aiLL 7/6/23 - The Generative AI Revolution is Here! ChatGPT, Stable Diffusion, and More!

Transcript

1. 0:00foreign
2. 0:22welcome to the AI learning lab I'm just going to get a news story
3. 0:48exciting news coming there's big things happening welcome everybody we'll let people get in here it's the AI learning lab my name
4. 0:58is Kyle Shannon let me pop up my little I generally start out with this looks like Ron's here's that you were on maybe I think so um
5. 1:14what was I gonna say oh the website everyone get in here share it far and wide people share far and wide we just have a handful of
6. 1:26people in here now we gotta we gotta get this in here hey yellow what's happening Victor Troy good evening Mr Decker was shaking
7. 1:38oh Silver Fox shared the live if you haven't shared the live you're you're sl you're losing ground it's very sad if
8. 1:45you haven't shared it yet many people are sharing this live [Laughter] welcome welcome welcome it is what day
9. 1:59is today Thursday night it's a hot Thursday night it's actually kind of cool here in Denver there's a chance of tennis ball sized
10. 2:09hail or nothing it's like 30 percent 30 percent that it's going to be catastrophic and then you know 70 that
11. 2:18it's gonna be just fine everything's gonna be just fine so I don't know what was it last time two nights ago I was on live I guess it
12. 2:25was July 4th and uh yeah the garbage cans washed into the intersection it's pretty pretty nasty uh welcome everybody
13. 2:33Silhouettes hope you're having a great day too appreciate it uh welcome my name is Kyle Shannon this is the AI learning
14. 2:39lab let me explain what all these URLs are on screen for I'm not associated with any of them um they're really just there more as a
15. 2:46public service announcement um the channel is about talking about AI sort of going on this Grand adventure
16. 2:54together of whatever the [ __ ] this generative AI thing is going to be I'm an entrepreneur of many years I've
17. 3:00started 12 or 15 companies in my day some of them got really big some of them were just abject failures that shouldn't
18. 3:05have been started in the first place my current company Story vines 11 years old and profitable and we do a lot of good work
19. 3:12in the healthcare and Pharma space so that's kind of fun and uh the past year I have been obsessed with this AI stuff and when
20. 3:21chat GPT came out I had an epiphany about it that was not dissimilar to an epiphany I had about the World Wide Web back in
21. 3:291994. where I thought holy [ __ ] the world just changed and no one knows it yet and that's that's a feeling about this stuff so what
22. 3:38what's here if you haven't played with chat GPT um what I will do most of the night is just tell you to go do that so check
23. 3:46that openai.com that's the official chat GPT website if you haven't played with it go there bing.com is also chat EBT if
24. 3:53you click on their chat button that's Microsoft Microsoft's version of it and it's gpt4 and it's free
25. 4:01um and you can generate images there and you can it's connected to the internet so it's it's fairly sophisticated not as
26. 4:10good in my opinion as the gpt4 up there but whatever it's free and then poe.com's got a whole bunch of different
27. 4:15things you can play with so uh go play with those prompts.chat is a a little document that'll explain to you what
28. 4:24prompting is and how to do it how to talk to these things and then it'll give you a giant long list of personas where
29. 4:30you can tell chat GPT to act like different personas like act like an accountant or a master scribe or a
30. 4:38nutritionist or whatever and and it gets better and better as you give it more context and then futurepedia.io is just
31. 4:45a big long list of uh AI tools and Kyle for president oh God that'd be a [ __ ] disaster you know my problem would be I tell the
32. 4:58truth and uh then things would be very bad oh yeah they took me in the back room here's what's going on with UFOs I know
33. 5:09all right anyway if you have questions about AI pop them down below the other thing I thought might be interesting is um
34. 5:16if you've got a problem that you're trying to solve and you're you know you're trying to figure out how do I do
35. 5:20that with AI just pop that down below as well see what I can see what I can do to um answer any of that stuff so welcome
36. 5:28welcome oh [ __ ] wrong shirt I'm gonna need to go change my shirt you can't you can't see a thing here
37. 5:35hope you're having a great day all right um all right looks like you have a popped any questions down there so feel
38. 5:42free to pop questions and I'm gonna go change my shirt so we can read the text
39. 6:02thank you [Music] all right
40. 6:16got a little Basquiat Cassius Clay nice all right shirt changing commences oh yeah I should have put on some music
41. 6:31[Music] um am I doing an AI demo this evening yes what would you like to see greetings earthlings hello Kevin
42. 6:43it's loud I like it yeah it was loud but it was not you couldn't read the text I couldn't read the text it was it was making my head hurt
43. 6:50and we don't want daddy's head hurting that would be bad what the hair is starting to misbehave again I got it I
44. 6:57got a haircut it was good for like three days four days now it's it's starting to head south again that's like a two-pay tussle laughs
45. 7:10how about teaching us something about prompts okay I can do that I'll tell you what might be good [ __ ] duck I get it
46. 7:23you switch those letters around and if I wasn't careful I'd say a bad word very clever of you [ __ ] duck huh I'm too smart for you
47. 7:35like a steel trap so are you doing an AI demo and teach us something about prompting so what you should do [ __ ] duck is pop down below
48. 7:46what's something that you're trying to solve so rather than me just randomly putting [ __ ] in there because I tend to
49. 7:51go to the same stuff just because I'm thinking fast on the fly so I go to I go to the same [ __ ] over and over again so
50. 7:57put something down there and then we'll see if we can uh I'll see if I can come up with a creative way to play with it oh there's some news
51. 8:06so if you pay for um open AIS API everybody has access to gpt4 now I didn't have access to gpt4 on the API side I do know everybody has it
52. 8:19and the other thing that you've been hearing me whine about for I don't know three weeks now four weeks is that I
53. 8:26don't have access to code interpreter for chatgpt apparently everyone who has a paid chat GPT member is getting access to
54. 8:35code interpreter in fact why don't we go check it now maybe they gave it to me tonight but everyone is supposed to get
55. 8:40access to it within the week beta features it's not there yet but apparently within a week we all got
56. 8:48access to code interpreter Ethan malaku I'm a big fan of said what does what did he say in his tweet hang on um let me see code interpreter
57. 9:03here we go mullik what he said he said where's this post
58. 9:18everyone who subscribes oh let me flip this around so Ethan Malik Wharton professor brilliant dude definitely worth following
59. 9:28everyone who subscribes to chat EPT plus is apparently going to get code interpreter in the next week it is the
60. 9:34mode of AI that I find the most promising interesting and unnerving he it's coming and if you don't know what
61. 9:45code interpreter it is you upload files to it and it reads them and understands them and then it can do [ __ ] for you so
62. 9:52like you can upload a big pile of data and then say to it uh what's interesting about this data and it will write you a
63. 9:58whole report and generate graphs and charts it'll write the python to generate it's insane it'll turn gifts into MP4s it'll pull
64. 10:10the colors out of a gif and create a color palette it I I don't know what it does because I don't have it yet
65. 10:17but when I have it I'll play with it all right let's see who's falling asleep [ __ ] up
66. 10:32how about teaching us something about prompts and then you're snoring on me now I'm not going to teach you something about prompts
67. 10:38that's just rude Kyle have you checked out threads got AI learning Labs set up over there um I set it up with uh
68. 10:50with my Kyle Shannon account but um no I mean I played with it I did one post um let me see can you do they have a
69. 11:04a web app like the Twitter threads uh meta -doo I guess not but if you go look up uh at Kyle Shannon
70. 11:19on threads my first post was a mid-journey image that I created today that was really cool you want to see it I'll see a cool image
71. 11:27flop ducks like what are you gonna teach me about prompting man that's just how he talks dude this is a little dark but that's the
72. 11:36image that I created Isn't that cool paper cut shapes and stuff like that that was my first post to threads I
73. 11:45figured I'd do a purdy image put a purdy image up there greetings I agree about the prompts wait given all that's going on in the world
74. 11:56wait earthlings is appropriate wait given all that is going on in the world earthlings is appropriate oh someone
75. 12:05said haircut looks haircut looks rocking that is well it does I guess it does it looks like bad what is that late 80s early 90s rocking
76. 12:16it's like it's like it's like I don't know Wham Wham with a receding hairline thank you thank you very much for that
77. 12:30Rose that's for the uh receding hairline yeah that's she's selling hair growth product right yeah she's she knows what's going down up here
78. 12:4020 bucks a month for um Poe yes is poe legitimate I would just cancel my open eye AI subscription okay so here's the deal
79. 12:53well the middle one's free so you can't pay for that I pay for open Ai and I pay for Poe I will cancel po as soon as open AI
80. 13:04expands their context window here's why I think they're both worth paying for and and the announcement about code interpreter being launched
81. 13:12within a week makes me tip heavily toward um chat gbt so the top one um so the top one if you pay for chat GPT plus you get gpt4
82. 13:27um you also get plugins when they when they resurrect um browse with Bing the the gpt4 connected to the internet they just
83. 13:40killed it three days ago because it was stealing um um content behind paywalls um I think I understand what's actually
84. 13:46what happened there but uh but they got to fix some [ __ ] so that'll be back so you'll have the default mode which is
85. 13:53gpt4 you'll you'll have uh it connected to the internet you'll have plugins which are interesting but flawed right
86. 14:00now and then you're gonna have this new thing called code interpreter so paying for chat GPT plus gives you all of those
87. 14:07there's other [ __ ] coming and and you'll get access to it first but code interpreter seems like it's a game changer uh just
88. 14:16just for it just seems like a game changer like it's it's I haven't been excited about something in a while about
89. 14:23like I want the new thing to come that's what I'm excited about okay so that's that now let's go to Poe pow pow pow
90. 14:32and here's why I think Poe is potentially worth paying for but you know what you're trying to accomplish better than I do
91. 14:39come on focus focus focus all right well it's not going to focus okay so when you come to Poe you've got these six models here Sage
92. 14:48gpt4 Cloud plus Cloud instant 100K Cloud instant and chat GPT the second the the two three and four gpt4 Claude plus and Claude instant 100K
93. 15:01you have to be a subscriber for Claude instant 100K what that means is it's a hundred thousand token chat chat model so translate it into
94. 15:12like if real people were to say that Kyle what would they say you can put 75 000 words into your prompt
95. 15:20so that's like a novel you can if you've got the Great American novel written and you want to summarize it or you want to
96. 15:27reorder the chapters or you want to make it written in the I don't know the tone of some favorite author of yours
97. 15:36um you can just paste it in there and start interacting with it 75 000 word context window so that's a big deal
98. 15:43um Poe I like like the the it's worth using um for the free version just because of the Bots because I think the the little
99. 15:50body ecosystem is pretty cool I think the Bots are actually probably more useful than plugins are right now
100. 15:57um but that's the difference between them recommendation um I don't know I would lean toward if you could only afford one I would go
101. 16:04toward open AI because I think code interpreter is gonna I won't change the game don't change the game what is the game what game are we
102. 16:13playing Kyle why is your hair so funny nice code interpreter I know where do you get the news on open AI on
103. 16:22the Twitter and maybe now on the threads did you see did you see that that tweet that [ __ ] Zuckerberg sent to Elon
104. 16:31Musk Elon Musk sent him a tweet going hey looks like you stole my you know intellectual property for for Twitter he
105. 16:38goes he said something about um Zuckerberg cheating and then and then Zuckerberg goes well what about you
106. 16:46cheating with uh Sergey brin's wife it's like oh as muscle says the girls are fighting oh my God I think all of this I I here here's
107. 17:00here's the theory for you maybe Zuck released threads just to Hype up the the cage match that they're gonna go do
108. 17:08can mid-journey do tasteful Modern Art more minimalist yeah can absolutely do that um let's go do one oh you know what we'll do we'll go to
109. 17:20Pub so here we are in po and I'm going to click the mid Journey bot and then I'm going to say painting let's see painting
110. 17:34of colorful abstract abs direct shapes modern
111. 17:50minimalist we'll do 50s and then let's just do that I'm going to let this thing rewrite it so what it re rewrote was
112. 18:00colorful paint color painting of colorful abstract shapes in a modern minimalist style from the 1950s bold and
113. 18:08vibrant colors geometric and symmetrical composition clean lines sharp edges textured surface acrylic on canvas Jackson Pollock mark rothko
114. 18:20well okay that's just [ __ ] random you can't have Pollock and Mondrian in the same prompt so so we'll unfuck that up
115. 18:30that's you know one of the cool things about playing with these these tools is like like with mid-journey and and
116. 18:36understanding what it can do and what it can understand is like you do get a bit of an art history lesson
117. 18:43that was me trying to do oh that's up with this Focus ain't there we go um this is me trying to do Sling Blade
118. 18:53it didn't quite get it it's in the neighborhood but it wasn't quite right all right image all right so we're gonna do that we're
119. 19:01gonna pull out Jackson Pollock because that's not minimalist at all that is maximalist so we'll see what this does oh and it
120. 19:10gave me a two by three aspect ratio which I don't like but we'll play with it we'll play with a couple of cool things here there's a new
121. 19:17thing called um zoom out which if you haven't played with it is is their word for out painting which Zoom outs much more clear
122. 19:26and then there's also panning you can now pan you can so if you've got an image here and you want to like extend
123. 19:32just one side of it you can pan it in that direction and it actually makes the image bigger all right so that's going off and doing
124. 19:41its thing looks like mid Journey's being slow so we'll come back to that let's see awesome awesome
125. 19:59where's the picture it's coming where's the picture it's just the most remarkable piece of technology in the
126. 20:07history of humanity why isn't it done yet my microwaves faster than this with a hot dog here they come
127. 20:18here they come they're coming is it done looks like it's done [Music] I don't like this let's see let's get let's take Mondrian out of
128. 20:34that let's do another one slash imagine pop that in get rid of Jackson Pollock let's just do Mark rothko get rid of mondrion
129. 20:45and then instead of bold and vibrant Colors Let's do um bold dramatic colors um symmetrical composition clean line
130. 20:58sharp edges textured surface acrylic on canvas and let's do 16x9 all right so we'll go do that and then
131. 21:08I'll come back and I'll show you some of the other stuff I just don't like those color palettes they just look I don't know what is this
132. 21:14what are we what are we Easter what is this Easter Eastern Modern Art no no we want some devil worship Modern Art
