aiLL 7/16/23 - The Future of Generative AI

Transcript

1. 0:08foreign
2. 0:28welcome to the AI learning lab got a situation going on in the hair hair department but you know it's part of the situation [Laughter]
3. 0:43welcome to the AI learning lab if you're new to generative AI that's what this is all about well even if you're not new to
4. 0:50generative AI it's still about this we are talking about Chachi p and the like and anything else we want to talk
5. 1:00about because that's how we roll welcome welcome welcome happy Sunday night I have not been thinking about AI much
6. 1:10today so I don't have a lot of a ton of new updates but good evening very sorry I missed your answer a couple of days ago about
7. 1:18best way to clone my voice oh yeah you missed a good one I did it good I did 11 labs and did I can show that again tonight Mr Stone Mr Stone
8. 1:37see is it hot is it I mean it is hot but is it cooling enough that I should have the window open or should I put on the air conditioner
9. 1:47hmm choices choices choices welcome welcome welcome do me a favor if you would hit the hit the like button share
10. 1:55the live let's see if we can get some folks out there join them in the fun my glasses are incredibly dirty I'm just doing just doing life
11. 2:10maintenance over here [Laughter] welcome everybody to the AI learning lab um Kyle Smith another Kyle in the house
12. 2:22looks like the numbers are not updating again how many how many folks we have in here now I'm showing 10.
13. 2:28and it looks like there's more than that that if I'm seeing names pop up nice likes oh yeah there's definitely a
14. 2:41day 132. holy crap oh yeah now it says 137 okay uh welcome everybody my name is Kyle Shannon this is the AI learning lab
15. 2:51um Let me let me flip over so so here's the deal I'm an entrepreneur uh my background is I've started 12 or 15 companies
16. 3:00um one of the companies I started in the mid 90s was called agency.com it was one of the first digital agencies so I got
17. 3:07to experience the early days of the World Wide Web and got to ride that um rocket ship up and down and through
18. 3:17through the.com Bubble Burst and uh ended up selling that company to Omnicom in 2002 but this generative AI stuff this this Chachi PT thing
19. 3:28what is this chat GPT thing uh it is very very similar uh to the um to the early days of the web and uh I
20. 3:39think it's going to be significantly more profound and I think it is coming much much quicker than the web came the
21. 3:46web took about six years to get to 100 million users chat GPT took about six weeks so this channel is all about
22. 3:53talking about kind of post chat GPT inflection point sort of you know for the last nine months we've been dealing
23. 3:59with this new generative AI thing what is it you know what's it mean uh that's what this channel is all about so if you
24. 4:05have questions about AI pop them in the comments below I'll do my best to answer them uh I swear occasionally so if you
25. 4:12don't like that get the [ __ ] out hahaha uh what else I tend to get a little smarmy as the night goes on I do try to
26. 4:22answer all the questions so fair warning I will get behind on the questions you can make money with Chachi mate thank you Chad [Laughter] um
27. 4:32yeah I don't like those people that say that but uh it's been become kind of a funny meme in the channel uh uh
28. 4:40I love you Kyle does anybody else love Kyle thank you Matthew I appreciate that you chase them away we're up to 170 now
29. 4:52we're down to 140. so I don't think they do I think they're like who's this old guy what the [ __ ] going on with his
30. 4:58hair why did he swear at us all right let's get started here oh let me let me uh tell you what the okay
31. 5:06so this used to be three sites now it's five and it's probably gonna be seven at some point and then I'm at some point
32. 5:12I'm gonna have to narrow it down but basically these are what I'm considering the major players in the chat GPT game
33. 5:20um there's another one out there called Pi Pi dot AI that is at least in the neighborhood of a major player they just
34. 5:26raised 1.3 billion dollars so let's call them a major player um but their their model is is not not quite as capable as is what a lot of
35. 5:36these are doing um it's just different it's a bit more conversational so chat.openai that's the official chat EBT website if you haven't
36. 5:44played with chat GPT just go there go there right now while you listen to this old man ramble on about this [ __ ] and
37. 5:50start playing ask it to write you a poem ask it to write you a song ask it to write you a python application just ask
38. 5:56have it uh create a meal plan plan for your family for the next week and then turn that meal plan into a shopping list
39. 6:02of ingredients and then have it sort those ingredients by section of the grocery store do anything just go do something
40. 6:11um the the thing down below here prompts.chat that's uh a document that'll talk to you about how to talk to
