50 reporters because all these National
58:52 conglomerates bottle bought up all the
58:54 local stations and basically just gutted
58:56 the newsrooms
58:58 um
58:59 you don't have local reporting anymore a
59:03 lot of what the local reporting was was
59:05 reporting on public documents and you
59:08 know their you know permits and you know
59:11 the building they're going to build and
59:13 the bridge they're going to do and are
59:14 they going to fix this and are they
59:15 going to fix that and how they're
59:16 changing streets and when are they doing
59:19 the water mains
59:21 um so a lot of that reporting is not
59:22 getting done but a lot of that reporting
59:24 could be automated with with these tools
59:29 um because if you got access to the
59:31 public records
59:33 you could write prompts that parse those
59:36 public records and write stories and get
59:38 the word out right so so I think it's a
59:40 double-edged sword you might not need
59:42 court reporters but you might be able to
59:44 dramatically increase Court Reporting
59:46 and local reporting and things like that
59:50 all right C sharp like using program
59:53 code to run from other file the main
59:56 file and using other file for info how
1:00:00 um
1:00:02 I don't quite understand your question
1:00:04 I'm also not an engineer so I'm not the
1:00:06 right person to ask
1:00:07 the the right person to ask
1:00:10 if you don't have a engineer buddy is
1:00:13 right there so instead of just having
1:00:15 chat GPT try to do the coding
1:00:18 have chat GPT tell chat GPD to act like
1:00:22 a senior c-sharp engineer
1:00:25 and and see if you can get it to tell
1:00:28 you how to use it to do that
1:00:31 because I'm sure it knows it might just
1:00:34 this might just be a sequencing thing if
1:00:36 you can get it to tell you how it would
1:00:38 do something in a step-by-step fashion
1:00:40 and then have it start to do that for
1:00:42 you that might help it might need that
1:00:44 context first it's one of the things
1:00:47 about these tools is they don't do well
1:00:49 with sparse instructions if if you just
1:00:51 say you know write me a poem it's going
1:00:54 to write a really generic poem If you
1:00:56 tell it you want it to you know act like
1:00:58 Emily Dickinson and
1:01:00 you know it's a master of her works and
1:01:04 her tone and her this and her that and
1:01:06 then you ask it to write a poem it's
1:01:08 going to be you know much more what
1:01:10 you're looking for so it's no different
1:01:12 than the programming stuff
1:01:13 I got it yesterday are there are there
1:01:16 rumors that being AI will have plug-ins
1:01:18 this would be a game changer against
1:01:20 open AI it's not going to be against
1:01:22 openai because
1:01:24 um so so Bing is from Microsoft they
1:01:28 have already announced it's not a rumor
1:01:29 they've already announced they're going
1:01:30 to support the openai plug-in
1:01:32 architecture that means chat GPT plugins
1:01:35 will work in bing and they're going to
1:01:37 work in all of Microsoft's office
1:01:41 Windows 11.
1:01:43 um
1:01:44 Dynamics 365 uh
1:01:48 GitHub co-pilot all of the Azure cloud
1:01:51 services
1:01:53 um
1:01:54 Microsoft invested 10 billion dollars in
1:01:57 open AI so even though they may have
1:01:59 competitive products like there's
1:02:00 there's some stuff that openai's looking
1:02:03 to do that may be competitive with some
1:02:05 of the applications that Microsoft is
1:02:07 doing they're they're
1:02:09 um they they call it in the business
1:02:11 World Co-op Edition
1:02:13 um there might be some areas where they
1:02:15 compete and there's going to be other
1:02:16 areas where they're deeply deeply
1:02:17 cooperating so that's where they are
1:02:19 right now so yes Bing will support
1:02:21 plugins and it is going to support open
1:02:25 AI plugins they're trying to make it a
1:02:27 standard I assume
1:02:29 and and the way if you look at the the
1:02:31 nature of how the plugins are created
1:02:34 they're very very
1:02:36 um
1:02:37 they're very very it's a very very
1:02:39 simple protocol so it would be a good
1:02:42 standard for lots of people to adopt
1:02:45 they're very easy to write they're very
1:02:46 easy to call you're basically just
1:02:48 describing
1:02:50 here's what this plugin does and here's
1:02:52 when you would use it and then that's
1:02:53 what chat GPT uses to determine
1:02:56 you know should I use a plug-in here or
1:02:59 can I just answer this
1:03:01 um so yeah
1:03:03 how do you sum object or function in a
1:03:06 file
1:03:08 um again I don't know I don't know the
1:03:10 function calling stuff I haven't gotten
1:03:12 into that yet and I probably won't it's
1:03:14 that's a level deeper than my pay grade
1:03:16 and my knowledge
1:03:18 so so just using Code interpreter right
1:03:21 now is is sort of bouncing up the the
1:03:24 upper limits of my intellect it's making
1:03:26 me feel like a [ __ ] idiot it was
1:03:28 funny I saw that on Reddit today so
1:03:31 someone someone put on Reddit
1:03:33 um does anyone else feel you know like
1:03:35 immediately stupid again yes
1:03:40 I was answering something from a long
1:03:42 time ago okay Silver Fox cool
1:03:44 um y'all hiring nope sorry
1:03:46 I wish we were
1:03:48 we're we're stable as they say which
1:03:51 Stables better than laying people off
1:03:52 right
1:03:57 where's the bottom
1:04:01 I got a bottom for you right here
1:04:05 name's Buck that's my bottom
1:04:12 a bot is oh
1:04:15 okay hang on hang on I lost my place
1:04:18 again
1:04:19 old man trying to read little text on
1:04:22 tick tock
1:04:24 it's the Tickety Tock it's the tickety
1:04:26 tog I'm playing on a Tickety Toc oh wait
1:04:30 now I'm I'm way far behind or like back
1:04:33 where I was whatever I'm still lost hang
1:04:36 on
1:04:37 I saw them building it in February the
1:04:39 biggest fear eyeball that thing's cool I
1:04:42 mean I was I was thinking about that
1:04:44 like like like the uh
1:04:48 AI like degenerative AI stuff and the
1:04:51 generative video stuff
1:04:53 you're gonna
1:04:55 they're gonna be able to come up with a
1:04:57 generative AI thing to just dynamically
1:04:59 generate stuff for that that sphere like
1:05:01 it's just gonna be insane
1:05:03 it's so cool and apparently the inside
1:05:06 of it's going to be even better
1:05:09 I think a word is a token but so is each
1:05:13 punctuation mark yeah it's something
1:05:14 like that but I think I think longer
1:05:16 words they even break up into multiple
1:05:18 tokens it's just it's whatever like it
1:05:21 doesn't really [ __ ] matter for for
1:05:22 real people like it matters for for
1:05:24 engineers right because because you know
1:05:27 they need to like for them a token is a
1:05:29 token and you know
1:05:31 they're doing whatever they're doing
1:05:32 with the token you know certain tokens
1:05:34 are getting thrown out other tokens are
1:05:38 creating probabilities of you know
1:05:39 semantic waitings or whatever
1:05:42 I don't [ __ ] know but but it's
1:05:44 basically three quarters however many
1:05:46 tokens it is three quarters of that is
1:05:49 about how many words it is but it but it
1:05:51 ain't exact
1:05:53 the cost depends on the model you're
1:05:54 using yeah the like the the GPT 3.5
1:05:57 models now are super super cheap gpt4 is
1:06:00 more embedding is more but that also
1:06:02 came down in price
1:06:06 Kyle's saying words though I am saying
1:06:09 words
1:06:12 what'd you do Saturday night
1:06:14 I watched some old guy say words what
1:06:16 did he say I don't really know
1:06:19 I don't know he seem to enjoy himself
1:06:21 Anna
1:06:23 I don't what am I doing with my life
1:06:26 what what am I doing
1:06:28 I need a drink
1:06:30 all right should I go get a wine I think
1:06:32 I should go get a wine yes it's so true
1:06:33 I was a small town newspaper reporter
1:06:36 for six years right
1:06:38 like I mean holy [ __ ] like
1:06:41 like all these National [ __ ] private
1:06:43 Equity [ __ ] that came in and just
1:06:46 [ __ ] decimated
1:06:48 um
1:06:49 local reporting and local news and local
1:06:52 journalism
1:06:53 it's like [ __ ]
1:06:59 um
1:07:00 you know it just feels like there should
1:07:02 be some boundaries in [ __ ] society
1:07:05 where you like you know
1:07:07 that's not for profit
1:07:09 you can make a profit but it's not its
1:07:12 sole purpose
1:07:15 there should be laws but we would need
1:07:18 to have a functioning legislature to
1:07:21 have those wouldn't we
1:07:23 bitter you
1:07:25 all right
1:07:27 automating public interest stories could
1:07:29 be awesome yeah I know it's I I like
1:07:30 that is I feel like that's a
1:07:33 that's a business that's business you
1:07:35 want to go start a business I've been
1:07:36 I've been giving out free business