133. 21:22[ __ ] some drama we want drama all right I lost my place in a little comment here okay here we are modern
134. 21:32tasteful art I I don't know what code interpreter means really okay I'll I'll come back to code interpreter uh I'll go
135. 21:39show some some things but base the basic deal with it is in your little chat box in chat it gives you an upload button and then you can
136. 21:48upload like say so you're in fundraising you could upload um you know all of the uh all of the foundations or non-profits performance
137. 21:58for the year let's say you got a big spreadsheet with all the performance data of all your non-profits and you
138. 22:03could say tell me what's interesting about this data tell me the trends tell me what's going on here and it will just
139. 22:10read all of the data and it'll write you charts it'll analyze the data it you can tell it to visualize stuff it'll write code for you it's
140. 22:22it's crazy I'll go find a post on Twitter these are actually pretty good it's the they're they're not quite the
141. 22:28color palette I would go for but they're this prompt right colorful painting abstract with Martha Mark rothko and then
142. 22:35and then um it um there's four different images here with with these abstracts it's hard to tell there's four different ones
143. 22:44um so you got one two three four and then you've got upscale one through four variations one through four and you
144. 22:50could just re-roll it basically just say I don't like any of them give me a bunch of new ones but let's see I think
145. 22:56I think one and three are interesting to me so let's do variations of one and variations of three and so now that's going to have
146. 23:06um uh mid Journey go do things that will look very very similar but they'll have slight variations and then we can blow
147. 23:12one of those up and then we could let's say we wanted this thing to be square instead of landscape we could do that as well
148. 23:21what's going on here
149. 23:40oh there we go that's that's looking interesting I like this one with the big dark circle they're not quite there yet this one's done
150. 23:53wait oh no they're still coming they're still coming 93 percent so you kind of watch these things if you
151. 24:01haven't watched these things work what they actually start with is like a fully noised image and then over a series of
152. 24:0730 or 50 or 70 steps it'll denoise it so I like three and I think two is pretty interesting so what I'm going to do is
153. 24:16I'm going to hit upscale two and upscale three and so now it's going to give me big versions of those
154. 24:30here by printing some website and then printed um are you kidding holy [ __ ] that was two recipes yeah I
155. 24:47know I'm so sorry it's all right all right so so there's one of them that's actually pretty beautiful
156. 24:57and then oh it's great and then there's another one all right so that's oh I was going to show you the zoom thing so so let's say um
157. 25:09let's say we we like this uh this dark circle here I like this dark circle here and let's say I want that more in the middle of
158. 25:18the composition so I can now say pan down so if I hit this down arrow what it's going to do is it's actually going to
159. 25:25add to the bottom of the painting so it's going to give us a whole new chunk of painting um so this is that's something that just
160. 25:34came out in mid-journey like yesterday I think or the day before so that's pretty cool so we'll wait for
161. 25:41that to do that people are going to flip when they get code interpreter I've had it for a while now oh the silver tell me
162. 25:46more tell me more talk about what it is yeah I I just like everything I've seen in fact Let Me Go let me go to the old Twitter let's see
163. 26:01so here here's the thing Ethan Malik did he he asked code interpreter to give him like different kinds of how did he describe it hang on
164. 26:14he goes I had no idea chat EBT with code interpreter could do flow charts until it did so I asked it what other exotic python visualizations
165. 26:25it could do and to make samples interactive maps pairwise plots word clouds hierarchies machine learning
166. 26:33diagrams Andrews curves so like it can just do this [ __ ] like you upload you upload data to this thing
167. 26:45and then you say like visualize this data on a map give me you know a set of Scatter Plots give me whatever the [ __ ]
168. 26:53these things are you know give me a word cloud and it writes the python code and Maps the data into it and provides you the graphic
169. 27:02I mean good flipping Lord really that's coming this week can you explain the math of how neural networks calculate gradient descent I
170. 27:15cannot I can tell you how large language models do what they do and I can tell you how things like mid-journey do what they do
171. 27:23but what you're asking for is a level more machine learning engineer than I am unfortunately so here's the uh
172. 27:33here's the extended image let me I'll just upscale one of these let's grab I think this bottom one yeah is better so we'll do upscale three
173. 27:46um so I don't understand the math of that just just a little bit about context about who I am I'm an
174. 27:51entrepreneur and a you know I'm a I'm an innovator with technology once it sort of hits um the the masses right so so when the
175. 28:01internet was just the internet you had to do command line [ __ ] I used it but I didn't use it a lot when the World Wide
176. 28:07Web came out I was like holy [ __ ] I started one of the first online magazines I started one of the first online agencies
177. 28:14and same thing with chat gbt when when this thing came out I'm like holy [ __ ] everything's different so I'm I
178. 28:21understand some of the math behind this stuff and how these things work but but think you know from chat gbt forward essentially
179. 28:29um so anyway so there's that thing that you know the this painting ended down here before and now it just added all
180. 28:35that stuff at the bottom I mean remarkable if you ask me you know you know just your everyday uh every day little making art on nothing
181. 28:52I haven't messed with the others yet but I have spent some time finding the edges of gpt's capabilities that's cool keep
182. 28:58digging on it push it harder um someone asked me about prompting before I'll do something here as soon as
183. 29:04I get a little inspiration about what the prompt I will do it retro bug the context window is my biggest challenge so if the contacts
184. 29:12Windows your biggest challenge it's probably worth paying for Poe and going to go go play with that hundred thousand
185. 29:19token window or if you don't want to pay for it there's a open source model called MPT 7B that has a 65 000 token window which
186. 29:29is what 45 000 words 40 000 words something like that 45 probably um if you're geeky and want to go play with
187. 29:38uh open source there's that or pay the 20 bucks and go play with the Claude and hundred thousand one where did the pick go which pick
188. 29:48um I made it go away code interpreter gets around the token length issue in some ways that's interesting where's the picture
189. 29:57there's the picture no it's gone again uh um when do you think Bing is coming back to chat GPT I have no flipping idea I don't
190. 30:09know how bad that bug was so so here's here's what I think okay so when you go to um when you go to the New York Times or
191. 30:19when you go to whatever site and it shows you the article and then it like pops up the thing that says you got to pay you flipping loser
192. 30:27there's kind of two ways to do that paywall one of them is kind of the cheap way where if you know how to inspect the
193. 30:34code of a page you can actually just sort of sneak you basically just say show me the code of this page and you can see the article so so
194. 30:44those kind of sites that kind of had the cheap version of the paywall for lack of a better term I think chat GPT was just
195. 30:51reading the code and presenting the paywalled content I I assume the ones that had better versions of that better
196. 31:00versions of the payroll while that didn't happen but but you know it could have been a major player that just said
197. 31:05if you don't fix that we're gonna you know make this a big deal so I I mean I mean listen Sam Altman and the team at open AI
198. 31:13they've got Microsoft as a partner and even though Microsoft doesn't have a controlling interest they don't want to
199. 31:20piss them off right Microsoft has a bunch of advertisers on Bing they've got massive corporate Partners so it
200. 31:27wouldn't surprise me if you know someone made a stink and maybe Microsoft has this dealt with on on their
201. 31:36implementation of chat jpt or on Bing chat um and maybe they said to open AI you got to deal with this or you know maybe
202. 31:44they're just doing it I don't know but I don't I don't have a [ __ ] clue it seems to me that it would be something
203. 31:49relatively simple to fix but if they don't know what the actual problem is if they don't know all the places where
204. 31:56it's sneaking around paywalls and and whatever other issues were raised um it might be a while I mean the good news is we've got Bing
205. 32:05good and it's not as convenient as using gpt4 connected to the internet there but that's the deal it's where we are yeah
206. 32:15I don't know why it's like this I want it to be so much better all of this has promise but it seems so you know
207. 32:28hard forking last week in AI or great podcasts I'm confused by the pricing is it charged by tokens or keys I don't know okay so
208. 32:39chat GPT is flat pricing charged monthly if you're paying for apis um the the GPT API on the playground
209. 32:51and and you're doing things like zapier and you're hooking into the API that's charged by the token so the the key just gives you access to
210. 33:01make the API calls and then the API calls are the combination of the prompt that you send in and the answer that it returns that
211. 33:08amount of tokens is what you get charged for each call I don't know if there's a I don't think there's a charge per call
212. 33:14I think it's per chunk of tokens per thousand tokens I think so like like uh GPT 3.5 turbo is I don't I don't even
213. 33:25know what it is two cents and then GPT for six cents like they they've they've changed like the GPT 3.5 turbo 16k is
214. 33:34way cheaper and then gpt4 was slightly cheaper so it is confusing though because you're paying if you're paying for chat GPT
215. 33:44you're essentially paying for an application layer if you're playing if you're paying for API access then that's
216. 33:50then you're paying for tokens you can put text files into code interpreter and greatly expand the context in a prompt oh that's a super
217. 34:00cool idea yeah because I think I think can't isn't it like a maximum of 100 megabyte file you can put in 100 megabyte data file
218. 34:10so yeah so you could put in 100 megabyte text file which that's probably a fairly sizable amount of text and then have that be the
219. 34:20the prompt holy [ __ ] that's amazing because you could do some some really sophisticated prompting in that
220. 34:29in the uh in that document wow that's wild is that 75 000 K of just your words or added together questions and
221. 34:37responses yes it's a 70 it's a hundred thousand token window context window so that's that's a combination of so you could put in a 45
222. 34:48000 word novel and get back a thirty thousand word response and then you'd still be under that and then it would
223. 34:55cut you off somewhere in that neighborhood hey I have a serious question well then I'll give you a serious answer please ask all right
224. 35:10do you think software developers careers are at risk due to AI well you said due to IA definitely not at risk due to I IA
225. 35:19but maybe due to AI I know I'm a little smarmy he's a little smart he's kind of a smart ass isn't he is he trolling me I
226. 35:27don't think you're allowed to do that yes you [ __ ] yes that's gone all right do you think um so so here's
227. 35:41the deal software developers jobs are so okay so there's a guy that I'm a big fan of named Paul rateser who started the uh
228. 35:51marketing AI Institute and he's got this thing he says ai's not going to take your job but something somebody using AI will
229. 36:00um I think that AI literate programmers AI literate Engineers are going to be fine for some time because there's a lot
230. 36:07of [ __ ] to be built now how engineers and programmers do what they do I think is going to transform quickly and
231. 36:15dramatically and and you know anyone who's using GitHub with copilot right now you know essentially doing pair programming with GPT tools
232. 36:24um knows this already right it's just a completely different experience you're you're sort of driving the intent of the
233. 36:30application and then GPT sitting there sort of executing you know big chunks of code and then you're sort of figuring
234. 36:37out is that code right is it did it do it in a way that I like right all that sort of stuff so so I think that there's going to be
235. 36:45plenty of work there but any developer right now that basically just sticks their head in the sand and just says
236. 36:52well hand coding is the is the way I you know I've always done it this way and I will be coding like this for all of my
237. 36:59days they're going to be in trouble because some little Punk's going to come along and be able to 10x
238. 37:05um the productivity and I don't care how tight their code is at some point they're just going to get priced out of
239. 37:11the market right they're gonna they're gonna be expected to do more than they can do with manual coding all right
240. 37:16serious question from Black Swan I have an idea I just don't know where to start well feel free to share what you are you feel free to share
241. 37:24um and I'll help or I mean I don't know if I'll help I will give you feedback on whatever that might be and and tell you where you
242. 37:32might start um yes they absolutely are we'll need less developers until New Industries expand yeah in the end I think AI is going to
243. 37:42create you know massive amounts of new jobs in the next two to three years it's going to be a really painful transition
244. 37:49until those new jobs emerge right right now all of these generative AI tools all the stuff that's going to [ __ ] up all
245. 37:57the jobs is very early and it's very kind of spit and duct tape it's it's really janky software it's not
246. 38:04sophisticated the autonomous agents that everyone got so excited about you know Auto GPT and baby AGI all those things
247. 38:11are just janky pieces of [ __ ] right now all of it is as these tools get more and more and more sophisticated and
248. 38:22instead of being disparate features and functions like we've got mid Journey over here that you have to do in Discord
249. 38:28but we've got chat GPT over here but we got this over here we got that over there all this stuff's going to start to
250. 38:34merge where the tools that we use are just going to be imbued with AI capabilities um and and and and that's when we'll
251. 38:44start to things will start to normalize a bit but that's going to take a while [Laughter]
252. 39:00I could listen to you all day sorry about that I wish you were my dad careful what you wish for yes [Laughter]
253. 39:17it wears thin I wear thin thin I wear this I'm a lot I'm a lot Camelot
254. 39:31all right I get Tick Tock just lost me in my place of comments again so here's the thing for those of you that have
255. 39:38been here before you know this about me I got a little OCD like I want to answer all the questions and then as I'm
256. 39:44scrolling to answer one question someone will post another one and Tick Tock jumps me to the bottom of the screen and
257. 39:49then I lose everything and so that makes me cranky and so then I just sit here scrolling old man scrolling on a tick tock old man
258. 39:59tried to read words old man eyes don't work old old man welcome to Elder talk children ah oh here's black Swan's serious
259. 40:15question all right I'm back Black Swan asked I want to take every piece of Congressional Record and feed it to an AI Okay so
260. 40:29Okay so good serious question I should I should answer it seriously and you know for talking about Congressional records um okay
261. 40:45so so there's a couple of things about what you want to do I think what you want to do will absolutely be done I
262. 40:51don't know what your background is what access you have to that if you're just talking about doing public records um
263. 40:58I think there are so many applications for that kind of work I was I was on a panel um about a month and a half ago uh with
264. 41:05a guy NX newspaper guy that was talking about local journalism and the fact that local journalism is [ __ ] and doesn't
265. 41:12exist and like all these you know uh multinational companies are buying up all the local markets and right so so
266. 41:18local journalism is kind of screwed but what he was talking about was what local journalists most of what local
267. 41:24journalists did was they went and got you know public records and they reported on them you know here's what is happening on that corner
268. 41:33lot they're putting up a skyscraper and if you want to keep the little corner candy store you better get out there and
269. 41:39protest right that's what local journalism was well that's all discoverable public data that's electronically discoverable that you could
270. 41:50create um you know GPT article writers with um automatically and so what you're talking about is the same thing right now it's it's it's
271. 42:01um in order to do that you can't just do it with chat GPT you can do your prompt engineering on GPT playground or on
272. 42:08chatgpt but ultimately what you'd need to do with that there's a there's a framework a project called Lang chain I
273. 42:15think it's short for language chain and Lang chain allows you to essentially Point um to certain directories full of
274. 42:24content and embed them and so embedding basically means the the English uh the the real human being phrase of embedding
275. 42:33is make it gptable so you basically take all of the Congressional records you embed them make them gptable and then
276. 42:40you store them in what's called a vector database and that just basically means they're searchable like chat GPT so it's absolutely doable
277. 42:48um and depending on the access you have um and you know you could probably find you know an engineer or two that's got
278. 42:55some experience with Lang chain and do that I would go just go hang out on Reddit um in the in the Lang chain subreddit
279. 43:03and there might even be some other projects or extensions of Lang chain um that you know that might make that relatively straightforward
280. 43:12you can do some pretty sophisticated stuff with not a lot of code and not a lot of time it all boils down to how
281. 43:17fast people want to adopt on the scale how do you display text on mid-journey you don't right now the text the text is just it's just not
282. 43:26there yet they'll get there with it if you want to do text you can use adobe's Adobe Firefly has a uh you can sort of do
283. 43:34um prompting inside of a font so you can make like fuzzy letters right the word fuzzy all the letters are fuzzy or you
284. 43:43know honeycomb and they're all Honeycombs so you can do that with an adobe Firefly there I've I've seen some
285. 43:49papers in the past month that that are talking about their methods for how to do text my guess is that we're gonna see
286. 43:56coherent text on stable diffusion before we see it on mid-journey um just because it's open source in the
287. 44:03developer Community kind of attacks everything on stable diffusion faster than the team at mid-journey can um so
288. 44:11everybody like like I feel like mid-journey dealt with the pretzel fingers and so those are good or getting better now
289. 44:19um and and they've they've dealt with some of the mushy faces and uh so I would say text is probably it's
290. 44:27probably up there in priority whoever gets text first is gonna people are gonna flock to that whatever that tool is
291. 44:33do I know if AI is being included for the future of Robotics yeah there's let me see if I can go find this one video
292. 44:39there's a there's a disturbingly fun video from Boston Dynamics which I want to like them but their robots are
293. 44:47[ __ ] terrified like why like if you're gonna have a robot company why would you make your robots look like the
294. 44:53ones that are in all the movies that kill us you know you know what I'm saying Am I Wrong Am I Wrong see Boston Dynamics dog chat GPT
295. 45:09oh doc dog there it is we use certificate to query information and the user can ask natural questions I
296. 45:25was just wondering if I could dictate to it to say something specifically what is your battery level battery level is
297. 45:32currently at 53 what is the voice coming from it's a Google Text-to-Speech we give the Json to try GPT and explain
298. 45:40what the structure is and how to read that Json Notch rgbt can answer questions about that case all right so
299. 45:47so what he said there if you if you're not into the geeky stuff that that basically they're they're turning the
300. 45:53chat GPT into Json which the robot can understand and then so they can send stuff to the robot that the robot
301. 45:59understands and then the robot can respond and chatgpt can can um take that in so they're doing that there what what one of the things that
302. 46:08large language models make possible is um is is putting an incredibly simple layer on top of incredibly complicated systems
303. 46:17so a robot is an incredibly complicated system of sensors and motors and batteries and you know all that sort of stuff um and and so
304. 46:29historically you know what you would have to do is you would have to code it programmatically go here go there go
305. 46:34there and then in the past 10 years they've taken um these these more specialized machine learning um models and put them in the robots and
306. 46:44that's how they do their navigation and things like that that's how Tesla's doing self-driving cars what the large
307. 46:50language models offer the potential of is now we can just talk to them so rather than having to have an engineer
308. 46:54kind of you know do all that stuff manually if you get that large language model interface right um You can um
309. 47:05talk to the robots and tell them to go do things and then when you combine that with the autonomous agents like Auto GPT
310. 47:11where you can just give them a goal then you can say hey robot uh I'm low on milk could you run down to the 7-Eleven
311. 47:21the robot will go sure boss I wonder what it'll be like when will it be each of our first moment where we're
312. 47:31just at a grocery store and a [ __ ] robot comes in and starts picking up peaches and sniffing them yeah that's right
313. 47:40to happen it's not going to be that long probably within five years if Elon Musk has his way right he's
314. 47:47figured out the manufacturing [ __ ] so he's gonna he's probably gonna pop some pop some stuff out into the world sooner than sooner
315. 47:57than the world is ready for it uh thanks for the I don't know if that uh coveruno I don't know if that art was
316. 48:04what you were looking for but I thought that was pretty cool actually that was pretty cool so yeah [Music]
317. 48:15any other wait any other law books like oh Amber II oh and other law books like yeah so so the stuff you want to do Black Swan
318. 48:31is very very doable you know I would go if if you if you're D if you're a little bit geeky even go read the documentation for embedding
319. 48:42um at open ai go to open AI I just go to platform.openai but let me see what what it is here foreign [Applause]
320. 48:59yeah if you go to platform.openai.com documentation and then there's a guide for embeddings go read that or go have chat GPT read
321. 49:11that for you and summarize it for you actually some someone wanted to know about that the about some prompting to
322. 49:19do so let's go let's go play with this right now let me see how long this document is oh it's not that long okay
323. 49:25so here's the here's the page on embeddings in openai's documentation okay so I'm going to copy it we're going to hop over to chat GPT
324. 49:39and this is one of these things that these tools are really really good at I'm going to go to gpt4 just because I like it better
325. 49:46and so I'm going to say based on this documentation and then I'm going to hit colon shift return does a line break so it doesn't
326. 49:58submit this and then I'm going to put um pound to pound pound and then I'm going to put another line break and I'm going to paste in all of
327. 50:06that documentation and then I'm going to put pound pound Pound what the pound pound does is let's chat GPT know that
328. 50:13that's the beginning and end of the document you don't have to do that like Chachi PT is incredibly good with
329. 50:19unstructured data with like shitty unstructured oh [ __ ] I just got rid of my my double camera I somehow slid it up can I get it back no ah
330. 50:35that what's that what's that no anyway okay so so sorry 80d Panic um okay so I said based on this documentation pound pound pasted the
331. 50:51documentation pound pound now I'm going to hit shift return again I'm going to say please summarize um what uh embeddings
332. 51:04mean for a non-technical user in you know two sentences
333. 51:19and so now it's going to go read all that for me and it's going to tell me what's up embeddings in simple terms are a way for AI to convert
334. 51:30and represent complex information like text into list of numbers which can be used to find similarities or differences okay listen
335. 51:40this is still too nerdy um please use um clear metaphors and um
336. 51:55bring the meaning to life with exam with real world examples real world examples so so someone before was
337. 52:08asking examples someone was asking about prompting the thing about prompting is it's it's a lot it's very different than
338. 52:18playing with Google with Google you you we've been kind of trained by Google to create the most concise possible answer
339. 52:27and what happens with chat gbt is it's more like you're having a conversation with this thing um so so if it gives you something you
340. 52:34don't like you can tell it that you don't like it and also tell it why you don't like it and then it will just you
341. 52:40know it will get better so let's see what it says I can simplify further imagine you're at a large party and you
342. 52:46want to introduce people to each other based on their interests the process of creating embeddings is like making
343. 52:52mental notes about everyone's Hobbies background characteristics later you can use those mental notes the equivalent of
344. 52:59embeddings in AI it's really cool to introduce people to one another find the most sports loving person or identify
345. 53:06someone who doesn't quite fit the crowd yeah that's really interesting so it's taking individual words in the context
346. 53:12of this Congressional you know legislation and it's saying you know semantically what what is that that word
347. 53:20is in the context of of what meaning the meaning of the other words around it right so like the person with their interests
348. 53:27it's like giving a computer a way to understand and compare the essence of each text similar to how you understood
349. 53:33and compared the essence of each person at the party that's a good explanation that's not bad what app are you using I don't know what
350. 53:42app you're talking about um let's see I'm gonna I'm gonna invite someone up here on screen I don't know who wants to hop up here
351. 53:54hair growth geek you want to hop up here because I think if I talk to you on screen then when I go back it'll give you back my hello
352. 54:05it looks like you're processing hello hello how are you how's it going good another turn the camera on hold on a sec
353. 54:16there I am hello hello hello how are you tonight I've been watching you this is this is so much fun I was at
354. 54:24my computer trying to keep up I couldn't yeah you know it'll it we I you know I mean you've been here enough I cycle
355. 54:33through the same thing over and over again awesome awesome how what can I do for you how can I nothing can you you said you turned on
356. 54:42your camera I don't see you okay there we go I made it smaller there we go hold on there hello you know save it there we go okay
357. 54:53okay cool hello um I don't know how are you doing are you how how are things going are you playing with any AI stuff uh yeah I uh
358. 55:01thank you I am I was actually sitting here and I was playing with um comparing Memento FM to get munch so we can break
359. 55:09down some of our YouTube content and take some shortcuts yeah yeah the minds were short for them and then we post it with repurpose dot
360. 55:18AI but it's not the great uh greatest on uh posting YouTube shorts so uh yeah I know all of the tools right now are a little janky like I I
361. 55:28um I've played with Vizio v i v i d y o that's another one that does that and it's like they're all they're all like okay
362. 55:36just like but that's just the nature of where we are in the news right now it'll get better well I gotta tell you you
363. 55:42know the repurpose.io has been helpful they're not all the way there but it's it's it's really a big help but what I
364. 55:50did this weekend have you heard of maverick.ai no okay so this is how I spent my July 4th so so you know how you have your clavio CRM
365. 56:01flow and you have your thank you card in the band or thank you email the abandoned email Etc so what I did was I recorded all those videos
366. 56:11and then they have you sit there and read names for like 15 minutes which I know sounds really archaic
367. 56:20but they can hit about eighty percent of the names so when you abandon your car oh yeah I saw this yeah
368. 56:29oh interesting yeah and um there's been some very good conversion rates with it so um and I went into Ai and I said
369. 56:41Pretender garyvee write me an abandoned carton thank you email and so I said act as garyvee act as Ogilvy act as who is
370. 56:51the best act as Michael cialdini um act as um Daniel Kahneman act as a neuro-linguistic programmer and
371. 57:02so all these people and in the scripts that came out were actually quite similar in their structure but they're all doing the same
372. 57:11thing did any did anyone show up better than the others you know who was the best garyvee yeah because he's got more content out there
373. 57:19he's just he puts so much content out there that makes perfect sense to me because every all the other guys all
374. 57:25their shits behind a payroll wall but garyvee's just like [ __ ] it here it is yeah so I will share with you the
375. 57:32results they're processing the videos now and we're going to set up our clavio flows tomorrow oh that's cool and then
376. 57:39we'll be running our our paid app to that as well as anybody who visits the website because we use retention and
377. 57:46other services to grab the website and add them to our clavio so even if you don't give us your email we
378. 57:54yeah well of course you know and we're spying on You by the keywords that you're using to search and when you go
379. 58:02to my competitor site so I gave them a list of my competitor sites I gave them a list of my top keywords and we're
380. 58:08going to try to match and then we're going to try to match you to your name and the product you're looking for oh
381. 58:13that's cool well that's I mean all the stuff that you're talking about I you know I'd imagine if you're doing that
382. 58:18much analysis you're probably paying a fair amount of money for that you're going to have a blast with code
383. 58:23interpreter I think you just take a bunch of your raw data and just dump it in there and start asking it questions
384. 58:30you'll you'll be able to do all sorts of custom reports that's going to be insane all right well well so definitely if you
385. 58:36start playing around with code interpreter let me know how it's going because yeah worthwhile yeah yeah yeah that's super cool awesome well
386. 58:44thanks for popping up here thank you yeah turn on all right I got my dual camera back thank you thank you thank you hair
387. 59:00geek all right there's that I got my dual camera back that was awesome thanks for popping up here um props.chat is very helpful check that
388. 59:11one out oh it's Shannon okay all right I lost my place in my questions again okay so who's here so I don't know who's here
389. 59:17I don't know how long you've been here anyway my name is Kyle Shannon this the AI learning lab as you can tell this is
390. 59:22a rambly sort of ADHD nightmare hellscape of enjoyable wacky fun with an old guy talking about new stuff
391. 59:33um if you have questions about AI pop them below I'll do what I can to answer them um if I can't answer them I try to be