1:07:38 launching advice that's one
1:07:40 go start the local news bot business
1:07:43 business
1:07:49 [Music]
1:07:52 again I've never stopped feeling stupid
1:07:54 I know me too I've like I feel like I
1:07:57 feel like oh I know everything and then
1:07:59 I go to try to do something and I fail
1:08:01 miserably like I don't even know where
1:08:03 to start what do I do [ __ ] it's just
1:08:07 it's been like that for for I don't know
1:08:08 it's like a year now
1:08:10 it's
1:08:12 like I even feel like on these on these
1:08:15 lives sometimes I'm like like am I
1:08:17 actually is anything progressing or am I
1:08:20 just saying the same [ __ ] over and over
1:08:22 again and then I realized like well
1:08:24 some people are still just starting but
1:08:26 also
1:08:27 like the fact that code interpreters
1:08:30 here now like that's going to completely
1:08:31 change how I use chat gbt I don't know
1:08:33 how yet but it is
1:08:36 and and so and and then things like they
1:08:38 take away
1:08:40 they took away browse with Bing and then
1:08:42 I saw on Twitter today someone else said
1:08:44 oh they also took away plugins but they
1:08:46 didn't for me
1:08:48 so so this stuff is just changing so
1:08:51 much that I think it's just I I think
1:08:53 the
1:08:55 the thing for all of us to do if you're
1:08:58 willing to do it is just think of it as
1:09:00 an adventure like we're on a [ __ ]
1:09:02 adventure and and what are Adventures
1:09:03 well they're super exciting what else
1:09:06 are Adventures they're super [ __ ]
1:09:08 challenging okay what else are
1:09:09 Adventures sometimes they're really
1:09:11 scary right sometimes they're really fun
1:09:15 but like but you're but you're committed
1:09:18 to it right I'm gonna get the Grail
1:09:21 we're gonna storm the castle right
1:09:24 whatever the [ __ ] Adventure is we're
1:09:25 gonna go kill the dragon
1:09:28 I think that's what I think that's the
1:09:30 mentality we have to be on for at least
1:09:32 two years because it's just the stuff's
1:09:35 moving too fast
1:09:37 and it's it's like the it's just the
1:09:41 changes are going to be so profound
1:09:44 that that
1:09:46 I don't know like like you know the the
1:09:48 people that were asking earlier about
1:09:49 well what job can I get how do I get a
1:09:51 job and where do I get a job I don't
1:09:53 know
1:09:54 I don't know like I feel like the thing
1:09:56 that's going to be super in demand in
1:09:58 the next six months is gonna be either
1:10:00 non-existent or just a completely
1:10:02 different set of requirements six months
1:10:04 out
1:10:06 it's crazy it's crazy I tell you my head
1:10:09 hurts welcome welcome to the welcome to
1:10:12 the adventure I know my head's my head
1:10:15 has been
1:10:17 like like I like I'm just
1:10:21 bursting at the seams educationally I
1:10:24 feel stupid every day I I feel like I'm
1:10:27 in a constant state of
1:10:29 it's not quite anxiety and it's not
1:10:32 quite shame but it's just that like that
1:10:34 thing where you know you're not good at
1:10:35 something but you really want to be good
1:10:37 at it
1:10:38 and with this stuff it's like I feel not
1:10:42 good at it I really want to be good at
1:10:44 it I'm getting better but I'm not really
1:10:47 good like I just it
1:10:51 I don't know it's hard right now it's
1:10:53 hard
1:10:55 Kyle's saying words versus characters
1:10:57 vis-a-vis tokens yes correct Thelonious
1:11:00 Thelonious Monk you are correct
1:11:04 you are a call to action field trip
1:11:07 buddy indeed this is your call to action
1:11:09 people
1:11:11 get on your Steed and head into the deep
1:11:14 deep dark Forest the one that shall
1:11:16 never be spake
1:11:19 laughs
1:11:21 take your [ __ ] swords go to chatgpt
1:11:26 get code interpreter or go make a bonnet
1:11:29 Poe or go I don't [ __ ] it doesn't matter
1:11:32 it doesn't matter it really doesn't
1:11:34 matter
1:11:35 I mean it matters deeply and profoundly
1:11:38 but what you do doesn't matter and you
1:11:40 know we were talking about this with the
1:11:42 AI Salon I've got this thing called the
1:11:44 AI Salon that if you want to hang out
1:11:47 with a bunch of adventurers
1:11:49 um
1:11:49 come join us go go to that URL the
1:11:52 salon.ai
1:11:54 and
1:11:56 we were talking about what is it about
1:11:58 this group that is
1:12:02 that that we like we've been around for
1:12:04 six months it's it's a it's a really
1:12:07 cool it's it people are generous they're
1:12:09 curious
1:12:11 um they share what they know
1:12:13 um they want to keep learning what is it
1:12:15 about them and and what we came up with
1:12:17 was this idea that we're all on an
1:12:19 adventure and like in a good Adventure
1:12:21 you have people with all different
1:12:23 skills right like you have the the big
1:12:27 dumb [ __ ] muscle guy who turns out to
1:12:30 have a heart of gold and you've got the
1:12:31 the really mousy person that is not you
1:12:35 know doesn't seem to have any skills and
1:12:37 then they get in into like an escape
1:12:39 room and she's brilliant at cracking the
1:12:42 code right like you have all these
1:12:44 archetypes right
1:12:45 and I feel like that's where we that's
1:12:47 that's where we need to be right now I
1:12:49 was like
1:12:50 the more we can connect and find
1:12:54 you know our little group of adventurers
1:12:56 to kind of go out and do this with
1:12:59 so rather than me having to become an
1:13:02 engineer like why don't I pile up with
1:13:05 someone who's really good at engineering
1:13:07 and they can do some of that [ __ ] and do
1:13:09 it really well and then I can do the
1:13:11 [ __ ] I do really well and someone else
1:13:12 could do this stuff they do really well
1:13:14 and and and you know together you can
1:13:17 kind of learn this stuff and at least
1:13:18 learn what's possible with it that's
1:13:20 that's where my head is lately
1:13:23 find your adventurers get on the
1:13:25 adventure get on the horse go
1:13:29 go go go go go go go go go go go go go
1:13:31 is it scary yes is it worth killing the
1:13:34 dragon well [ __ ] yes
1:13:36 because if we done with the dragon's
1:13:37 gonna kill the village
1:13:40 right someone's got to kill the dragon
1:13:46 but that seems scary yeah
1:13:52 is it scarier than the dragon coming and
1:13:54 eating the village
1:13:59 I don't know why I went to you know
1:14:01 fantasy metaphor but I think that I
1:14:03 think that's where we are
1:14:06 [Music]
1:14:06 um
1:14:09 speaking of the legislature
1:14:11 wait
1:14:14 where'd that go
1:14:17 oh here we go
1:14:18 what do you anticipate will be the first
1:14:21 proposed legislation around AI oh [ __ ] I
1:14:24 don't know
1:14:30 if they do as I if if they do as my
1:14:34 cynical self expects them to they'll ban
1:14:37 it
1:14:38 but I think I I think they uh you know I
1:14:42 I think the the Biden Administration is
1:14:44 actually taking it seriously
1:14:46 um
1:14:50 I think it'll probably be something
1:14:53 around
1:14:56 um
1:15:00 companies like open Ai and Google like
1:15:03 like some sort of you know legislation
1:15:05 around regulations around
1:15:08 um
1:15:10 what they're allowed to do what they're
1:15:12 what some boundaries around what's
1:15:14 called alignment which is how aligned
1:15:17 are these powerful models with the goals
1:15:20 of the human race
1:15:22 so I think they'll probably do some
1:15:24 regulations like if you're a company
1:15:26 with as much resources as Google or open
1:15:29 AI or Microsoft or meta
1:15:32 you can't make the evil bot that's going
1:15:34 to kill kill us all right like it it'll
1:15:37 probably be something around that
1:15:41 um
1:15:42 I don't know I don't know
1:15:45 I just I you know our government right
1:15:48 now they don't understand what [ __ ]
1:15:49 Wi-Fi is and they're they're you know
1:15:51 they're just not functioning as a
1:15:54 government right now they're just
1:15:55 they're just functioning as a money
1:15:57 raising machine that's that's you know
1:16:00 polarizing all of us so we don't pay
1:16:02 attention to the grifting that's going
1:16:04 on
1:16:05 um
1:16:07 and I think that this AI stuff actually
1:16:10 requires
1:16:12 good bipartisan
1:16:14 functional government
1:16:17 to like
1:16:19 learn about it learn what it does well
1:16:22 learn how it can be a competitive
1:16:24 Advantage for our country and for the
1:16:27 world
1:16:30 um and get their [ __ ] heads around it
1:16:31 get their heads out of the the [ __ ]
1:16:34 petty [ __ ] that's going on
1:16:36 do I have much hope that that's going to
1:16:38 happen no I don't
1:16:42 [Music]
1:16:43 um
1:16:44 do I have deep hopes that something will
1:16:48 shift yeah I do
1:16:52 um and it's also you know
1:16:57 I think it's why I'm doing these lives
1:17:03 because I don't think we can wait for
1:17:05 government I
1:17:07 what can they [ __ ] do
1:17:09 like like the the open source large
1:17:11 language models are have no safety
1:17:14 restrictions on them and are already
1:17:16 almost as powerful as the stuff that
1:17:18 open Ai and Google have produced
1:17:21 so that [ __ ] ain't gonna be legislated
1:17:24 and then
1:17:27 Europe's gonna do their AI legislation
1:17:29 and America is going to do their AI
1:17:32 legislation
1:17:34 but everyone else in the world is just
1:17:37 going to keep rolling ahead forward so
1:17:39 if either the EU or or the U.S make it
1:17:43 too restrictive they're going to
1:17:45 absolutely kill Innovation and the rest
1:17:47 of the world is going to kind of subsume
1:17:50 that like all the models right and and
1:17:53 and you know the momentum
1:17:55 um and they're gonna do whatever they're
1:17:57 gonna [ __ ] do so it's like
1:18:00 I think that it's
1:18:02 it's really important for us as
1:18:04 individuals to start using these tools
1:18:07 in ways that bend it toward the light
1:18:11 like it's the only way I can think about
1:18:13 it right like go do the adventure we
1:18:15 gotta go kill the dragon but I don't
1:18:17 know how to kill dragons it doesn't
1:18:18 [ __ ] if we don't kill the dragon The
1:18:21 Regulators or someone or China or
1:18:23 whoever something's gonna happen right
1:18:25 so
1:18:27 it's on all of us to start learning
1:18:29 these tools get AI literate and and
1:18:32 start putting [ __ ] in the world that is
1:18:34 bending it toward the good
1:18:36 I think that's gonna have I think it's
1:18:38 gonna make a difference it's it at least
1:18:41 lets me sleep at night right because I
1:18:44 understand it as much as I can
1:18:46 and I'm doing what I can to get the word
1:18:48 out about it and that's why this is here
1:18:51 so
1:18:52 I know it's very philosophical tonight
1:18:54 dragons and
1:18:56 so many business ideas there are
1:18:58 business ideas in August I'm going to
1:19:00 see my 87 year old mom going to
1:19:02 introduce to Chachi PT and watch her
1:19:04 head explode that's [ __ ] awesome I
1:19:07 know I know it's so it's one of my great
1:19:10 joys in life is showing people that
1:19:12 either have no clue what it is or or or
1:19:15 sort of like cocky but dismissive yeah
1:19:18 yeah I've played with chat GPT and then
1:19:20 you actually show them what it can do
1:19:22 and they're like
1:19:23 I can do that
1:19:25 really
1:19:26 could it do my [ __ ] uh-huh I'm gonna see
1:19:29 not really here look
1:19:35 it's [ __ ] the best
1:19:38 because it's like you can't go like once
1:19:40 you once you have that Epiphany about
1:19:42 this [ __ ]
1:19:44 like I don't think you can't go back you
1:19:46 can't unlearn what it does what it makes
1:19:48 possible
1:19:50 but that's going to be awesome that have
1:19:52 fun with that who has time for a job
1:19:54 we've got three three hours of the AI
1:19:56 learning lab every night exactly I know
1:19:58 I feel the same way
1:20:01 this is my job I looked at my I figured
1:20:03 out where to go look at my uh my my
1:20:05 balance for all my earnings I have
1:20:08 earned
1:20:09 um 24.32 so far doing these lives I
1:20:12 can't take it out until it hits a
1:20:14 hundred but you know
1:20:15 so so this this is my job if I on an
1:20:20 hourly basis of making I think I'm
1:20:21 making like 3.8 cents
1:20:26 Thelonious you're evil can you make a
1:20:28 video you know showing this stuff to
1:20:29 your mom it would actually make a good
1:20:30 tick tock learning never ends exactly
1:20:34 my mom is going to sus site she keeps
1:20:36 hitting links the shuts her PC down yeah
1:20:39 that's rough that sounds that's what she
1:20:41 calls exactly control alt delete mom
1:20:45 exactly oh that's hilarious
1:20:55 [Music]
1:21:00 okay
1:21:02 the salon.ai yes that is the correct URL
1:21:06 just it should go to a link tree
1:21:09 um and then maybe just do HTTP instead
1:21:11 of https I probably set up the site
1:21:13 wrong or something like that if it's not
1:21:15 going there but that's the correct
1:21:16 that's the correct URL
1:21:20 um and we've got a there's a meet up
1:21:21 there so we meet every other week
1:21:24 um on a Google meet and in person here
1:21:26 in Denver and then we've got a Discord
1:21:28 where you can play with mid-journey and
1:21:30 ask questions and hang out and then
1:21:33 we've got a LinkedIn Group which isn't
1:21:35 as active but it's just a different kind
1:21:36 of group
1:21:37 do you think we can use our GPU to
1:21:39 provide computing power for AI just like
1:21:41 Bitcoin mining yeah sure if you've got
1:21:43 like a 30 90 if you're on a PC I'm a Mac
1:21:46 guy so if you've got a decent GPU yeah
1:21:50 that can be used for this kind of stuff
1:21:51 and it's you know
1:21:53 um
1:21:54 if you want to use anything big like a
1:21:56 65 billion parameter
1:21:59 model you need
1:22:01 Big Iron
1:22:04 like George hotz the guy from Kama he's
1:22:06 he started a company called tiny Corp
1:22:08 and they're creating a computer called
1:22:10 Tiny Box
1:22:11 which is going to be a fifteen thousand
1:22:14 dollar machine learning
1:22:17 um computer that's got like six gpus in
1:22:20 it and like just massive throughput and
1:22:22 processing power and I don't know some
1:22:25 stupid amount of gigabytes of you know
1:22:28 GPU RAM and [ __ ] like that
1:22:31 um and that'll be able to run big models
1:22:34 but if you've got a single one you can
1:22:35 like go to go to Reddit and Twitter and
1:22:40 just start geeking out on open source
1:22:42 llms
1:22:44 like the seven billion uh parameter
1:22:47 models and they've got a bunch of them
1:22:49 that have come out it's it's like every
1:22:50 three days they've got a new model I
1:22:52 don't keep up with them anymore but a
1:22:54 lot of those can be installed on local
1:22:56 machines like I know they've got some
1:22:58 models now you can install on M1 Max and
1:23:01 then M2 Max are apparently doing really
1:23:02 good with this stuff and then there's
1:23:05 other ones that you can install on PCS
1:23:08 um the the CEO of stability AI says he
1:23:12 thinks that by the end of next year so
1:23:15 by the end of 2024
1:23:17 will have GPT 4 level
1:23:22 large language models that you can
1:23:24 install on your phone that don't need an
1:23:26 internet connection so these these
1:23:29 things are like very powerful tools
1:23:31 you're going to be able to install
1:23:32 locally so yeah if you've got a GPU you
1:23:34 can do that and you can start playing
1:23:35 around with embeddings and all sorts of
1:23:38 crazy [ __ ] if you want to geek out
1:23:39 I've been playing with code interpreter
1:23:41 today for my restaurant and it's
1:23:42 brilliant oh Silver Fox tell me tell me
1:23:44 more like what what um what are some of
1:23:47 the things you've done that sounds
1:23:48 really cool
1:23:50 um
1:23:51 then there'll be those of us still
1:23:53 secretly using it yes well that's the
1:23:56 that that goes back to that Ethan Malik
1:23:58 thing the secret cyborgs I think there's
1:24:00 gonna be a lot of people that use these
1:24:01 tools certainly short term for like the
1:24:03 next two years
1:24:05 where
1:24:06 you know your friends just seem to get a
1:24:08 little bit more productive the big
1:24:11 secret is are you using chat GPT for all
1:24:14 this no how dare how dare are you
1:24:18 hahaha
1:24:20 uh can't legislate AI anymore than you
1:24:24 can encrypt with encryption I know
1:24:25 exactly it yeah yeah exactly like that's
1:24:29 what that that's where I come back to
1:24:31 like it it comes back to
1:24:35 it comes the more the more people that
1:24:39 get AI literate the more it's going to
1:24:41 help us
1:24:42 stabilize its direction
1:24:48 um I don't I if we wait for government
1:24:50 to do it they're just going to [ __ ] it
1:24:51 up and they're going to take too long
1:24:52 this stuff's moving too fast so I think
1:24:54 it's on us
1:24:58 get on your Steed strap on the sword
1:25:00 head to the forest
1:25:04 oh you're an Archer fine bring your bow
1:25:07 I like hammers you too Thor
1:25:13 you can bring your hammer
1:25:17 um
1:25:24 all right Silver Fox it looks like
1:25:26 everyone wants to know what's going on
1:25:27 with your restaurant how are you using
1:25:28 it for your restaurant we want to know
1:25:30 curious minds want to know
1:25:32 what's the best AI bot you suggest for
1:25:34 trading stocks the ones that I have on
1:25:37 screen for the entire
1:25:39 three hours a night and if you're asking
1:25:42 me is is open AI paying you to do this
1:25:45 no but they should be
1:25:46 [Laughter]
1:25:49 um go there like well okay two things if
1:25:53 you want to find a trading specific AI