392. 59:39straight about that too um the chat GPT isn't good at writing people on my last nerve yeah I know I I know it's it's like um
393. 59:54like I just kind of feel like it's like you know I'm boring and uninspired and I asked chat GPT for something and it gave me
394. 1:00:04something boring and uninspired like well yeah can you explain prompt creation yes so that's what someone was saying to you Shannon
395. 1:00:18um go to the prompts.chat um this thing go there and that will explain to you what prompting is so debate so so
396. 1:00:31prompting is just a geeky term for talking so so you're or here here's a here's maybe a better way to put it what prompting is is
397. 1:00:43finding ways to ask better questions to get better answers so what what chat is when you go to chat GPT it looks kind of like Google
398. 1:00:55and in fact let me go since you're you're kind of starting from the the ground floor on the prompting stuff I'm gonna do
399. 1:01:03um what I call some parlor tricks because the the let me give you a little bit of primer sort of technically some of the things
400. 1:01:14that are happening here it's called chat GPT the GPT part of that stands for generative pre-trained Transformer
401. 1:01:23you don't need to know much of any of that but generative basically means it's generating words so it's actually
402. 1:01:29writing original content for you based on what you ask it pre-trained mean it means it's been pre-trained you don't
403. 1:01:36need to supply it with data you can just ask it questions what's it been trained on it's been trained on essentially all
404. 1:01:42of the public internet for as much as open AI could get over the past four or five years and it's been trained on like all of
405. 1:01:51Wikipedia and all of the blogs and all of the websites and all of the research papers and all of the programming
406. 1:01:58languages and all of the documentation and just like everything you can imagine that's been dumped onto the public
407. 1:02:04internet for the past 35 or 40 years is inside this thing right so it looks like just a lovely simple little little
408. 1:02:12um prompt message but but you're it it's you're not searching for links on the internet like what Google does it's got
409. 1:02:21all of the content of the internet just sitting in there and then the Transformer part of of uh GPT is it's
410. 1:02:30just the technology that allows these tools to be trained on all this data and Google invented it and now they're at at
411. 1:02:37they're they're at risk because of what chat gbt has done okay so some of the parlor tricks so you can kind of ask it
412. 1:02:45anything so I could say um let's see write me a sonnet about um how lovely
413. 1:03:02thunder storms are and it will write a lovely sonnet and it did that actually quite well it did it better than it's done it for me
414. 1:03:13in the past so in the skies we're Tempest dance and thunder roars a spectacle unfolds Enchanted sight the
415. 1:03:20air Ablaze with electricity's might as rain Cascades lightning strikes the shores right so so okay good Palm poem um
416. 1:03:30can you make that a country song boom and it will just write you a country song it knows what verses are it
417. 1:03:39knows what choruses are it it writes them it matches them um what's the chorus here oh lovely thunderstorms rolling on through your
418. 1:03:48crackling lightning and raindrops too you paint the sky with your electrocues in this country boy's heart I'm singing the blues
419. 1:03:58um I could say things like um please add sad country um chords and it will now add it adds in
420. 1:04:14um D C's and G's which if you're in a country song we can do that but I could say um those are boring chords um and progressions
421. 1:04:30make them more interesting and I misspelled a bunch of words there it doesn't care about that so it threw
422. 1:04:38it into a minor and now it's giving me some more interesting chords okay so I can do that so now I can go um please um make this a table
423. 1:04:53and compare the main concepts with Kenny Rogers The Gambler why that just popped in my head hey Gen
424. 1:05:09X Gen X is hard man Gen X is hard so now it's it's literally writing a table comparing that shitty song I just wrote
425. 1:05:17about thunderstorms with Kenny Rogers The Gambler theme Beauty and power of Thunder thunderstorms theme wisdom and
426. 1:05:24life lessons setting Heartland wild winds stormy skies setting gambling table train journey uh so it does that so now I could say um
427. 1:05:37make me a python application um that asks someone for a theme and then writes
428. 1:05:51them A Country Song certainly boom there's your code so prompting is the like as painful as it
429. 1:06:07is to watch me type and I I get told that a fair amount you know I didn't just come on this free internet channel
430. 1:06:14to watch you type for an hour well apparently it did butthole because now you're commenting on it uh so but but prompting is is that
431. 1:06:23interaction and and it's it's um I saw it referred to a couple of days ago as as um um AI psychology or GPT psychology that
432. 1:06:39that you're almost trying to get in the head of okay if I want to get this thing to respond to me as if it were William Shakespeare
433. 1:06:48what would I have to tell it to do that well I could tell it I want you to act like William Shakespeare and I want you
434. 1:06:55to respond uh only in iambic pentameter and I want you to you know make every other line rhyme and whatever it might be um
435. 1:07:05and so so that's what prompting is and so prompts.chat the you know that that document right there will give you some
436. 1:07:13some hints and lessons into that direction and then once you sort of get it my you know I say this a lot that
437. 1:07:18that my My Hope for everyone on this channel is that you play with chat GPT long enough until you have a Kevin
438. 1:07:24McAllister moment which is from home alone right because at some point when you play with this tool you'll crack through those parlor tricks
439. 1:07:35and you'll do something that's relevant to like your identity or something you're good at or like you know maybe
440. 1:07:41you're good at whatever writing recipes or gardening or copywriting or writing strategies or doing marketing
441. 1:07:50um hair growth geek was just up here she's obviously really good at the marketing right so as you start to play
442. 1:07:55with this thing in the area that you're good at at some point it's going to do something where you're where you're just gonna go oh
443. 1:08:06really should I be excited or should I be scared yes like the answer is yes like like I'm the most excited I've been
444. 1:08:17in decades and and I'm totally unsettled because everything's changing everything in my business is changing everything in
445. 1:08:27marketing is changing if I owned an advertising holding company right now I would be terrified everything's changing so hope that helps
446. 1:08:44can you demo something from prompts.chat sure let's go do that that's a good idea let's go to prom Star Chat yeah prompts.chat
447. 1:08:59so this is prompts.chat it's awesome GPT chat GPT prompts just some dude made this and then I think other people can
448. 1:09:07add to it I don't know if this is even being updated anymore but it doesn't really matter so this just sort of explains what chat
449. 1:09:14GPT is it it gives you a free ebook about writing prompts I think it sells you at ebook about writing prompts I wouldn't buy that um
450. 1:09:24build your own GPT prompt app that talks about the API and then we get down into the the sort of meat of
451. 1:09:33the document so the the top thing is just a primer and the meat of the document is act like an English
452. 1:09:39translator right act as a Linux terminal act as a blank position interviewer so like if you want to practice for a job you could say
453. 1:09:50um act as a project manager interviewer right and you can just change out um that thing so let's go find something
454. 1:09:57interesting here and then we'll play with it act as a travel guide plagiarism checked or Advertiser Storyteller football
455. 1:10:04commentator stand-up comedian I don't want to do stand-up comedian because I'm a big fan of Comedy fan and when it does
456. 1:10:11bad comedy it's going to make me cranky act as a debater that could be an interesting one movie critic all right let's do a movie critic
457. 1:10:20all right so we'll read what the prompt is I want you to act as a movie critic you will develop an engaging and
458. 1:10:27creative movie review you can cover topics like plot themes and tone acting and characters Direction scores cinematography production design
459. 1:10:39special effects editing pace and dialogue the most important aspect though is to emphasize how the movie has
460. 1:10:45made you feel which is interesting um what really resonated with you you can also be critical of the movie please
461. 1:10:53avoid spoilers my first request is okay so we're going to copy that we're going to hop over to Chachi BT or as they say
462. 1:11:02on Tick Tock you can make money foreign if you if you listen to those people then you're no longer welcome you can't
463. 1:11:15listen to those people they're all full of [ __ ] they're just trying to take your money you know how they're making money
464. 1:11:19with chat GPT getting you to click on a stupid [ __ ] link okay so we're gonna modify this a bit I want you to act as the
465. 1:11:34movie critic Roger Ebert um [Music] you could also be critical about the
466. 1:11:49movie please avoid spoilers um okay let's see um um write your review
467. 1:12:06so that it is formatted with you know bold headlines and such and organize
468. 1:12:23the review into distinct sections like d distinct sections okay now why am I doing that I don't know
469. 1:12:38because distinct um because one of the things that that you can start to do is you you can you you can push it to go move just
470. 1:12:51beyond a writing exercise and start to think about how are we actually going to present this right so
471. 1:12:57um write your review so that's formatted with both headlines and such and organize the reviews into distinct sections each review should be
472. 1:13:07um I don't know 600 words so so we'll go to length we'll try to make this consistent my first request is
473. 1:13:14I need to write a movie review for the movie Interstellar now we're going to do it for um Pulp Fiction all right Bang
474. 1:13:30so notice that it's created um bold things there's a thing in chat EPT called markdown um which it it actually is sort of
475. 1:13:41coding this on the back end so that it displays things as like bold and normal and italic and things like that
476. 1:13:47so title Pulp Fiction Pulp Fiction a symphony of style and substance the intricate dance of narrative and theme
477. 1:13:53Quentin Tarantino's Pulp Fiction isn't merely a film it's a paradigm shifting spectacle that wriggles and dances its
478. 1:13:59way through a genre blender creating a mesmerizing mix of high art and low culture Tarantino's audacious narrative
479. 1:14:07structures spending time and playing with right so solid writing spectacular performances and authentic
480. 1:14:14characters right so it has now figured out that because I told it I wanted to break this into how did I put the prompt
481. 1:14:22um organize the review into distinct sections so it's doing that and I thought it was going to give us generic things like
482. 1:14:28pacing or plot but no it's actually writing headlines so spectacular performances and authentic characters in
483. 1:14:35the center of the swirling cinematic Vortex or Tarantino's fully realized characters brought to life by Stellar
484. 1:14:40ensemble cast John Travolta Samuel L Jackson Vincent Vega that's not right is it wasn't it isn't it Vincent
485. 1:14:50it's not Vincent Vega I think that's wrong anyway um a Thurman's Mia Wallace is more than a mafia mole she is a mirror reflecting
486. 1:14:59the desires and fears of men a master class in Direction and writing Artistry behind the camera critique style over substance right so
487. 1:15:09brilliant okay so there's there's our there's our review I say um how about um a review of Caddyshack
488. 1:15:24again show my Gen X Stripes [Laughter] Larger than Life not so larger than laughs the cast of Caddyshack is an
489. 1:15:39ensemble comedians who deliver coarse and comedic timing Carl spackler Chevy Chase all right so we'll let that keep going
490. 1:15:49so that's writing that one so that's interesting so now I'm going to say um give me a table that compares the characters
491. 1:16:05and plot elements of these two classics and so here's a comparison so now it's writing a it created a table
492. 1:16:22so here it says here's your main characters here's your supporting characters primary plot secondary plots
493. 1:16:38character arcs the character Arc for Caddyshack is generally shallow prioritizing comedy over dramatic growth yeah that's the point
494. 1:16:52um so there you go so so the it's it's um this idea of of telling chat to act as
495. 1:17:08as a specific thing it it actually why this works is because because these GPT models have been trained on such massive amounts of data
496. 1:17:20when you tell it to act as a movie critic or in this case Roger Ebert right you're kind of narrowing the world of
497. 1:17:27data that it's looking at to to a much smaller subset so if I just said write me a movie review of Pulp Fiction it'll
498. 1:17:34just go look at everything out on the internet and it'll pull up like you know Twitter posts and like whatever whatever it has
499. 1:17:41in there just any idiot that said anything about it you know from IMDb to you know to some blog to Roger Ebert but
500. 1:17:49if you say I want you to act as Roger Ebert it kind of filters out all that other crap and focuses it so if if you
501. 1:17:56want Chachi PT to help you with your quantum mechanics problem solving or conceptualizing or understanding
502. 1:18:05um point it in that direction and you'll get much more focused and generally more accurate less hallucinations because the
503. 1:18:11the way hallucinations work is if if you're casting too wide a net it will just kind of average out you know
504. 1:18:19all of the data that it has and it's a prediction engine so it's just predict predicting one word after the other and
505. 1:18:24so it will just include [ __ ] that has nothing to do with reality because it doesn't know any better it's not
506. 1:18:30actually looking at what it sends to you all right janky shame is my love language you can make money with Chachi VT so Kyle isms
507. 1:18:44it's so true I do use janky a lot and shame is my love language all right I just got a job engineering automation systems cool
508. 1:18:55AI is getting implemented right now yep yep uh Ai and automation this generative AI [ __ ] and automation oh it's gonna be
509. 1:19:03amazing it's gonna be like magic because if you've ever done any sort of automation historically or just any kind
510. 1:19:11of computing any kind of programming like you have to think of like all of the logic trees right like like you just have to like
511. 1:19:22anticipate every possible path people might go down and then you program in all that logic you don't need to do that
512. 1:19:30with this thing it's just it like knows all the [ __ ] that's that's why I'm so excited about code interpreter coming because
513. 1:19:38it's you can give it imperfect data and it will turn it it will clean up your data and turn it into perfect data it'll
514. 1:19:44find anomalies in your data and fix it right so so you've got a bunch of people entering um you know your programmer didn't do
515. 1:19:53you right and didn't break up your Fields enough when you're you know when your customers are inputting their their
516. 1:19:59addresses and all their addresses are all [ __ ] up you'll just go fix all that um I just built a slack integration with
517. 1:20:09chat EBT yeah I've got one of those too that's it's super cool I don't my problem right now is I'm I'm here's something to I'm starting I'm
518. 1:20:18starting to get smarter about this but it's taken me a while um I'm falling into the Trap of being intimidated by there's too much
519. 1:20:27possibility like this thing can do too much so I kind of I get like a blank screen syndrome right where like blank page
520. 1:20:35syndrome where it's just like since you could write anything like you don't know where to start I'm starting to find I'm doing that with
521. 1:20:44both AI images with with chat GPT with with a lot of the tools that we're using with automation tools like zapier
522. 1:20:51the thing that hit me today is I need to flip my mental model to not think about what's possible but to actually start with what's a
523. 1:21:04problem I have right now like I don't I don't care if I know how to solve it or not well that's a problem that I have
524. 1:21:09right now that needs solving and start with that and then from that say okay how would I solve that with
525. 1:21:17chat gbt or the mid-journey with this or with that or with that um and that's that's that's something that I'm I just feel like I have to
526. 1:21:25shift because it's this this stuff is very seductive like once you have your Kevin McAllister moment it's very
527. 1:21:32seductive you get you get sucked into it and and then and then you'll find yourself self spending four hours like
528. 1:21:39writing the Great American novel you know um in iambic in Shakespearean iambic pentameter and then you'll realize I've
529. 1:21:49got this book now that No One's Gonna read and I don't even really care about but I did it because it was there right
530. 1:21:56that it's it's a little like that um so so I think the idea of focusing on a specific problem and then just you
531. 1:22:03know being less seduced by the technology is a big one that I need to work on um appreciate your efforts you're a
532. 1:22:10great teacher thank you very much I appreciate that please tap it's freezing oh is it still freezing sorry is there an application
533. 1:22:20that wouldn't you stream from your computer there is but it's for Windows only and I'm a Macintosh guy I'm an apple Fanboy from way back
534. 1:22:30uh I've hated Microsoft for 35 years so I'm very conflicted now that they're acting like apple and they're more
535. 1:22:37they're more Innovative than really anyone out there right now because because of what they're doing um you know in conjunction with open AI
536. 1:22:45I think they're they're they're up there so I pay for a chat GPT for just for email stuff I'm still afraid to share
537. 1:22:54what I do at work yeah there's there's a uh there's a really fascinating um article by this guy named Ethan Malik
538. 1:23:01who's A Wharton Professor called detecting the secret cyborgs and Ampersand you are what Ethan Malik would call a secret cyborg which
539. 1:23:10basically what he's what he says in the article is these generative AI tools are incredible personal productivity tools
540. 1:23:18but they're not really ready for prime time for businesses and and they're also potentially very disruptive for businesses and they're
541. 1:23:27also potentially legally in privacy and security risks for businesses so businesses are like yeah we can't really
542. 1:23:35[ __ ] with that right now and I don't know if I want to because I think that can screw some things up and the
543. 1:23:41employees are like holy [ __ ] [ __ ] I just got done in four hours what normally takes me the whole week uh I'm
544. 1:23:48not gonna tell anyone because that's [ __ ] awesome so those are the secret cyborgs and what Ethan Malik's saying in the piece is
545. 1:23:55companies need to get their [ __ ] together and say hey you've got sort of you know a whole class of employees here
546. 1:24:02that's becoming secretly AI literate that's not benefit it's benefiting your business but not directly right
547. 1:24:11um and so so the businesses need to create safe zones to invite these secret cyborgs Out of the Shadows
548. 1:24:20um and and really start innovating as as organizations so I think it's absolutely fascinating oh Lord he's going into
549. 1:24:26character it happens it's it I haven't done it too much tonight a little bit but as I get more tired I tend to go
550. 1:24:33into more characters it's also like I've noticed like I don't know I don't know why I'm doing self-analysis on on my
551. 1:24:39Tick Tock performances but um but I've noticed like if I'm if I'm in a bad mood or cranky or just low energy I
552. 1:24:47tend to be just like AI is fantastic here's how you learn it and then if I'm in a good mood I'm like wow so
553. 1:24:56where am I tonight I'm somewhere in between I'm I'm Midland Midland on stage do you use zapier with chech EPT and if
554. 1:25:06so in what capacity have you found it most useful so the most useful things I'm using zapier for right now is just
555. 1:25:12the first things I used it for like you know how when you first sort of crack a technology it's like that that use case embeds on you
556. 1:25:21so it's like um the first thing I did with chat CPT was I built a Content generation engine so it was a thing that asks someone
557. 1:25:29questions and then generated a bunch of social media content content I did it first in Python and then I I went to I
558. 1:25:37recreated it in zapier and I called it the social media marketing machine um and it's cool I mean it it works you
559. 1:25:45answer some questions about a topic and it generates a bunch of actually really good social media content that really doesn't need editing
560. 1:25:52and so so I do a lot of that I did one for my business so my business is called storyvine and it's a video
561. 1:25:59um technology platform we have an app that guides it it provides prompts for someone we're a prompting company we have been for 11 years
562. 1:26:08um it provides prompts for someone to answer to tell a specific story so if it's someone who was diagnosed with
563. 1:26:13cancer it'll prompt them you know when were you diagnosed and um you know what did it feel like when
564. 1:26:19you heard the news and how have you been treating it how are you feeling and all that sort of stuff and so writing those
565. 1:26:25prompts is a very particular skill that we've gotten good at over the years so one of the engines that I built was an
566. 1:26:32engine that automatically writes those prompts for US based on some questions that we ask a client so we get on a
567. 1:26:38phone with a client and an agency which historically would take us three to five creative meetings just to get them to
568. 1:26:46understand how we think about questions versus how the agency does versus and it was always this confusing mess
569. 1:26:54and so now what we do is in the course of the meeting we ask them questions what's the kind of story you want to
570. 1:26:59tell who's the audience you know what's the message what are you trying to accomplish and then when we submit that form it
571. 1:27:07goes off to chat GPT it generates three sets of questions a short medium and long set of questions for that specific
572. 1:27:14story for that specific client and then it sends it to our slack Channel and so the account team now and when when we're
573. 1:27:22on one of these creative calls we ask the client questions it it generates these prompts they literally copy and
574. 1:27:28paste them into a word doc tweak them kind of as we're talking you know they might add one or subtract one and
575. 1:27:34they're good they're like we don't need to tweak much and then we we literally just pop them up on screen and say well
576. 1:27:41so based on the the answers that you've given us here's a starter set of prompts and so we don't have to explain how we
577. 1:27:49write prompts or what we do they're just done and then all we're doing is tweaking the prompts that are there what
578. 1:27:56we would have done historically is say well based on this story what are the kinds of questions we want to ask and
579. 1:28:01then the agency would go well why don't we go off and we'll come back in a week or two and give you that and then three
580. 1:28:06weeks later they would send us 27 questions for a two-minute video and we're like you can't ask 27 questions in
581. 1:28:14a two-minute video and there would be all this back and forth so that's gone so that's what I'm doing with it um we're also we've got
582. 1:28:22um open ai's whisper transcription tool built into our system now so every video in the system now has interactive transcriptions
583. 1:28:31and I'm using zapier to prototype um omnichannel content generation based on a video transcript right so here's
584. 1:28:40what the person said in the video here's the title of the video here's the YouTube description here's a tweet and
585. 1:28:46doing all that sort of stuff so it's I'm using it a lot it's pretty cool and there's there's so much more you can do
586. 1:28:53but that's that's a fun one that I've been been playing with would you be willing to beta test an AI app I'm working hard as hell on to get
587. 1:29:01launch ready I'll take a look at it yeah just my my contact info is in my um what's it called profile there's a
588. 1:29:08there's a link tree there oh you know what I wanna I wanna make a request to everyone on the on yield Channel so a couple of days ago
589. 1:29:19someone had asked hey you know do you do a workshop or do you have a workshop and I said well no not really um but
590. 1:29:29I now have a Graphic for a workshop so I'll put my little head here uh like I'm like I'm actually talking to you so um
591. 1:29:38so I'm thinking about putting together a workshop but I it it it's a lot of work which I'm not afraid of work I don't
592. 1:29:44mind it being a lot of work but what I don't want to do is just um create a workshop and a vacuum so if
593. 1:29:49you go to that URL below the link tree slash real Kyle Shannon the first link there is a link to a survey that's got
594. 1:29:57three questions in it it's like it's like your name email and three questions and so the questions are just like what
595. 1:30:02do you want this Workshop to be like what are you trying where are you in in your AI Adventure you know what are you
596. 1:30:08hoping to learn and then like where do you hope you are a year from now it's they're actually really simple questions
597. 1:30:13so if you'd be so kind hop over there while I'm rambling and just if you'd be interested in the workshop and then that
598. 1:30:20way when I put together the workshop I can do it in a way that that serves as many people as possible as well as
599. 1:30:25possible because I'm codependent I want everyone to be happy if you're not happy I failed as a human all right
600. 1:30:39um so yes I'd be willing to to at least talk to you about testing it I'm listen I have ADD here's what I can tell you
601. 1:30:45about my Beta Testing if you want all of the things tested I'm not your guy if you want me to break your app as quickly
602. 1:30:52as possible I'm your guy I can usually break an app within 30 seconds my my developers hate me for that
603. 1:31:02because they're like okay we got to this is good now it's solid I'm like 30 seconds in fact I just got so we have a we just
604. 1:31:13got rid of one developer we added a full stack developer and the full stack developer is actually pretty good and
605. 1:31:20um it was the first time in I think four years that I was given a beta version of our app and I didn't break it
606. 1:31:29um in in two entire sessions first time in in a in a long long time so that was pretty cool in the beginning that was
607. 1:31:37the word and the Word was with God and the Word was God oh my God foreign that's funny okay did the AI Salon
608. 1:31:47Meetup happened last night yes it did Miss Birch and it was it was good we did a thing so so normally at the AI Salon
609. 1:31:53let me let me give you that info as well [Music] um so the AI Salon is it was funny we were actually working on the um
610. 1:32:04the positioning today and we're like What is the salon so the salon is this we call people AI adventurers in the
611. 1:32:10salon so it's a community of people they're all basically um starting the adventure of getting into generative AI are there already on the
612. 1:32:21adventure but like the the common theme is that it's people that see this as an adventure there's some [ __ ] to learn
613. 1:32:27here it's gonna [ __ ] some stuff up like a good Adventure there's reasons to be excited and like a good Adventure
614. 1:32:34there's reasons to be terrified right and that's um what the salon is so if you go to the salon.ai it's a link tree
615. 1:32:41that's got a link to the Meetup we meet every other week on on Tuesdays we met last night because the fourth was on Tuesday
616. 1:32:48um and we normally have a speaker come in that is you know in the industry somehow they're either an artist or
617. 1:32:54there's someone in business and then we just talk to one another and we have the speaker talk and people can ask
618. 1:32:59questions it's it's a really great community and what we did last night was we didn't have a speaker but we had
619. 1:33:07um what we call AI confessionals and so everyone in the group who wanted to had 10 minutes to talk about something they
620. 1:33:15learned some failure they had some app that they're working on some business that they're starting some cool
621. 1:33:21experience that they had some art that they want to show um and it it was just great and you know one guy just got on and said okay you
622. 1:33:29know you're all are talking about prompting here's some you know prompting tips I came up with so and we record
623. 1:33:35them they're two hours long and they go up on uh YouTube so we've got recordings of back back ones that that link is
624. 1:33:41there and then we also have a Discord so the Discord where I show mid-journey stuff here that's the AI Salon Discord
625. 1:33:47and we have a LinkedIn group as well so yeah so it happened and if you want to be a part of it please join
626. 1:33:58is GitHub co-pilot useful for non-devs does it have a prompt like uh GPT um I assume it is I have not deeply all the
627. 1:34:10sort of coding stuff I've done I've done within chat GPT um just because I'm not really an engineer so I'm usually just sort of
628. 1:34:18[ __ ] around trying to get some little chunk of code modified rather than you know building a whole application uh it
629. 1:34:25is it is ostensibly GPT um you know that's that's um optimized for coding um but I don't I don't know how it's
630. 1:34:33implemented enough and what the I think it's more like a pair programmer where you're kind of programming your thing
631. 1:34:38and it's it's sort of assisting you along the way rather than a distinct experience so I don't actually know
632. 1:34:46um you can do a lot of coding in chat GPT gpt4 in particular is quite good Adobe text is awesome but you can't use
633. 1:34:55it on a commercial site they'll change that MCAT you know it's it right right now it's in beta so they're basically
634. 1:35:01saying we don't we don't until we have this thing dialed in um to the point that we're happy with it we don't want people using the outputs
635. 1:35:09commercially because you know someone will produce something that they'll get whatever sued for someone will write a tweet
636. 1:35:17about and then that'll get back to Adobe and Adobe shouldn't have had you know boobies in their model or whatever
637. 1:35:24stupid [ __ ] thing will come up so it'll get there um it's just not there yet you're not wrong they must do it on
638. 1:35:32purpose too smart to make that mistake I'm not sure what I was talking about there Kyle will you explain your setup
639. 1:35:38in detail I don't get how the little window works yeah so when when I first go live there's like all these like eight little
640. 1:35:48icons that you can set up like goals and effects and all that sort of [ __ ] one of those options just says dual
641. 1:35:59d-u-a-l and it's dual camera so it's like a little icon with two two cameras stacked on top of one another
642. 1:36:05and then when you turn that on you can either choose a square window a round window or or none I don't I suppose none
643. 1:36:12just means you're not turning it on um and then all it's doing is it's using my back camera you know for that and my front camera here
644. 1:36:22and then if I tap in that window so I can move the window like over my face which that's probably better for everyone right solid although
645. 1:36:34wait maybe we should just cover up the neck fat there [Laughter] so I can do that and then if I tap on it it it flips them
646. 1:36:46and so that's all it is there's nothing fancy here except me I'm fancy I do declare it's so fancy did you notice his hair what was going
647. 1:36:58on with it does he do that on purpose good Lord all right robot dog right right my term paper and do a flip exactly and the robot dog will
648. 1:37:12robot dog stuff humping my leg robot dog start humping my leg robot dog What is love exactly you know you joke we're not that
649. 1:37:25far from that we really aren't oh my goodness um what's going on with the sound I don't know it could be my shitty voice