1:25:56 tool there's probably a bunch of them at
1:25:58 futurepedia.io
1:26:00 almost all of them are probably pieces
1:26:03 of steaming fresh [ __ ]
1:26:06 that look pretty because they used
1:26:08 mid-journey to make the images
1:26:10 um
1:26:13 there are probably some there but
1:26:16 uh
1:26:19 chat EPT especially with code
1:26:21 interpreter now
1:26:23 you can upload piles of financial data
1:26:26 and have it help you with the trading
1:26:29 stuff teach you but like I don't know
1:26:30 where you are in your education if
1:26:32 you're clueless it can teach you if
1:26:35 you're really sophisticated it can help
1:26:36 you write Bots and could do all sorts of
1:26:38 crazy [ __ ] so
1:26:40 um for me I would be learning how to do
1:26:44 it myself at chat.openai.com rather than
1:26:47 paying someone 20 or 30 or 50 bucks a
1:26:49 month for their glorified prompt sitting
1:26:53 on top of chat GPT because that's all
1:26:55 those things are going to be with some
1:26:57 minor exceptions they're so let's say
1:26:59 there's 4 000 tools at Future pedia
1:27:02 there's going to be 400 that are
1:27:04 mind-blowing real companies with really
1:27:08 cool technology built on top of this
1:27:10 [ __ ] and or or have their own technology
1:27:12 and then there's going to be what's left
1:27:14 3 600 that are just steaming piles of
1:27:17 [ __ ] it's probably going to be 10 that's
1:27:20 probably being generous
1:27:22 we have to identify the dragons first
1:27:24 don't we yeah we do I mean well or we
1:27:27 just need to know that
1:27:28 um we keep hearing Rumblings in the
1:27:31 darkness of the woods and we are pretty
1:27:34 sure those are dragons we don't even
1:27:35 know if they're dragons or what they are
1:27:36 yet I I agree with that the guardrails
1:27:40 scare me information control well Beer
1:27:42 Hunter I know like
1:27:45 that that's
1:27:49 that's a good conspiracy theory you know
1:27:53 skeptical place to start
1:27:56 um
1:27:59 if you listen to Sam Altman talk he's
1:28:01 very thoughtful about it
1:28:04 um I would get thoughtful about this
1:28:05 stuff because
1:28:08 it's easy to demonize a single company
1:28:10 and say wow Sam Altman this and Sam
1:28:12 Altman that and Elon Musk this and Elon
1:28:14 Musk that and Mark Zuckerberg this and
1:28:16 Mark Zuckerberg that there's a whole
1:28:18 open source Community out there
1:28:20 building these models with no safety
1:28:23 guard rails on them
1:28:25 and
1:28:26 it wouldn't surprise me at all that
1:28:28 within the next year
1:28:30 one of those non-guard railed models
1:28:35 someone's going to use it to do some
1:28:37 awful [ __ ]
1:28:40 um so it's not
1:28:43 it's not like we don't we we need to be
1:28:46 talking about and thinking about safety
1:28:49 and alignment and and when we're talking
1:28:51 about alignment with these large
1:28:52 language models it's basically
1:28:56 as as these
1:28:58 tools get more capable
1:29:01 and they exceed
1:29:03 our intellect
1:29:05 are they aligned with human values
1:29:08 that's what when they talk about
1:29:10 alignment that's what they're talking
1:29:11 about
1:29:13 so the guard rails are really just these
1:29:16 big organizations that have the money to
1:29:18 do incredibly powerful tools you know
1:29:21 looking at
1:29:24 how do we keep these things in line with
1:29:26 human values because if they get too far
1:29:29 apart
1:29:30 that's when you get into the Sci-Fi
1:29:32 dystopian scenarios so it's fine to be
1:29:37 cynical that oh you know the man is
1:29:40 gonna is gonna suppress all of us
1:29:44 um
1:29:45 and
1:29:48 we need to be thinking about these
1:29:50 issues
1:29:51 three cents an hour I know I got the 3.8
1:29:55 in my head so I guess I'm okay at math
1:30:01 all right
1:30:02 [Music]
1:30:06 he's approaching the end again people
1:30:08 ask more questions you're asking that's
1:30:11 it if I get to the end it's over you
1:30:13 know that right so like like we got a
1:30:16 nice tight crew in here uh normally we
1:30:19 have a few more so I'm normally way
1:30:20 behind on the questions but I am getting
1:30:22 toward the end
1:30:30 I love that you guys are talking to each
1:30:32 other
1:30:37 [Music]
1:30:40 PC versus MAC for using chat GPT well
1:30:44 for Chachi BT it doesn't [ __ ] matter
1:30:45 because you're using Microsoft servers
1:30:48 you're using Microsoft Azure every time
1:30:50 you go to
1:30:51 um chatgpt you're not it doesn't you
1:30:54 could run Chachi BT you can run it on
1:30:56 your phone it runs great on your phone
1:30:57 because all of the computes happening
1:30:59 out there
1:31:01 if you're talking about if I'm going to
1:31:03 install a large language model locally
1:31:05 what's better a PC or a Mac I don't know
1:31:08 I think that the uh it looks like the M2
1:31:10 Max are they're getting some pretty
1:31:12 impressive performance out of it and if
1:31:15 you paid attention to the Vision Pro
1:31:16 announcement from Apple the R1 chip so
1:31:20 so you have two chips over your eyes and
1:31:22 those goggles one is an M2 Chip which is
1:31:24 an incredibly capable processor the
1:31:26 other one's called an R1 chip
1:31:28 and that's doing all the machine
1:31:30 learning processing from all the sensors
1:31:34 um so they now have their own silicon
1:31:36 that's just doing machine learning
1:31:38 processing
1:31:40 so I think long term apple is going to
1:31:42 be amazing but
1:31:44 if you want like a the thing George hotz
1:31:48 is building like a six GPU monster
1:31:50 machine you're going to want a PC
1:31:53 so it depends on what your goals are
1:31:55 like I think I think just sort of out of
1:31:57 the box Computing Apple's gonna be
1:31:59 amazing at this
1:32:01 [Applause]
1:32:02 um but PCS are obviously way more
1:32:04 scalable and way more flexible
1:32:10 um
1:32:13 I gave it my menus it broke them down by
1:32:15 ingredients oh that's cool I had to make
1:32:17 new recipes oh my God that's so cool
1:32:20 were they good
1:32:22 now to work it all out before someone
1:32:24 else does that's super cool wow
1:32:28 huh
1:32:32 it's just this that's the thing about
1:32:34 code interpreter it's like like the fact
1:32:36 that I can upload an image and it can
1:32:38 process data from the image
1:32:41 I this is something I don't know if you
1:32:42 know this
1:32:43 so so I went to go get data sources and
1:32:48 and sometimes you'll get like like I
1:32:50 downloaded all my LinkedIn data because
1:32:52 I thought it might be fun to like go
1:32:53 like process my LinkedIn data and and
1:32:56 see what I can do with it
1:32:58 so your LinkedIn data it doesn't come
1:33:00 down as a single spreadsheet it comes
1:33:02 down as a bunch of individual CSV files
1:33:06 that are all individual
1:33:08 and so I was trying to figure out well
1:33:09 how do I get them into a workbook and
1:33:11 can I do that automatically and it was
1:33:12 all just not fun
1:33:15 and then it turns out
1:33:17 um you can upload a zip file you can
1:33:20 upload a compressed file that has like a
1:33:23 hundred different csvs in it and it will
1:33:26 write the code to unzip the file it'll
1:33:28 read all of the csvs and then talk about
1:33:31 them in context so you can upload
1:33:33 multiple files to code interpreter by
1:33:36 putting it in a zip file
1:33:39 and like and and then Ethan Malik did a
1:33:41 thing where he uploaded
1:33:44 he uploaded a spreadsheet that had
1:33:49 that had prompts
1:33:51 in in one of the it had prompts in in
1:33:55 the first row
1:33:56 and then his command was
1:33:59 um
1:34:01 his command was something like take the
1:34:03 prompts in each of the cells in the
1:34:04 first row and write the second row and
1:34:08 then write the so so he basically
1:34:10 uploaded an empty Excel file
1:34:12 and code interpreter filled in the Excel
1:34:16 file based on his prompt and then gave
1:34:18 him a link to download it back so he's
1:34:21 he's doing
1:34:23 like I I
1:34:28 gloo what to do with
1:34:30 it seems like it's it's like a Swiss
1:34:32 army knife that you can do almost
1:34:36 anything now it's only in Python so it
1:34:39 there's I'm sure there's libraries that
1:34:41 it doesn't have I'm sure it's a little
1:34:43 Walled Garden right now
1:34:44 but just conceptually what it's doing
1:34:47 when they start expanding that out to
1:34:49 all the programming languages into all
1:34:52 the libraries and figure out where the
1:34:53 safety boundaries are within it
1:34:55 it's just insane
1:35:03 [Music]
1:35:07 don't even bother with stocks Aladdin
1:35:10 has that coverage there you go somebody
1:35:11 knows I don't know [ __ ] about investing
1:35:14 so there's I like like here's the other
1:35:16 thing about AI you know we're talking
1:35:19 about this AI stuff like it's only been