650. 1:37:36it could be my shitty internet um the absolute nightmare of Robo dog running into the grocery store by me
651. 1:37:43thanks for the nightmares I I'm telling you man it's I am guessing within three years someone on there's 58 people on this right now I
652. 1:37:54assume that means including me no but probably in addition there's nearly 60 people on here I think within three
653. 1:38:00years one of us is going to be in a [ __ ] grocery store and we're gonna hear we're gonna go no no you're gonna turn
654. 1:38:08around and there's gonna be a [ __ ] robot picking up a box of Fritos Cheetos free whoa those are in bags I
655. 1:38:16can't even I I'm so upset about that perspective meeting that I don't know what's in a box or a bag anymore
656. 1:38:24all these robots walking around Costco filling up the carts and like mathematically perfect piles [Laughter]
657. 1:38:36a little general intelligence I don't know um Ray Kurzweil is predicting 2029 for the singularity which I think is in in the
658. 1:38:46neighborhood of that um feels plausible to me I mean there's there's a group of Microsoft researchers
659. 1:38:55put out a like a really long paper like 120 page paper um three or four months ago um where they they um were doing tests on gpt4 and they
660. 1:39:07said gpt4 is showing signs of general intelligence um but but there's a couple of things about it that that just aren't there
661. 1:39:16right it it it doesn't have memory it doesn't have reasoning it doesn't understand what it wrote even though the
662. 1:39:22things it's writing feel very human and and it will respond empathetically sometimes and and in ways that looks
663. 1:39:30like reasoning and some of the stuff it does borders on reasoning it is there there are you know unanticipated
664. 1:39:37emergent properties the technicians call them the the emergent properties are like holy [ __ ] I didn't know it could do
665. 1:39:44that we didn't train it to do why is it doing that could it do that so so that's happening um but it's not true reasoning it doesn't have
666. 1:39:55memory right so so once you start to create these autonomous agents where they start behaving on their own
667. 1:40:02and it start has starts to have some self-awareness right I think I think the building blocks for general intelligence
668. 1:40:09there's components of we we see components of them in a lot of the research papers that are coming out
669. 1:40:15right now where they're trying to work on alignment alignment basically meaning we want these robots to um align
670. 1:40:26with the human values like humans are valuable we want that alignment so so so open AI just announced two days ago
671. 1:40:35um they're they're dedicating 20 of their resources to um what they're calling super alignment which is basically don't kill us please
672. 1:40:46so I think I think 20 of their resources to building a tool that has AGI super Intel or general intelligence first
673. 1:40:56um and then putting 20 of your efforts towards making sure that it's good for Humanity that's good it's better than zero
674. 1:41:05might want to up it to 50 but let's start with 20. let's see how evil it gets how how smart it gets and then how evil it gets
675. 1:41:13um can you talk about neural networks uh and and or deep learning I can talk a bit more about the about how chat GPT
676. 1:41:22works I I don't have enough expertise on the neural networks and deep learning piece of it to really probably do anyone any good
677. 1:41:32um if you want to know like how chat GPT Works how the the diffusion models work I'm good with that um I'm just I'm just too inexperienced
678. 1:41:43and frankly it's it's it's not stuff that I could contribute anything to anyway so um I'm not really learning it like I
679. 1:41:53don't know it was five or six years ago and Google did whatever sort of nice press release that that was like
680. 1:41:58tensorflow is now available to play with and you can do machine learning and I'm like oh good I'm gonna go learn machine
681. 1:42:03learning and I went and I started reading the paper for tensorflow and I'm like no I'm not the words in it and I didn't have chat
682. 1:42:13gbt to be able to you know um distill it for me explain it to me like I'm five explain it to me like I didn't go to
683. 1:42:21Stanford explain it to me like I have a degree in acting which I do so yeah so it's probably not good for me
684. 1:42:31to talk about neural networks um I mean what I can tell you is Google invented this thing called the Transformer which
685. 1:42:39allowed these neural networks to be the machine learning model sitting on the neural networks to be trained on massive
686. 1:42:45data sets and the more compute power you throw at them the better they get and that's how openai did what they did with gpt4
687. 1:42:55um I also heard who was it that was saying it um oh George Hots I don't know if you know George hats he's from a company called
688. 1:43:02comma he was he he um he was the first guy to jailbreak an iPhone when he was 17. and he and he became famous for that and then he started
689. 1:43:12a company called comma which was a self-driving car that he was designing out of his garage in San Francisco
690. 1:43:20um and he sort of spilled the beans on gpt4 is actually apparently like 16 different models and eight different
691. 1:43:28weights I don't know there's it's apparently a hell of a lot more sophisticated and complicated than they're letting on so it's it's
692. 1:43:37significantly different than 3.5 structurally apparently according to George hotz but he likes he likes [ __ ] with people anyway so
693. 1:43:47if you've not watched videos of George hot's programming if you want to see the human equivalent of GPT like the the
694. 1:43:55speed with which that so so if you haven't seen GPT do programming um so let me see um write a C sharp application that
695. 1:44:09combines Flappy Birds with snake
696. 1:44:31first the game's rules oh so so this is inventing a game so now it should code now now we'll need several several classes
697. 1:44:41so here's the classes so it's writing us the documentation of the game it's going to build and now it's building the code
698. 1:44:48so watching George hot's program is he's faster than this um make it a python uh script
699. 1:45:06that um I don't know uh sorts um ingredients from a shopping list
700. 1:45:23sure here you go so watching George shots program is like he like types this fast and he talks this fast like if you think I talk fast
701. 1:45:36George hotz is like 10 times faster it looks like my sound is [ __ ] up somehow so embeddings is like an SQL code query it's
702. 1:45:50no well all right so you want to see okay so someone was asking about neural networks I won't tell you about them I'll tell you how
703. 1:46:01how large language models work I've showed this a couple of times before if you've seen this before feel free to take a nap or take off um
704. 1:46:10Keynote this is how this thing works is like like the more you understand how how this thing works the more mind-blowing it actually is
705. 1:46:29Okay so let me get my little head out of the way yeah oh no I just lost my camera again God damn it all right whatever [Music]
706. 1:46:45Okay so we have this thing called chat GPT GPT generative pre-trained Transformer I talked about that earlier generative
707. 1:46:53meaning meaning it's generating stuff pre-trained means it's pre-trained on all of the internet basically and or it
708. 1:47:00can be on whatever but in the case of chat GPT it's trained on on the the public web and then Transformers the
709. 1:47:06technology that lets that happen the large language model is is the core based model the like GPT 3.5 and gpt4
710. 1:47:14that's the large language model um and then the latent space this is this is how it works so so it's there are similarities to a SQL query
711. 1:47:28but the SQL queries for the most part are explicit you have to have good data well organized and then you have to
712. 1:47:35query on that good data good well-organized data what happens with these large language models is when when you talk about embedding
713. 1:47:46you're embedding a document it's taking all of the words in that document and it's clustering them into these
714. 1:47:52mathematical Vector spaces so they call it a vector database so what you see on screen right here is just
715. 1:47:58um two two Dimensions right so you've got an x y coordinate and then these these clusters are numbers each of these
716. 1:48:06dots represents a number in this x y coordinate space and then they're clustered by meaning and then each one
717. 1:48:13is tied to a specific word right so the the real models are like 500 Dimensions or a thousand Dimensions or even multi
718. 1:48:22multiple thousands of Dimensions so it is impossible to comprehend how complex it is but this is the simple 2D XY
719. 1:48:31version of it okay so that's the latent space the way chat GPT works is that based on your prompt the large language model
720. 1:48:40predicts the next word mathematically and then repeats that over and over and over so if your prompt is the quick brown fox
721. 1:48:49jumps over the lazy blank we all know as humans well that's dog because we've heard this a thousand times but how did we get to that word
722. 1:48:57dog well how um how chat gbt and how these large language models do it is is in this latent space when you type your prompt
723. 1:49:04you're creating a mathematical sort of probability function that says there is a word that is the most likely to be
724. 1:49:13there and so what it does is it kind of deselects all the Clusters that are not relevant and then it selects like I got
725. 1:49:20a bunch of light blue dots here and these might all be animals right and so one might be a cat one might be a
726. 1:49:26whatever and then one of those numbers is going to have the highest probability so that says 96 if you go look at what's
727. 1:49:34actually you know at that Vector space it's just a long string of numbers that are the address of that Vector space
728. 1:49:42and then that string of numbers is attached to the word dog and then what the large language model does is it
729. 1:49:48sticks the word dog index fence and then it does that over and over and over again so when it writes you a 500 word
730. 1:49:56response it's doing it one word at a time by just doing this over and over and over again and that's why
731. 1:50:04um that that idea of tell chat GPT to act as a screenwriter it will like narrow your search into into one of
732. 1:50:12these semantic clusters basically so why you'll get better answers is because it got it didn't get smarter it got less
733. 1:50:19smart you you made it specialize a little bit right because it's designed on everything it's got all the latent
734. 1:50:26space it's got all the words in all the semantic clusters and when you narrow it down you're narrowing down that semantic cluster
735. 1:50:35trippy right isn't that insane I lost my uh oh that's my double camera again so I'll put on I'll put on a background
736. 1:50:45we'll do we we've we've been in this Rodeo a few times we know how to deal with this stupid [ __ ] yeah man
737. 1:50:56um so I hope that answers your questions it's like a SQL query if instead of writing SQL queries you just said hey
738. 1:51:04can I have some [ __ ] and it gives it to you hmm and code interpreter which is coming this week to a chat GPT near you apparently
739. 1:51:14um is going to do that even better even more better hi professor Sonya what's happening this is dope I appreciate that oh it just
740. 1:51:27jumped me to the bottom again you Tick Tock I'm afraid I know I know the call is coming from inside the house
741. 1:51:42it's crazy this shit's crazy this shit's crazy it's crazy oh Vincent Vega is right from Caddyshack great that's good thank you
742. 1:51:53yeah that's pretty legit you pay for this one uh I pay for chat gbt if that's what the question was prompt engineering
743. 1:51:59needs to be better here's the deal with prompt engineering you ain't gonna need to do it if you're an engineer if you if
744. 1:52:07you are if you are building applications you're gonna have to be good at prompt engineering but even then um these things are going to start
745. 1:52:13prompting themselves and they're going to start assisting us it's going to get much much easier to just say make me a
746. 1:52:19poem about the sky and it will figure out does it need to ask you more questions about that does it just go
747. 1:52:25ahead and pick you know based on all of your preferences it's going to pick a poet that it knows that you like
748. 1:52:31um so that the prompting stuff is going to drift to the background increasingly um and probably pretty quickly probably
749. 1:52:37within a year maybe two that's amazing I wonder how well the code runs uh the code runs well my first um the code that I was showing you
750. 1:52:46before um my real breakthrough my Epiphany with chat GPT happened in December so like it's like a week or two after it came out
751. 1:52:55I'm sitting in the living room with my wife she's watching some TV show that I was not interested in and I'm like I'm
752. 1:53:01like you know adding out on chat GPT chassis with you it's exciting so I'm doing that and I said oh I I I read on
753. 1:53:11Twitter that this chat EBT can code and I've heard that I should learn Python and you know I'm I just I don't have the
754. 1:53:17attention span to do coding but I'm like maybe this can code so I said hey write me write me a a python application that takes an input
755. 1:53:27and it did it wrote this python code and then I did exactly what you did uh I said I realized I realized it wrote
756. 1:53:36python code I said I don't know how to read python code I certainly don't know how to run it right so I said I don't
757. 1:53:41know how to run python code it said sure no problem just go to your terminal and type in this command line and then do
758. 1:53:46this and install this and this library but I'm like no no no I don't want to do anything on the command line isn't there
759. 1:53:52a website I can just paste this [ __ ] in and see if it works and it said oh sure and it gave me replit and it described
760. 1:53:58okay copy this code from here paste it into the left pane on replit and hit the Run button I was like okay and I pasted
761. 1:54:06the python code into replit I hit the Run button and it went error so I copied the error and I went back to
762. 1:54:12Chachi PT and I said I did what you told me and it gave me this error it said well you don't have this Library installed and I'm like
763. 1:54:19how do you install the library right like working with chat GPT is like working with a very simple intern
764. 1:54:26that's the smartest [ __ ] intern in the history of interns like knows everything but needs Direction on everything right
765. 1:54:36so it said sure you go click on this icon a search window will pop up in the upper right search for this Library name hit
766. 1:54:44install and then it should run and within 90 minutes I had built a functioning content generating application in Python
767. 1:54:55that I was running in replit that would ask you for a topic then it would ask you three questions about that
768. 1:55:02topic and then it would write you five pieces of social media content it's a hashtag Cloud a blog post
769. 1:55:11and and uh what else did I do oh YouTube script it's like in 90 minutes never having written a line of python fully functional so yes
770. 1:55:21it runs now if you're a real engineer real Engineers are going to look at that [ __ ] and go well it's sloppy or it's too too tight
771. 1:55:28or not tight enough or whatever like there's nuances to it so being an engineer and and sort of pair programming with this stuff is the way
772. 1:55:36to go because experienced Engineers know what's good and what's not so they can take the best of what these things do
773. 1:55:42and and you know leverage them it's also really good at documentation you can take code and say um write me documentation for this code
774. 1:55:50or for this API and it'll do it or just pop the code in there have it grab some code out of a random bot out of some
775. 1:55:58stock trading site just pop the code in and tell it to tell you what that code does it will tell you that it's a bot and
776. 1:56:07it'll tell you probably what model it uses so she's doing data analytics this is dope I don't know where I was before okay
777. 1:56:17here I was this is where I was okay good um let's see regenerate geez that's amazing I wonder how the code runs we just did that one