1:35:21 around for six months it's been around
1:35:22 for like three decades and and like the
1:35:25 stock trading
1:35:27 um AI stuff's been around for at least a
1:35:29 decade probably more than that because
1:35:32 those guys are on it you know stock
1:35:34 Traders are the ones that like run their
1:35:36 own fiber because if they can get
1:35:38 something three microseconds faster than
1:35:40 the other guy they can make money
1:35:42 um so they're on top of this [ __ ] but if
1:35:45 you're talking about sort of consumer
1:35:46 grade level tools
1:35:48 um if if Aladdin's a good one that's
1:35:50 great
1:35:51 my mom asked me to call her home insurer
1:35:54 chat GPT giving a list of questions for
1:35:56 the insurance agent oh this is this is
1:35:59 one of the things that I'm most excited
1:36:00 about is
1:36:02 um
1:36:03 the ability for all of us for the for
1:36:06 the every man the every woman to
1:36:09 um
1:36:10 [ __ ] with the machine right to be able
1:36:13 to call United Healthcare and say well
1:36:18 based on the policy here are your rules
1:36:21 and here are the things it says you
1:36:23 should cover Well normally people don't
1:36:25 read that part of the documentation well
1:36:28 I did well chat GPT did
1:36:31 but you know here's my prepared
1:36:33 statement
1:36:34 would you like to talk to my lawyer no
1:36:36 no well why don't we'll just settle you
1:36:38 know what that's covered now
1:36:40 we're gonna get to do that soon we're
1:36:43 gonna get to [ __ ] with people I like
1:36:44 that I got a mosquito bite right on my
1:36:47 elbow look at that little [ __ ]
1:36:51 what in Florida now they have malaria
1:36:53 did you hear this the [ __ ] is happening
1:36:56 it's like God damn it
1:37:06 how high the Moon
1:37:08 let's see
1:37:11 foreign
1:37:17 how high is the Moon
1:37:23 well I think it's obvious
1:37:26 384 400 kilometers or obviously 238 900
1:37:32 miles everybody knows that
1:37:35 the measurement is taken from the center
1:37:36 of the earth to the moon I'm going to
1:37:38 ask
1:37:39 um does it
1:37:42 change
1:37:44 um depending
1:37:47 on the orbit
1:37:50 um
1:37:51 cycle or is it perfect
1:37:57 perfectly circular
1:38:02 um
1:38:03 it's not perfectly circular ha
1:38:09 so at perigee
1:38:11 which I assume is the narrowest
1:38:16 yes
1:38:18 225 000 miles
1:38:20 at the apogee
1:38:22 251. there you go see
1:38:26 there's chat CPT for you making us all a
1:38:29 good scientist
1:38:30 I can mansplain even better now
1:38:33 [Laughter]
1:38:36 it seems like he knows a lot
1:38:39 what's that earbud he's not listening to
1:38:42 anything with that is he
1:38:47 that's gonna be there's a good
1:38:48 application just have a little earbud
1:38:50 that gives you dating advice you're on
1:38:51 like your first date and it's telling
1:38:53 you what to say so you don't look like a
1:38:54 schlub that would be pretty good
1:38:57 what's a good eye to summarize audio
1:38:59 files the the open AIS Whispers really
1:39:03 good
1:39:05 what major challenge do you think AI can
1:39:08 help solve in the next five years lots
1:39:10 and lots of them um
1:39:14 I think it's going to automate out thank
1:39:16 you very much for the hearty heart
1:39:18 um I think it's going to automate out a
1:39:20 lot of the low-level repetitive work
1:39:21 that is
1:39:24 um kind of Soul crushing for people
1:39:27 um I think the scientific advancements
1:39:30 are going to be insane I think medical
1:39:32 advancements you know advancements in
1:39:35 Cancer Treatments genetic diseases
1:39:37 neurological diseases brain injuries
1:39:39 spinal cord injuries
1:39:42 um
1:39:43 I think I think the advancements are
1:39:45 going to be insane um Sam Walton talks
1:39:48 about
1:39:48 um
1:39:49 I should find this I should find this uh
1:39:52 this link and put it somewhere but
1:39:56 it was an interview and he says he he
1:39:57 thinks within the next five years or I
1:40:00 don't know if he said five years maybe
1:40:01 it was ten years
1:40:02 in the next short amount of time we're
1:40:05 probably gonna have a single year where
1:40:08 we have as much scientific advancement
1:40:10 as the previous 500.
1:40:12 and if you want to
1:40:14 get a little mind [ __ ] go to chat GPT
1:40:16 right now
1:40:17 and search for what were the major
1:40:19 scientific
1:40:22 advancements over the past 500 years and
1:40:25 so just look at all those look at the
1:40:26 list it gives you have it write you a
1:40:27 table
1:40:29 and then imagine a single year where the
1:40:32 equivalent the modern equivalent of that
1:40:34 much advancement happens in a single
1:40:36 year that's a very real possibility in
1:40:40 not much time from now so probably
1:40:42 within five years it's a lot there's
1:40:45 probably gonna be advancements in
1:40:47 climate science in energy
1:40:50 like like
1:40:52 um
1:40:53 like not needing
1:40:55 oil energy
1:40:58 um
1:40:59 yeah lots so
1:41:02 I don't know I've created a community
1:41:04 Kyle Amelia's wife is our mayor that's
1:41:07 awesome
1:41:07 [Laughter]
1:41:10 thanks for the answer you're welcome
1:41:13 Aladdin is back
1:41:20 it is like a little family in here it's
1:41:22 cool I like it
1:41:24 what's what's nice is you know someone
1:41:26 comes in and starts trolling me you guys
1:41:28 you know like pounce on them
1:41:29 appropriately if they have if they have
1:41:31 questions I already answered you let
1:41:32 them know if I went to go deal with the
1:41:34 dog you're like he'll be back he's
1:41:35 changing the shirt so we can read it's
1:41:38 good
1:41:39 I feel like you got my back it's nice
1:41:41 I was I was a bartender in New York
1:41:44 um New York City for like a year was it
1:41:47 yeah like a year and a half
1:41:49 and my first bartending gig was right
1:41:51 near One Police Plaza
1:41:54 um which is you know the the police
1:41:55 station in in Manhattan that all the cop
1:41:57 shows are at right
1:41:59 and so there were there were two bars
1:42:01 like One Police Plaza was here and then
1:42:04 there was this Irish Barco I think it
1:42:06 was called O'Neill's
1:42:08 and then I was at this like low rent
1:42:11 neighborhood bar called castillos
1:42:15 and so all the police officers went to
1:42:18 O'Neill's and like all the lower level
1:42:20 cops and administrators came to my bar
1:42:24 and so one night I'm bartending and and
1:42:27 they were they were
1:42:29 wild people they were wild like this one
1:42:33 guy was this really short guy I forget
1:42:34 his name
1:42:35 but he was uh
1:42:37 he was like yeah yeah I became a cop so
1:42:40 I could kick people's ass
1:42:42 I just like kicking people's ass that's
1:42:44 why I became a cop so so anyway so one
1:42:47 night I'm bartending in like these two
1:42:48 like college punks come in and they
1:42:51 start [ __ ] with me
1:42:52 and the the little short guy the one
1:42:55 that said he liked to kick people's ass
1:42:57 he was like he was like five four maybe
1:42:59 and skinny
1:43:01 well a Puerto Rican guy and and he goes
1:43:05 he goes hey
1:43:06 um hey can I ask you something come
1:43:09 you know let's can I ask you something
1:43:11 outside and so they they went outside
1:43:13 and they were gone for like 15 minutes
1:43:15 and he came back it was a little sweaty
1:43:17 he's like yeah they're not gonna [ __ ]
1:43:19 with you again
1:43:21 he drank for free for the for that week
1:43:23 so it was pretty good it's good when you
1:43:26 got you got your uh
1:43:28 you got your uh low-level cops to uh to
1:43:32 have your back when some jerks come in
1:43:36 um
1:43:37 I gave it my menus oh oh it's already
1:43:40 here what's going on Tick Tock why do
1:43:43 you move me around so
1:43:47 there's the audio file
1:43:50 who's that
1:43:53 Aladdin
1:43:54 Aladdin is Blackrock yeah I know
1:43:58 no no one's going to outperform Aladdin
1:44:00 well I mean
1:44:03 I know like like the early days of the
1:44:06 World Wide Web
1:44:07 there were there was a lot of Shangri-La
1:44:09 kind of possibility as well
1:44:12 and all that sort of Consolidated like
1:44:14 all the open source stuff and all the
1:44:15 really cool design stuff that all sort
1:44:17 of went away and it all did consolidate
1:44:20 so I assume the same thing's gonna
1:44:22 happen with this AI but
1:44:24 I think if the big companies try to
1:44:26 consolidate it too much the the the open
1:44:29 source Community I I feel like this is
1:44:32 I feel like this is squeezing through
1:44:35 the fingers of the control mechanisms
1:44:37 but we'll see we'll see
1:44:41 uh
1:44:44 uh now I'm gonna run her policy through
1:44:45 chat EBT thanks for that oh cool Oh
1:44:48 that's oh yeah oh that's great yeah
1:44:49 exactly yeah yeah take everyone's
1:44:52 complicated Health Care policies and you
1:44:54 know home and auto insurance policies
1:44:57 and yeah have code interpreter you know
1:45:01 make you charts and graphs about you
1:45:03 know
1:45:04 how you could
1:45:06 what are the like you could probably ask
1:45:08 it things like what are the sneaky ways
1:45:10 that they make this process overly
1:45:13 complicated that we could simplify and
1:45:17 get [ __ ] paid
1:45:19 I bet that's in there that's actually
1:45:22 something that would be fun to play with
1:45:28 evil [ __ ] policy writing people and
1:45:31 lawyers that are like well
1:45:33 if rather than having them fill out one
1:45:36 form we had them fill out 10 forms at
1:45:38 six different times over the course of
1:45:40 three months
1:45:42 and then at each of those steps we try
1:45:44 to deny the claim
1:45:47 I would think that only Point
1:45:49 0.625 of the people would actually
1:45:51 pursue that
1:45:53 right you know there's people thinking
1:45:55 like that in those [ __ ] companies
1:45:57 so now we can think like that too we can
1:46:00 actually read their [ __ ] and and
1:46:02 decipher oh that's how the rule works
1:46:05 okay
1:46:06 let's find another rule that contradicts
1:46:08 that and then let's take them to court
1:46:10 and watch them settle before we get
1:46:12 there
1:46:13 that would be good
1:46:15 thing they're genetically engineered
1:46:17 mosquitoes that can kill off the spread
1:46:18 of malaria oh you're being serious yeah
1:46:20 and zika yeah that's good
1:46:23 they're doing they're doing all sorts of
1:46:24 remediation right now down there about I
1:46:28 assume it's just lots of chemicals
1:46:29 Monsanto is probably happy what if AI
1:46:32 determined our government on alive JFK
1:46:36 um
1:46:38 the conspiracy theorists would have more
1:46:40 evidence to say that our government on
1:46:43 the live JFK I don't know I feel like we
1:46:45 we're living in a in an era right now
1:46:48 where
1:46:49 um logic doesn't matter anymore truth
1:46:52 doesn't matter anymore like we're in
1:46:54 this post-truth
1:46:55 sort of world that's very very
1:46:57 frustrating so I don't [ __ ] know
1:47:01 I don't know what would be nice is if AI
1:47:03 could snap us the [ __ ] out of whatever
1:47:06 pipe dream we're in the middle of right
1:47:08 now we could just get back to being
1:47:09 human that would be good
1:47:11 whatever happened to IBM's Watson it
1:47:13 flamed out Watson Health was just a
1:47:15 disaster apparently they just relaunched
1:47:18 because
1:47:20 I
1:47:21 so it's just remarkable that like the
1:47:24 the amazingly quick adoption of chat GPT
1:47:28 just forced everyone out of the woodwork
1:47:30 so
1:47:31 IBM was really big they did their big
1:47:34 blue thing they beat Gary Kasper off at
1:47:36 chess that was sort of the first big
1:47:39 headline AI can beat humans
1:47:42 because they said no one will ever be
1:47:43 able to beat a grand master and they did
1:47:46 it
1:47:47 and then they had this Watson brand was
1:47:49 just [ __ ] brilliant and then they
1:47:51 said we're going to do Watson health and
1:47:52 we're going to cure cancer and we're
1:47:53 gonna do all this stuff and then it just
1:47:55 kind of flamed out and you know it was
1:47:58 it was machine learning that was not
1:48:00 this Transformer based stuff that we're
1:48:02 playing with today so it was much more
1:48:03 specialized and so I don't know why it
1:48:06 it didn't succeed but it didn't succeed
1:48:09 um and then about four months ago IBM
1:48:13 announced Watson GPT I think they called
1:48:16 it or or something like that they
1:48:18 basically said we got AI too
1:48:22 we we did the chest thing and we Watson
1:48:26 GPT and then I haven't heard [ __ ] from
1:48:28 them since but I did hear from uh a
1:48:31 buddy that that knows
1:48:33 um some folks at IBM they put together a
1:48:37 what do they call them
1:48:39 um a center of excellence around prompt
1:48:42 writing and their IBM they have 400 000
1:48:45 employees or 280 000 now it was 400 000
1:48:48 when I was working with them
1:48:50 um and
1:48:52 um apparently they've put a thousand
1:48:54 people on prompt writing
1:48:57 because they can
1:48:58 so I have a feeling we'll hear from IBM
1:49:01 but yeah it's it it sort of flamed out
1:49:05 well well ain't you a cyber Sore Eyes
1:49:08 thank you your show is something else I
1:49:10 gotta tell you well
1:49:12 I would agree it's something else
1:49:17 I hope it's good that was good I want
1:49:21 you all to be happy you know what I mean
1:49:23 it would be nice if you was happy I
1:49:26 don't want you to add a partner I'll
1:49:27 tell you what you know
1:49:29 uh
1:49:33 all right
1:49:38 whatever happened to IBM SWAT so we just
1:49:40 did well well same for Sore Eyes
1:49:43 I am now I am now I am now
1:49:46 you catch my drift sir
1:49:48 yeah that's your name and it's in your
1:49:50 thing
1:49:54 all right which bar did the defendants
1:49:56 go to
1:49:59 I think the defendants got the [ __ ] out
1:50:00 of lower Manhattan the defendants were
1:50:02 like I'm out of here they got they got
1:50:04 in there they got in there uh in the
1:50:07 subway and headed to the boroughs or to
1:50:09 Jersey
1:50:10 or to the Upper East Side
1:50:12 [Laughter]
1:50:15 um what's something you'd wished you had
1:50:17 chat GPT for 20 or 30 years ago ah
1:50:20 [ __ ] writing screen place I I went
1:50:22 through a two-year period where I wrote
1:50:23 seven feature-length screenplays with a
1:50:26 writing partner
1:50:28 and for me
1:50:30 like
1:50:31 the hardest part of writing is is the
1:50:35 low level [ __ ] like I love the
1:50:36 concepting
1:50:38 I love
1:50:41 outlining while it's still creative
1:50:45 and I like
1:50:47 I like when the screenplay is either at
1:50:51 zero to fifteen percent done
1:50:54 or 85 to 100 done because those are
1:50:59 really creative things
1:51:01 the part of the screenplay that is just
1:51:03 for me is just like grinding my face
1:51:06 against a brick wall having have having
1:51:09 some strong person grind my face against
1:51:11 the brick wall is that middle the grind
1:51:14 the the you just got to keep writing you
1:51:17 feel like it's a piece of [ __ ] you don't
1:51:19 know what's gonna have you you're not
1:51:20 you don't follow the outline anymore so
1:51:24 just
1:51:25 what's happening with me with chat gbt
1:51:28 right now is
1:51:31 the whole idea of creative blocks is a
1:51:34 thing of the past
1:51:36 like if I don't know where to start or I
1:51:38 don't know what to do
1:51:41 um I can just with chat EBT and get to
1:51:43 80 like that so I can just stay at a
1:51:45 much higher creative level so almost
1:51:47 all of my history of of all the writing
1:51:50 I've done over the years
1:51:52 um I just wish I had it for that I wish
1:51:53 it were just part of my life because
1:51:55 it's it's it's it's staggeringly
1:51:58 powerful to be able to have
1:52:00 a concept
1:52:02 that historically would not have been
1:52:05 brought into the world
1:52:06 just because it was it wasn't a good
1:52:09 enough idea or it wasn't
1:52:12 um it was just too much to deal with to
1:52:15 actually pull it out of my head or I
1:52:16 didn't have the energy or whatever now I
1:52:18 can just go I kind of have a brain fart
1:52:20 of an idea and I can go huh let me see
1:52:23 if I oh
1:52:26 oh there's a oh that's not quite
1:52:29 quite the right tone but you know what
1:52:30 would be the right tone
1:52:32 boom there's the right time oh that's
1:52:35 actually pretty good let me copy and
1:52:36 paste that and throw that over here so
1:52:38 like I'm sort of getting ideas out of my
1:52:39 head much more quickly and then there
1:52:42 I'm getting them to a higher level of
1:52:44 completion much much more quickly what
1:52:45 do you got to do you got to pee why are
1:52:47 you whining huh
1:52:49 you gotta go pee pee
1:52:51 no
1:52:52 meow
1:52:54 little fat dog
1:52:56 you little fat dog no
1:52:58 yes all right I'll be right back
1:54:52 foreign
1:54:59 still here 30 people these are amazing
1:55:02 all right
1:55:06 the dog is happier
1:55:11 all right let's get some hearts going
1:55:13 see if we can get some more people in
1:55:14 here bang on the hearts share the live
1:55:16 if you can
1:55:18 and I'll do another what time is it it's
1:55:20 10 o'clock I'll do another I don't know
1:55:22 15 minutes half hour or something like
1:55:23 that
1:55:24 I love watching the art evolve in front
1:55:26 of my eyes yeah it's amazing
1:55:29 uh what's the big break on your show
1:55:32 gonna happen
1:55:33 let's get this show on the road buddy
1:55:35 Hey where's oh where's my big break on
1:55:38 your show gonna happen I don't know man
1:55:40 let's get this show on the road
1:55:44 I just started a new job and trying to
1:55:47 compare the insurance choices is
1:55:51 are you so Emilio's wife if you started