778. 1:56:28damn because it'll hit the news and three days later it's patched that's pretty legit not much substance he was the character Vincent Bagels
779. 1:56:40right pretty damn good it's just parsing info it scrapes wrong Chad it's not I just explained to you how it worked it's well okay
780. 1:56:50well it's not parsing it it's it's not parsing it it's not that's not how it's working it's it's it's doing word
781. 1:56:57probability so so what chat EBT is is is just like when you're in Google when you're writing a Gmail and you
782. 1:57:05write most of a sentence and it shows you the four predicted words that come next that sentence completion that's what it's doing
783. 1:57:14um so it's it's what it's not doing is it's not taking chunks of like when I said act as Roger Ebert it's not taking
784. 1:57:22chunks of his review and copying and pasting them into the answer it's kind of amalgamating all of the possible
785. 1:57:33um words in that semantic region and saying of all these words which is the next most probable single word
786. 1:57:40and then it takes that entire prompt and that word and says okay based on all that what's the next most probable word
787. 1:57:47and then what's the next most probable word so it's literally the generative part of generative pre-trained
788. 1:57:52Transformer is it's generating original content it's mind-blowing that it does what it does like how anyone figured
789. 1:58:00this [ __ ] out and figured out latent space and Vector databases and you know thousand Dimension mathematical like
790. 1:58:09like just the level of math in this [ __ ] is just mind-numbing but this goes back to my favorite there's an Arthur C
791. 1:58:16Clarke saying any significantly Advanced um technology is indistinguishable from Magic this shit's magic it's magic
792. 1:58:28if you understand how it works I think it gets more magical question chat cpt4 includes all of the visible internet probably not all of it
793. 1:58:39but basically all of the public internet so anything that's behind firewalls like private corporate data like Bloomberg
794. 1:58:46just introduced a thing a few months ago called Bloomberg GPT where they took all of their proprietary financial data and
795. 1:58:54they created a GPT you know sitting they did all the embeddings of that and and did a private version of Bloomberg data
796. 1:59:02that you can interact with with GPT um so that stuff's not in the model but basically as much data as they could
797. 1:59:11find as open AI you know could find and tag and and do what they needed to do for the embedding um that was on the public internet
798. 1:59:19that's what it's been trained on so so yeah it's essentially all of the visible internet like realistically it's probably some
799. 1:59:28percentage like 70 I don't know like they might know like there might be there might be a percentage that they
800. 1:59:35know it is and and I assume gpt5 will be you know if if GPT 3 was 50 and 4 was 70 I'm sure five will be like 95 plus a
801. 1:59:48bunch of proprietary data sets where they cut deals with those companies um how it resonated from Wikipedia so
802. 1:59:57Wikipedia's reception from Wikipedia Wikipedia is a general opinion site he definitely paid for this I don't know what I definitely paid for
803. 2:00:06but possibly Caddyshack sucks Chad you're so dark if you're gonna troll please be creative boring trolls not interested
804. 2:00:17um it parses prompt engineering Are You The Prompt engineer no I'm a dude I'm an entrepreneur um but I'm a writer being good with
805. 2:00:26words helps with the prompting I sent an email yesterday had to edit it to not sound so friendly grumpy old man
806. 2:00:33obviously you can if it if it makes [ __ ] that's too um flowery for you you can tell it to limit Pros it'll make it more sparse you
807. 2:00:42can tell it to make it more serious you can tell it to make it more firm um I saw I saw a tick tock today I was I
808. 2:00:50was driving and I was flipping through tick tocks and this guy this guy goes uh what do I say rewrite this resignation
809. 2:00:58letter I [ __ ] hate this company I [ __ ] hate the way you treat me the stupid [ __ ] dinners you forced me to
810. 2:01:04go to every other night or a waste of my [ __ ] time you don't pay me what I'm worth I mean he just [ __ ] went on
811. 2:01:11this you know [ __ ] three minute rant into his phone and then it rewrote this professional resignation letter you know
812. 2:01:18my experience with the company well positive has been you know it was pretty good so yeah you can you can make it go the other direction
813. 2:01:30all right I have I have 15 more minutes and then I gotta I gotta leave it looks like we got some new folks here
814. 2:01:36so my name is Kyle Shannon I normally have a dual camera going but that went away whatever tick tocks special um
815. 2:01:44I'm an entrepreneur I'm the CEO of a company called storyvine This is the AI learning lab all these URLs behind my
816. 2:01:50head why they're here let me get out of the way um if you haven't played with chat GPT I'm obviously a little passionate about it
817. 2:02:00you know enough so that I spend like two or three hours a night talking about this [ __ ] um that's the official chat GPD website
818. 2:02:06so go there right now start playing with it if you haven't bing.com is also chat gbt but that's Microsoft search engine
819. 2:02:15and if you click on the chat button it's gpt4 and it's connected to the internet and you can make free images there so
820. 2:02:22this is these are the guys that invented it these are the guys that gave them 10 billion dollars and are completely
821. 2:02:28incorporating it in everything they're doing right they're going to roll it into office 365. it's in GitHub it's
822. 2:02:34going to be in Windows it's in all the Azure cloud services it's going to be everywhere so that's that and then poe.com is quora
823. 2:02:42if you know this the question and answer site quora that's their search engine or that's their chat bot but you can do
824. 2:02:49there's two interesting things here one is you can play with six different large language models so you can kind of
825. 2:02:54compare them one of them has a hundred thousand token context windows so you can put a lot of words in it and then
826. 2:03:00they also have a bot maker and so there's a lot of bots here you can play with which are just think of them as
827. 2:03:05like specialized chat gbts that one will rewrite your mid Journey prompt for you one will help you write emails there's
828. 2:03:13like thousands of them now and you can make your own bot so you can practice your prompt writing uh by making a bot and then prompts.chat
829. 2:03:21will teach you how to prompt these things how to talk to these things and how to make them act in ways that are
830. 2:03:27valuable to you and then the final Thing Future pedia is a directory of like 3 700 uh AI tools so can you move the blue window it's
831. 2:03:39covering your face uh that was because I think I moved it on my face audio's fine Milk Bones there's already a restaurant with a
832. 2:03:47robot weight staff yep and I think I think Wendy's is testing a uh a synthesized voice that talks to you at the when you pull up to the
833. 2:03:56drive-through it's not a real person I think Wendy's is testing that right now Kurzweil predicted AGI and aging are
834. 2:04:04going to be overcome by the 2030s Yeah Yeah his a recent interview with Kurzweil said um 2029 is is that was the year he predicted and
835. 2:04:13he said he's sticking by it he feels like that's on track open the pod bay doors hell I'm sorry Dave I'm afraid I can't do that
836. 2:04:25spelling that's correct h-o-t-z um wild intelligent crazy as [ __ ] dude but brilliant it like whenever whenever
837. 2:04:36I want to be humbled there's there's two guys right now that that are in that space for me one is George hotz although
838. 2:04:44I haven't watched him in a while he's sort of resurfacing because he's doing some of this AI stuff and the other one
839. 2:04:48is this guy named David Shapiro um who's neuro spicy as he describes it um you know on this on the Spectrum and um
840. 2:04:58and uh he's just brilliant and he's written uh David Shapiro has written five books on on artificial intelligence
841. 2:05:07and he's got a number of Open Source projects to to create AGI he wants to create an open source artificial general
842. 2:05:14intelligence so he wants to do it in parallel with what open AI is doing but do it with open source you know
843. 2:05:23um large language models and with tools he's writing and he's building a whole Community around it but he's [ __ ]
844. 2:05:28Wicked smart and he does like these these long like PowerPoint presentations on theories of like economics and like
845. 2:05:38uh uh uh Universal basic income and and like what are the implications like if these things are smarter than us and the
846. 2:05:44alignment issue of like we want to make sure that these robots are aligned with our values so they don't de-prioritize us
847. 2:05:53um really really smart guy I appreciate your willingness to share and respond to questions will you remind me of the a uh
848. 2:06:00the AI Salon URL yes I will that's it the salon the AI and then this thing down here is a newsletter that I did
849. 2:06:12with a guy named Greg mushin so that's the Twitter handle everyday AI news um but yeah that's the salon so if you
850. 2:06:21want to join the salon we've got a Discord we have meetups we've got a LinkedIn group and it's just it's like a
851. 2:06:27group that are just kind of on this adventure together trying to figure it out and there's people everyone from
852. 2:06:33like people have been working in machine learning for like you know 20 years like you know baller kind of Engineers to
853. 2:06:41like people that are like I heard this say I think it's something I should pay attention to I'm scared is it gonna kill us
854. 2:06:49everyone in between those two ranges it's it's a cool group do you think in the future AI can create new jobs I think AI is going to create
855. 2:06:59lots and lots of new jobs like like jobs we can't even imagine right now um I think it's going to create a lot of
856. 2:07:05them I think it's going to be a net dramatic positive um assuming we can get the alignment thing right and um but
857. 2:07:15short term it's gonna [ __ ] up a lot it's gonna disrupt a lot of jobs it's going to change a lot of jobs any job that's kind of a tactical
858. 2:07:23execution anything where your your some form of a cog in the in the wheel of this big complicated machine where you're doing repetitive
859. 2:07:36like Soul crushing the same kind of work over and over and over again whether it's cold calling as a salesperson or
860. 2:07:44doing data entry or cranking out you know resizing images for social media in a social media department or cranking
861. 2:07:51out you know copy for 57 variations for a b testing anything where you're a cog in a wheel or you're a cog in the in the machine
862. 2:08:02of your company or or Society those jobs are going to be automated out now are those people going to be
863. 2:08:12automated out not necessarily because you know someone's got to operate the machine that the cogs are in so if if the cogs
864. 2:08:20get automated and you can make more things faster and more efficiently well what what does that mean it potentially
865. 2:08:27means you don't need as many people but it also potentially means you could use the same amount of people to generate
866. 2:08:33you know more output right so you could give lower prices to your customers or you could give higher quality or you
867. 2:08:39could turn things around faster with the same amount of work some companies are just going to jettison employees others
868. 2:08:46are going to find ways to keep employees and basically upskill them to improve the company um but it's gonna be a it's going to be
869. 2:08:54pretty ugly for my guess is two to three years and it's already starting like we're already starting to see I forget I
870. 2:09:00just saw one today the 40 4 500 jobs were being eliminated from a big company direct result of of generative AI
871. 2:09:10um which AI platform generate graphics for free so bing.com so let me pull up that graphic again bing.com is
872. 2:09:20um if if you're in you go to bing.com and then you click on the chat button in the chat window just say make me an
873. 2:09:26image of and describe whatever you want and then if you want go to poe.com and there's a there's a bot for Dolly
874. 2:09:34d-a-l-l-e and and you can type in a little prompt for an image and it will rewrite it to to give you a better image
875. 2:09:42so you can go here find a dolly bot and then go to bing.com and paste The Prompt in to bing.com and they'll make you one so that's free
876. 2:09:54um if you're if you have Adobe Creative license Creative Cloud it's called um the you you can get access to Adobe
877. 2:10:03Firefly which is generative AI so it's not free but if you if you've already got a Photoshop license you can do generative fill within that
878. 2:10:11um sign up for a there's a site out there called leonardo.ai like Leonardo da Vinci leonardo.ai and that requires you to sign up for a
879. 2:10:24wait list but it's very worth it because they give you like 150 credits a day which is I think creating an image
880. 2:10:31there's like eight credits so you can do you can do a fair amount a day uh and then I don't think it's super expensive
881. 2:10:37to do like their lowest tier um so that's that one um stable diffusion um there are all sorts of free options
882. 2:10:47but you got to be geeky you gotta like set up a virtual machine or you gotta have some decent iron or you got there's
883. 2:10:54like there's just a fair amount you got to do to get it running but stable diffusion is open source and there's an
884. 2:10:59incredible developer Community like there's if you want flexibility and power stable diffusion is the way to go if you want quality
885. 2:11:08um Mid journey is the way to go and that that you you have to pay like I pay 10 bucks a month for Mid journey and that
886. 2:11:14gives me enough I don't I haven't really hit my boundaries on that I don't I don't make a ton but it gives me enough
887. 2:11:19but it's from a quality standpoint there's nothing like mid-journey right now um if you're in a business and you want
888. 2:11:26it to be commercially safe and viable Adobe Firefly and generative fill because it's trained on non-copyritten images
889. 2:11:36uh the crappiest one is Dolly the thing that sits underneath Bing but that'll get better at some point um but it's free
890. 2:11:44and then there's another one called Blue Willow that's kind of like mid-journey you log into a Discord server for Blue Willow
891. 2:11:52and then you type uh commands to make your images I think it's built on top of stable diffusion but that's free and I
892. 2:11:59have no idea how they're doing that for free I don't understand who the [ __ ] is paying for that but maybe it's maybe
893. 2:12:05it's one of those models where they raised enough money where they're like let's go get customers first and then
894. 2:12:09we'll figure out how to monetize it that's too stressful for me I couldn't start that business I started a service
895. 2:12:17business and then my current business storyvine we sold it first like we basically didn't build storyvine until we had someone to pay for it
896. 2:12:25uh uh the salon.ai you're helping each other that's awesome not that trippy up it just jumped me to the bottom again I lost my pace and
897. 2:12:35that's my place in the questions people I have seven minutes left until I turn into a pumpkin and then I must leave I
898. 2:12:42must go to the ball Papa I must oh papa why do you treat me so do you need to be great at math to learn
899. 2:12:57Ai No well depends on which side of the inflection point you are so here's here's the here's the deal with this
900. 2:13:09pre-november 30th 2022 the answer was essentially yes like you need to be you don't need to be good at math but you
901. 2:13:15need to be pretty geeky you need to be an engineer you need to understand programming you need to understand python you need to understand
902. 2:13:23yeah so the answer is [ __ ] yes November 30th 2022 that site launches so that the application that you play with there
903. 2:13:39chat GPT the original version of it that they launched was written in two weeks two weeks how did they do it that quickly because
904. 2:13:52the documentation for chat GPT was actually uh the the instructions for how to build chat GPT we're in the