1:55:54 a new job and you're trying to compare
1:55:56 Insurance choices
1:55:58 um
1:55:59 go play with code interpreter if you
1:56:02 haven't because I bet
1:56:08 I bet
1:56:13 you could grab you let's say let's say
1:56:16 you've got three three choices or maybe
1:56:18 there's ten you could grab the PDFs of
1:56:23 whatever the [ __ ] they're trying to sell
1:56:25 you
1:56:26 throw them in a folder zip that folder
1:56:29 and then upload that to code interpreter
1:56:32 and have it compare them for you
1:56:36 I bet it does it I bet within an hour
1:56:38 you could you could knock some stuff out
1:56:42 the silver lawyers are going to be the
1:56:44 reasons why it's stalled more I agree
1:56:46 with that with the big companies for
1:56:48 sure I mean that's why that's why browse
1:56:50 with Bing is no longer in chat gbt they
1:56:53 know what it means for them lawyer look
1:56:55 the lawyers are
1:56:57 they're pre-ordering their Yachts now
1:57:00 because so many lawsuits are going to be
1:57:03 flying in so many directions when you
1:57:05 have the level of disruption that's
1:57:07 coming
1:57:08 everyone's going to be suing everyone
1:57:10 you stole my business I mean elon's
1:57:12 doing it right now with Twitter right
1:57:13 with threads
1:57:15 um you know all the Publishers are now
1:57:18 saying to to open AI you trained this on
1:57:21 three billion documents so you owe us
1:57:23 all money okay this is just gonna the
1:57:27 lawyers are gonna make a [ __ ] fortune
1:57:29 on this stuff
1:57:31 um yeah
1:57:33 AI does not determine now ai is a
1:57:36 probability engine
1:57:38 it will determine it doesn't right now
1:58:26 thank you
1:58:29 what
1:58:32 okay
1:58:35 all right whatever
1:58:39 and I'm back
1:58:43 uh-oh lots of loves loves going on
1:58:48 a keyboard rephraser app I can talk like
1:58:51 Joe Pesci sure
1:58:53 Killarney Rose can I solve crimes like
1:58:56 who killed JonBenet there's enough
1:58:59 well what what AI can do is it can take
1:59:03 whatever evidence it could it can
1:59:05 analyze [ __ ] really well so if there's
1:59:10 evidence to be analyzed sure if there's
1:59:13 not it'll make it up I mean that's the
1:59:15 problem the hallucinations with this
1:59:18 [ __ ] are are a lot although I what what
1:59:22 it sounds like with code interpreter is
1:59:24 that a it can do math because it's
1:59:26 writing python if you watch how it does
1:59:28 what it does it's mind-blowing it's
1:59:29 writing python so it can actually do
1:59:31 math and apparently it hallucinates yet
1:59:34 less because it's analyzing a specific
1:59:37 set of data
1:59:40 all right all right you bunch of pansies
1:59:42 let's put an end to those damn creative
1:59:44 roadblocks exactly
1:59:47 Don Rickles
1:59:48 dog ain't crying
1:59:51 champ is Hefty not fat what did the
1:59:54 doctor call him she she goes the vet we
1:59:57 took him to the vet and she goes well
2:00:00 he's doing okay she goes he's a chunky
2:00:03 boy he's the best
2:00:07 when your vet says your dog is chunky
2:00:10 [Laughter]
2:00:16 he's cute though
2:00:19 all right
2:00:21 to show up
2:00:24 uh felonious the the thing with Don
2:00:27 Rickles and Frank Sinatra was hilarious
2:00:29 that's a great story
2:00:31 um do you think teachers should prohibit
2:00:33 the use of Chachi PT to write essays no
2:00:36 I don't
2:00:39 um go read
2:00:42 Ethan Malik's piece called
2:00:45 I'm gonna find it
2:00:48 uh the homework apocalypse
2:00:53 m-o-l-l-i-c-k
2:00:55 home
2:00:56 work apocalypse yeah the homework
2:00:59 apocalypse
2:01:01 go read
2:01:03 that
2:01:05 it basically says
2:01:07 the internet made cheating really easy
2:01:09 AI makes it it it's done
2:01:14 and so it's not that
2:01:21 I think the intent is good
2:01:24 with teachers saying
2:01:27 if chat EPT writes your essay you're
2:01:29 cheating and you shouldn't do that so
2:01:30 I'm gonna find
2:01:31 things that determine that and I'm gonna
2:01:34 block it and
2:01:35 the intent behind it is right they want
2:01:37 the kids to learn
2:01:39 but I think how kids learn
2:01:42 has to change it has to fundamentally
2:01:45 change
2:01:46 because education preach at GPT was
2:01:49 about giving kids enough information
2:01:52 so that they could memorize it
2:01:54 assimilate it learn it get the knowledge
2:01:59 you now have a tool
2:02:01 that can do the Tactical execution of
2:02:04 almost anything at some level at some
2:02:07 level of competence
2:02:10 um
2:02:12 so if you're moving into a world of
2:02:14 infinite
2:02:18 um
2:02:20 content creation and infinite sort of
2:02:23 tactical execution what's left
2:02:28 critical thinking is left
2:02:30 so
2:02:31 the teachers that I see that are doing
2:02:33 this right are bringing this [ __ ] into
2:02:36 the classroom they're saying let's have
2:02:38 chat GPT write the paper
2:02:40 and we're all going to look at the same
2:02:42 paper and what you're going to do
2:02:43 tonight is you're going to go home and
2:02:45 you're gonna
2:02:47 you're gonna read how Chachi Pita did it
2:02:50 and you're going to come up with ways
2:02:52 you would have done it differently and
2:02:53 and you're going to find out things that
2:02:55 did wrong and you're gonna
2:02:57 find out if it used this logical
2:03:00 argument or that logical argument or
2:03:03 some other logical argument
2:03:06 um
2:03:06 and we're gonna have discussions and and
2:03:09 and you know teach them to think
2:03:11 critically so so I think that becomes
2:03:14 more and more and more important have I
2:03:16 been to Casa Bonita since it's reopening
2:03:18 I have not and I need to get there
2:03:20 because apparently the food is better
2:03:22 the food was so bad
2:03:24 at the old one that I was told by six
2:03:27 people like it's not even
2:03:30 like the kind of bad that's funny it's
2:03:33 just like disgusting don't go there so
2:03:37 I've never been actually
2:03:39 my uh my co-founder my business has been
2:03:41 and she said the food she said it was
2:03:43 worse than dog food it was it was
2:03:45 horrific but apparently it's good now so
2:03:47 I want to get there
2:03:49 um I want to learn to code an app I
2:03:51 looked up API pricing and it's all about
2:03:53 tokens I don't understand so basically
2:03:55 so when you when you um make an API call
2:04:00 so so let's say you're doing an app and
2:04:02 you want your app you want someone to be
2:04:03 able to say oh what what was what what
2:04:06 was it uh
2:04:08 I think it was Silver Fox who's got the
2:04:10 restaurant who said she took her menus
2:04:12 and popped them into chat GPT and kind
2:04:16 of came up with new recipes right so
2:04:18 let's say you want to do an app like
2:04:20 that and you know you scan
2:04:23 you scan your current menu and it
2:04:26 automatically creates a list of
2:04:27 ingredients based on those things and it
2:04:30 gives you new things
2:04:32 well you're gonna need to do an API call
2:04:35 to
2:04:37 translate the photo of the menu into
2:04:40 something and that's going to cost you
2:04:42 two cents
2:04:44 and then
2:04:45 you're going to need to take the OCR
2:04:47 text from the menu and you're going to
2:04:50 send that to open AI so so let's say
2:04:52 there's
2:04:53 a thousand let's let's say there's 750
2:04:56 words on that menu
2:04:58 and so you you have a prompt that says
2:05:00 take this menu
2:05:01 there's the menu and you send that to
2:05:03 open Ai and then it sends you back a
2:05:05 list of ingredients let's say between
2:05:08 sending the menu plus your prompt and
2:05:10 getting the response that's 1500 tokens
2:05:14 so your API will cost
2:05:18 that the API call to do that will cost
2:05:20 whatever the tokens times a thousand is
2:05:23 so it's like two cents or six cents or
2:05:25 whichever model you're using
2:05:27 so if you did 1500 instead of two cents
2:05:30 it would be three cents right
2:05:32 um and so so it's basically just
2:05:35 quantity of [ __ ] sent and quantity of
2:05:37 [ __ ] sent back and then they they add
2:05:40 that all up and it's it's very very
2:05:43 reasonable if you're doing a small
2:05:45 amount of thing like if you've got an
2:05:46 app that only has a thousand users and
2:05:50 they're only doing you know on average
2:05:55 500 tokens a day it's not going to cost
2:05:57 you very much if you have a hundred
2:05:59 million users and they're doing you know
2:06:02 10 000 tokens a day your bill is going
2:06:04 to be big but it but it's it's quite
2:06:06 cheap if you're doing small prototyping
2:06:08 stuff so
2:06:11 like I like I've got I haven't been
2:06:14 playing with it a ton lately but I I I
2:06:17 haven't had a bill more than
2:06:20 like 10 or 10 or 12 dollars on any month
2:06:23 that I've been playing with the API
2:06:25 stuff I don't but I don't call it a lot
2:06:27 but it's it's pretty cheap
2:06:30 I know you're the guy to ask
2:06:34 sir can we collaborate where can I write
2:06:36 to you if you go to my
2:06:37 um profile my there's a link Tree in my
2:06:40 profile that's got my contact info it's
2:06:43 also got a
2:06:45 a survey for
2:06:49 this thing I'm putting together a
2:06:51 workshop so some folks on here have have
2:06:54 said hey why don't you put together a
2:06:55 workshop
2:06:59 so I made a graphic for a workshop that
2:07:03 counts right
2:07:07 um but if you go to linktree slash real
2:07:09 Kyle Shannon that's the link tree the
2:07:11 first link there is the
2:07:13 um survey to just there's only three
2:07:15 questions there I just want to find out
2:07:16 what what do people want to uh
2:07:20 what do they want to learn if I did a
2:07:21 workshop and then I'll I'll put
2:07:23 something together so anyway my contact
2:07:25 info is there I am now
2:07:28 [Laughter]
2:07:30 what am I doing oh yeah
2:07:35 um
2:07:40 I am now you can DM Kyle it's true
2:07:44 I actually look at that [ __ ] because you
2:07:46 know why not
2:07:49 can AI build me a YouTube channel it can
2:07:51 write you all the scripts
2:07:54 and then you can use other tools to
2:07:56 automate all that [ __ ] you can use
2:07:57 canvas canvas got their bulk create now
2:07:59 so you can bulk create and canvas have
2:08:02 chat EPT write all the scripts
2:08:05 yes
2:08:06 oh here's the thing
2:08:09 it's just we're we're moving into
2:08:13 an era
2:08:15 where everything's going to be automated
2:08:18 everything
2:08:25 and then and then what gets really
2:08:28 interesting I think is what becomes
2:08:31 valuable in in a in a world where you
2:08:34 can automate everything from creative to
2:08:36 music to writing to project management
2:08:39 to just everything
2:08:41 what becomes valuable and my gut is
2:08:44 telling me right now
2:08:46 authenticity
2:08:48 I think authenticity makes a comeback
2:08:51 live music like in-person things
2:08:55 um
2:08:56 real people telling real stories
2:08:59 may rise above the noise I think that
2:09:01 but you know it's going to take two to
2:09:04 three years for all that stuff to start
2:09:06 to happen but
2:09:07 that's where we're headed
2:09:10 you have to know your thing otherwise
2:09:13 it's almost useless for for instance
2:09:15 encoding yeah I agree that's what I say
2:09:18 about the about this AI [ __ ] is
2:09:21 how to get involved if you're new with
2:09:23 this stuff if you're trying to figure
2:09:24 out how do I how do I get my head around
2:09:26 this stuff
2:09:28 what
2:09:29 there's kind of three phases there's the
2:09:31 there's the dick around with it phase
2:09:35 right the the parlor tricks phase write
2:09:37 me a song write me a poem write me a
2:09:40 marketing strategy
2:09:42 he's [ __ ] around with it kind of
2:09:44 generally getting the capability then
2:09:46 they're sort of looking at well wait
2:09:49 what do I do what's the thing you're
2:09:51 good at and I think your point there is
2:09:53 really smart like if you're really good
2:09:55 at engineering and coding
2:09:57 you're going to be able to do things
2:09:59 with the engineering and coding this can
2:10:01 do that are at a whole different level
2:10:02 than what all the rest of us can do
2:10:06 um so start to understand how to use
2:10:08 these tools to do what you do like like
2:10:10 basically disrupt yourself figure out
2:10:13 how you can do what you do better
2:10:16 and then I think the next phase after
2:10:17 that is then start to look more broadly
2:10:20 and more horizontally and say well [ __ ]
2:10:24 it took me 30 years to get to this level
2:10:26 of being an engineer I don't have to
2:10:29 take another 30 years to get to the
2:10:31 level of being a good illustrator
2:10:33 because now I have mid-journey so if I
2:10:35 could do this level level of coding and
2:10:37 this level of image Creation with
2:10:39 mid-journing
2:10:40 I can maybe put together a project I
2:10:43 never would have even considered before
2:10:44 and then if you do that with five or six
2:10:47 different disciplines or ten
2:10:49 all of a sudden you start to become a
2:10:51 super being right so so these things
2:10:54 become Humanity amplifiers right
2:10:58 and and some people are going to be
2:11:00 really good at this the ones that are
2:11:01 good at critical thinking and thinking
2:11:03 horizontally are going to be [ __ ]
2:11:04 rock stars
2:11:09 makes unbelievable errors it does you've
2:11:12 got to be vigilant with
2:11:14 the errors although code interpreter I
2:11:17 think is going to help
2:11:20 thin that out
2:11:22 if you're not aware you're lost agree
2:11:27 um yes I can build you a YouTube channel
2:11:34 so how do I get an app created is there
2:11:37 a prompt for that well so it depends
2:11:39 what you mean by an app if you want an
2:11:42 app like that's going to show up on your
2:11:43 iPhone then you need to do stuff with
2:11:46 xcode and Swift and things like that and
2:11:48 and chat CPT can help you write all that
2:11:51 but you still need to get it in
2:11:52 Frameworks it can teach you all that
2:11:55 um if you don't feel like learning all
2:11:57 that [ __ ]
2:11:58 you know down at the at the Lego block
2:12:01 level
2:12:02 um use a no code tool like zapier
2:12:06 um which now has a a new feature called
2:12:08 interfaces where you can create a
2:12:10 website and Bots and things like that or
2:12:13 use there's another no code platform
2:12:15 called Bubble which you can do you can
2:12:18 create mobile mobile apps with bubble
2:12:20 and I think it handles all the all the
2:12:22 back end [ __ ] and submitting it to app
2:12:25 stores and Google Play
2:12:27 um
2:12:28 or go find
2:12:30 an engineer who knows how to do app
2:12:33 development and ideally find one that's
2:12:36 doing app development using these tools
2:12:39 where they can knock that [ __ ] out
2:12:40 really really quick so I wouldn't put
2:12:42 that all on yourself necessarily
2:12:44 you know find a partner or a you know
2:12:47 a young person that's hungry like
2:12:50 finding a young person that's hungry
2:12:51 right now is kind of hard because
2:12:53 they're all a bit cynical and they're
2:12:54 like I like working from home and I
2:12:56 don't really this whole thing where
2:12:58 you entrepreneurs with your expectation
2:13:01 of 40 full hours of work
2:13:05 so if you can find not that
2:13:08 but someone who's actually excited about
2:13:10 this stuff and thinking about an
2:13:11 entrepreneur entrepreneurially
2:13:14 um you can probably get that app done
2:13:15 very very quickly
2:13:23 but go look at bubble I mean bubble and
2:13:26 zapier
2:13:28 the no code tools are
2:13:32 they are more simple than coding but
2:13:34 they're still complicated so you you
2:13:36 need to know your level of sort of
2:13:39 geeky geekiness tolerance right if you
2:13:43 kind of like geeking out on stuff
2:13:45 you'll probably enjoy them and you'll be
2:13:47 success like it's not hard to be
2:13:50 successful with them you just need to be
2:13:52 a good Problem Solver because you still
2:13:54 have to do a lot of problem solving even
2:13:56 though they're no code yeah the problem
2:13:58 with no code is there's still a learning
2:14:00 curve that's exactly right like zapier's
2:14:02 really got all sorts of quirks and and
2:14:05 weird things and yeah that the other
2:14:07 point is they're quite limited that's
2:14:08 the the second problem with no code is
2:14:12 you'll you'll you'll figure out the no
2:14:14 code stuff you'll get to 90 of what you
2:14:17 want and you're like I just need this 10
2:14:19 that's the 10 that the no code tool
2:14:22 doesn't do it's like oh that yeah that
2:14:25 you're gonna actually have to write some
2:14:27 Code for ah [ __ ] really
2:14:30 that's totally my experience with
2:14:33 totally my experience
2:14:35 uh yeah zapier is more like a tool for
2:14:38 connecting other tools yeah it is zappy
2:14:40 so but I think bubble is is more of a
2:14:42 app development framework
2:14:45 um but anyway
2:14:47 um listen it is 10 18 here in Denver
2:14:49 Colorado looks like I've gotten to the
2:14:52 the bottom of the list here got to the
2:14:54 bottom of the barrel really nice hanging
2:14:56 out with you all tonight I appreciate
2:14:57 your time as always have a fantastic
2:15:00 evening
2:15:01 peace out
2:15:04 um I will oh Silver Fox there's also
2:15:07 remote Engineers on Fiverr yeah that
2:15:08 could that could be interesting
2:15:11 so peace out everybody thanks from
2:15:13 Bucharest awesome bottom of the barrel I
2:15:15 meant bottom of the question Barrel not
2:15:18 that you human beings are the bottom of
2:15:19 the barrel how dare you Emilio's wife
2:15:22 you know better than that
2:15:24 peace out everyone