905. 2:14:00documentation of gpt3 and nobody built it and so Sam Altman Microsoft's getting ready to make this billion dollar
906. 2:14:07investment announcement and he's like we gotta have something to show because we can't just show them the the dev tools
907. 2:14:12so go build chat GPT so they built it in two weeks and they launched it on November 30th 2022. what that tool does or what that tool
908. 2:14:22did was it basically turned the English language or your native language but let's just say for the purposes of the words I'm saying here
909. 2:14:32it turned the English language into a programming language so you just go there and you just tell it words and it does [ __ ] for you you
910. 2:14:41can say make me an application that calculates the average speed between this object and that object and it will go right that application
911. 2:14:52and it will probably work and if it doesn't work it will help you debug it um you can also just ask it to help you
912. 2:14:59write a paper or a proposal or a marketing plan or copy or whatever or create a table so um no you don't the other amazing thing
913. 2:15:10about that date is from the day this launched until they got to a hundred million users was six weeks
914. 2:15:18it was the fastest adoption of consumer technology in history the World Wide Web when the World Wide Web launched it took
915. 2:15:25six years to get to 100 million users this took six weeks so you've got at this point probably two or three hundred
916. 2:15:33million people that are using this thing where the English language now can deliver like magical results right
917. 2:15:42you've got all of the knowledge of the public internet at your fingertips and knowledge not just information not like
918. 2:15:49I've got to go read an article about Python Programming no no it will write the python code for you so um knowing math
919. 2:16:01probably doesn't hurt but I have a degree in acting like I get like I've got a I I will occasionally
920. 2:16:12stumble upon a thing I'm like oh I could use algebra to solve this and I'm like wait a squared plus b oh [ __ ] like I
921. 2:16:19gotta go back to like eighth grade like all the time so so like no you don't you don't need to be good at math at all
922. 2:16:31Reflection Point if you want to if you want to play on the machine learning and build models and do things like Lang
923. 2:16:36chain and do your own embeddings and all that sort of [ __ ] yeah you should be good at math you want to use these tools nope
924. 2:16:43what you do need to be good at is critical thinking that's the new thing like if you're just sort of a cog in the wheel I just show
925. 2:16:49up at work I do my thing that shit's gonna get automated out the new skill is being able to to understand what you want to accomplish
926. 2:17:01and how to figure out which of these tools can accomplish that and then when you when it starts generating stuff go oh that's
927. 2:17:09good that's bad that's good okay if I took that and put that here that critical thinking [ __ ] the value of that goes up dramatically
928. 2:17:19all right it learned to auto complete programming languages by modeling predictive tests text on flawed code oh
929. 2:17:26that's cool I didn't know that big game I love the name that's awesome the input has lots of flaws so the
930. 2:17:33generated output will also end up with similar flaws yeah exactly that's I mean the hallucination thing they'll get that
931. 2:17:39worked out it's not magic it no no I didn't say well it is it is perceived as magic right any significantly Advanced
932. 2:17:48Technologies indistinguishable from Magic that's that's what I was saying when when that thing [ __ ]
933. 2:17:54writes you something that you know would have taken you two hours and it does it better than you in 30 seconds that's
934. 2:18:01[ __ ] magic how it does it is not magic how it does it it's a [ __ ] calculator but it's it's a it's a sophisticated
935. 2:18:09enough calculator that it's like that's [ __ ] Magic when when you watch it write code and then the code [ __ ] runs I did this
936. 2:18:16with a 10 year old he he had a a clone of Flappy Birds and it had a bug in it he couldn't figure it out and we copied and pasted it into
937. 2:18:26chatgpt it found the bug you know of course like that it fixed it he copied it back into um into unity and hit play and it ran
938. 2:18:38and he did this he was like and he looked at me and he goes this changes everything that's [ __ ] magic I'm sorry it's magic
939. 2:18:52I mean Magic's not magic right you know magic tricks aren't magic it's sleight of hand and it's like 30 years of practice
940. 2:19:00but if you don't know how to do that it's magic right same thing with these tools it's 30 years of neural networks
941. 2:19:07and Mathematics and Technology and [ __ ] ups and you know Jeffrey hinton's original you know neural network and
942. 2:19:16large language model was whatever it was like 18 words that you know the original prediction was this in
943. 2:19:23incredibly small thing so all of that's in this right all of that history all that Legacy is there the 30 years of a
944. 2:19:30magician practicing sleight of hand so it shows up as magic so chat GPT shows up on November 30th it's [ __ ] Magic
945. 2:19:39because the 100 million people that signed up for that weren't building neural networks weren't engineering this [ __ ] they were just like
946. 2:19:50cut it right could it could it write a blog post oh [ __ ] but it but it can't write a tweet good right oh but but it could it it doesn't do
947. 2:20:01hashtags oh oh my is my job in trouble corner of the internet their little
948. 2:20:15corner of knowledge worker of creative work they're realizing isn't quite so special after all and so what do you do with that
949. 2:20:26that's existential right this is a lot of people's identity is in the work they do and Along Comes A Tool that can do it like that
950. 2:20:38it's intense it's just a load of Matrix manipulations or multiplications it is and it's magic I'm a prompt employee to see now that's
951. 2:20:53solid Polonius that's good that's a solid see that's somewhere between a troll and a compliment this is
952. 2:21:00this is the kind of trolling we like at the AI learning lab smart smart trolling love it and that's not even trolling that's just
953. 2:21:08like a good joke I'm a prompt employee you get it great session lot learned thank you very much appreciate it love this thank you
954. 2:21:16what's the salon url url it's the salon.ai can you give some examples of where you expect big opportunities
955. 2:21:25will be with autonomous agents thank you for the rose um and I gotta go um yeah I I think autonomous agents are
956. 2:21:36ultimately where this goes so right now we've kind of we kind of have to go to the to the AI right we have to go to
957. 2:21:43chat GPT right we have to go to Bing when when um these guys add code interpreter you're gonna have to go there and upload
958. 2:21:51your spreadsheet right the autonomous agents are going to be like so so imagine that we were all like a big cigar
959. 2:22:03smoking CEO in the corner office and you know we got we got the the the the you know the the uh like the 50s secretary
960. 2:22:12March get your ass in here you know go get me lunch so imagine if everyone had an agent that you could ask it to do anything and
961. 2:22:22it'll just go do it that's where it heads right so so I think initially it's going to start out with relatively mundane stuff like
962. 2:22:32um I wanna I want to um you know plan you know I want to plan a nice vacation for the family right so right now how would you do that well
963. 2:22:42you'd go you'd you'd go talk to the family and you'd say you know who wants to go where and maybe we'll go Google
964. 2:22:48and look at destinations or I've always wanted to go here right and you get in the fight you get in the fight about
965. 2:22:53where we're going and then and then everyone's in a bad mood and then you go to work the next day and then you go
966. 2:22:58Google you know the your least favorite place because it's like the it's the one that you know people won't kill each
967. 2:23:05other at right so so you go there and then you start googling right and what does Google do it presents you with like
968. 2:23:128 000 [ __ ] options and um 60 of them are paid SEO injected [ __ ] sites just trying to sell you some rip-off thing right so most of the
969. 2:23:27time trying to like plan your vacation is sifting through horrific SEO just [ __ ] [ __ ] right like like the usefulness of Google has just
970. 2:23:40plummeted right as as their Monopoly increased and everyone started following their rules for SEO and as SEO agencies
971. 2:23:48got better and better in what they did and writing all this fake content to generate the views to get the people to
972. 2:23:53click that that right it became useless so with an autonomous agent I might just give it a go you know
973. 2:24:04um my wife wants this kind of vacation I want this kind of vacation my kids want this kind of vacation can you suggest
974. 2:24:11five places that will make us each the least pissed off here's five I forgot about whatever you know Gatlinburg Tennessee that's where
975. 2:24:26Dollywood is okay that could work never would have thought of it good can you go now book me um you know we want to go in September and
976. 2:24:37and uh we don't want to spend more than this much on travel and accommodations sure Boom the agent will just go off it
977. 2:24:45will automatically find the best deals book all the [ __ ] because you'll have given it permissions for whatever
978. 2:24:49accounts it will go book all that stuff and then while it's doing that and it'll probably do that in 10 minutes maybe
979. 2:24:57five like depending on how quick those apis are it'll just go do that [ __ ] and it'll come back and it'll list it'll
980. 2:25:02provide you with a list here's some options do these work for you oh no I forgot we hate flying in the morning we
981. 2:25:08want to fly in the afternoon so we just get to the hotel and go to sleep okay great it'll just go rebook everything right
982. 2:25:14and then while it's off doing that you say well what are we going to do during the day because we'll one day we'll be
983. 2:25:20Dollywood and but we're gonna we're gonna be there for five days so we want one day where we do some shopping stuff
984. 2:25:26that the wife likes and one day we want to do outdoorsy stuff that I like and one day we want to do something that the
985. 2:25:31kids like right so you give those instructions to your agent you just say here's here's my goal I want to make
986. 2:25:37these people happy we have four days to do it we don't want to spend more than 500 a day on activities and food go book us daily itineraries
987. 2:25:46foreign so right now you can do that you can go to chat GPT and with plugins you can use some of the plugins and things like that so
988. 2:26:00all right I gotta go let's see so anyway so so that's where they start I think is with little personal projects shopping travel
989. 2:26:13probably a lot of work stuff project management autonomous agents would be great at project management why because
990. 2:26:20all the documents you've already got all the all of the project management tools are out there we also
991. 2:26:27know how complete all the documents are like we know that that document wasn't finished and we know that Sally was working on that document
992. 2:26:37so the agent can just go hey Sally how's that thing coming along would you like me to help you finish that document sure that would be great
993. 2:26:44here's thirsty can you just give that the thumbs up Sally and then we'll move that on to the next step and stay on
994. 2:26:50track with this so you'll see it and work and then I think they start to get more sophisticated and where they
995. 2:26:56ultimately go is you know Elon Musk starts to turn his robots into um a purchase that you would consider
996. 2:27:06buying just like you buy a car right so you're going to have a personal robot for the family just like you have the
997. 2:27:11family car you'll have the family robot and then you can just say you know you're you're out back working on the I
998. 2:27:19don't know the the porch the board popped up on the porch and you realize you got the wrong kind of nails and
999. 2:27:24you're just like oh [ __ ] uh Jimmy you named your robot Jimmy because why wouldn't you you like Jimmy and Jimmy
1000. 2:27:31comes over like could you run to the Home Depot and take these nails back and get me the nail the right Nails sure
1001. 2:27:37it'll go off and I'll hop in your car I'll drive to Home Depot it'll walk in it'll freak the [ __ ] out of the people
1002. 2:27:43at Home Depot for the first year and it'll get your nails and they'll come back right so so It ultimately
1003. 2:27:49plays out there like it plays this all this [ __ ] plays out to The Jetsons it really it really does it plays out
1004. 2:27:58like like we're living in Jetsons and by the way three days ago the FAA approved the first fly-in car um design
1005. 2:28:10three days ago the FAA said yes this is we we will consider this you know you know we're going to allow you to put
1006. 2:28:19this design into testing and if they approve it then we're going to have a you know an FAA approved flying car so
1007. 2:28:26we're going to literally have the Jetsons it's coming like we're not that far away sorry Dave I'm afraid I can't do that
1008. 2:28:38all right everybody listen I know I didn't get to all the questions uh we'll do one more what AI company should you invest in for
1009. 2:28:47a risk reward perspective you're asking an actor what company too invest in I don't know um I like what Microsoft is doing I
1010. 2:28:55think Google's in trouble but they're Google and they get a bunch of smart [ __ ] people there and they invented this [ __ ] so um
1011. 2:29:03I would probably be equally surprised if Google didn't come out with something remarkable um as as if they came up with something
1012. 2:29:13that like like they're probably going to come out with something quite remarkable um Nvidia like like AMD is trying to get
1013. 2:29:24into the Nvidia game but nvidia's got it kind of locked up right now but they have a trillion dollar market cap so I
1014. 2:29:30you know do they go to 2 trillion I don't know but right now they're kind of the only game in town for this GPU [ __ ] for AI
1015. 2:29:38um I think anthropic's really interesting but I don't think they're public yet um I think Apple's gonna [ __ ] kill it I
1016. 2:29:54think apple with the autonomous agents back to the autonomous agent thing what Apple's got that no one else has is
1017. 2:30:01they haven't caved on privacy like they will not give the government backdoors to their technology for privacy so
1018. 2:30:09they've got they've got some real credibility with privacy um and and that's going to get increasingly important and and the other
1019. 2:30:17thing that Apple's doing is they're building their own silicon and they're they're able to do a lot of machine
1020. 2:30:22learning processing locally so instead of having to send [ __ ] out to um big Microsoft servers like when you
1021. 2:30:30do an open Ai call that's going to Microsoft service so you're sending your data to the to the machine to the cloud
1022. 2:30:35Apple's going to be able to do all that processing locally so so I think Apple their combination of user interface like
1023. 2:30:43Kick-Ass technology and and the Privacy thing they're going to be able to do um autonomous agents really well so I
1024. 2:30:52think they're going to be amazing um and if you want to bet on the Vision Pro goggles um I mean you know Apple's always interesting
1025. 2:31:02um I don't know those are the those are the ones I'm paying attention to right now it's it's a fairly small list
1026. 2:31:08um for for me the one that that is I guess I guess the two that are most interesting to me right now are are anthropic and
1027. 2:31:18um open AI although what's the one there's there's a guy that found a deep mind inflection he just he just got funded
1028. 2:31:29for 1.3 billion dollars for their series B 1.3 billion dollars for a series B um so that's probably an interesting company to look at
1029. 2:31:40but anyway all right everybody um I gotta run I do appreciate you hanging out with me it's uh it's always
1030. 2:31:48humbling to know this many people can hang out and I hope this helps I know it was kind of all over the place tonight
1031. 2:31:53but that's what you get you get add some Focus sometimes it's not peace everybody have a good night