41. 6:18these tools these these are all sort of different flavors of the same thing um bing.com is also chat GPT it's gpt4
42. 6:26it's free so if you go to bing.com click on the chat button that's gpt4 you can generate images there you can uh
43. 6:36read images there like you can upload an image to it and tell it to describe it to you or write advertising copy about
44. 6:42that image and it will do it uh it's free it's connected to the internet it's quite remarkable what you get to do for
45. 6:49free there uh gpt4 at open AI sites 20 bucks then another free one this one's brand new this is Claude 2
46. 6:59um and it's anthropics competitor to gpt4 it's got a hundred thousand token context window meaning you can upload
47. 7:07about I don't know somewhere between 50 and 75 000 words into your prompt into your chat so you could take your great
48. 7:15American novel pop it in there and start editing it with that thing uh or whatever it's amazing and then uh
49. 7:22poe.com has a bunch of different models and bard.google.com is getting an honorable mention right now it wasn't on
50. 7:29the list for a long time because it sucks so bad but it's gotten better um and and it's they're on their way
51. 7:37it's it's still not great but it's they're on their way so that's that and then if you really want specific tools like something for
52. 7:46music or video or things like that futurepedia.io is a big list of AI tools so with that let's get to the questions
53. 7:53let's go to the videotape I'm a physician I did AI research 30 years ago who's advancing the medical front
54. 8:08um I think lots of people are advancing the medical front I mean the the thing about about the AI World AI ai's been around
55. 8:16for like you say decades the the thing that's different now is um since the Advent of the Transformer model that Google introduced in 2017
56. 8:27that's the thing that's been driving all these um innovations that we're playing with today so what the Transformer allowed uh
57. 8:37researchers to do is train these large language models on massive data sets so they trained them on you know kind of all of the internet
58. 8:46um and so these tools are just in incredibly powerful you don't have to bring your own data you can you know
59. 8:51augment your data with these large language models um so I think there's lots of people working on medical you know the people
60. 8:57that have been doing it before but now you're going to have a whole rash of people doing things on top of it Google
61. 9:04just announced three days ago a thing called Med Palm let me get make sure I got the name right which is a large language model that's
62. 9:13been trained up on medical data specifically I think it's Google Google Med Palm yeah Med Dash Palm so if you go to
63. 9:27sites.research.google slash medpalm a large language model from Google research designed for the medical domain
64. 9:41and mm-hmm [Music] she's trying to look at when it came out that was May 11th they had a they had a
65. 9:55new announcement this week oh medpom too here we go no that's April 13th anyway go look up medpom
66. 10:09there's here's July 8th okay this says this is from The Verge Google's medical AI chat bot is already being tested in hospitals the Mayo
67. 10:20Clinic has reportedly been testing the system since April okay so it came out in April and it looks like you know it's
68. 10:28it's in the world Mayo Clinic there you go you know Clinic's doing it hey man really important I gotta stick
69. 10:44oh Joe Black you're cute I'll work itself out it'll be tasty all right please do it the daily summary at
70. 10:54the beginning middle and end of The Tick Tock uh okay the I would I didn't I was busy doing a bunch of other stuff today so I didn't
71. 11:05do a ton of research on I didn't see any major new things it was a big end of last week so I don't have I don't have a
72. 11:14big update today on on news so I mean I can talk about what I talked about last week but I don't even have the URL still up um [Music]
73. 11:33[Music] no I don't I don't have the uh I don't have the article still up so I'll get back to him um Leonardo came out with a new model
74. 11:46check out leonardo.ai if you're not on it um get on it get on the wait list and then apparently with leonardo.ai if you
75. 11:56get on the wait list and go to their Facebook Facebook group and sign up for their Facebook group they'll immediately
76. 12:03Grant you access uh to leonardo.ai um but it's it's a very very good site uh that is built on top of stable diffusion
77. 12:13um and it's easier than mid-journey you don't have to go to Discord it's just a website uh and it's quite good and they give you
78. 12:22150 credits a day and it and it's about 18 or 20 credits to produce four Images so you know you can do it for free you
79. 12:34can do a handful of images and then if you pay for it you can do lots Lots [Music] I don't like Bard at all I hated Bard
80. 12:51for a while it's at least tolerable now and it's you there's some interesting things you can do there
81. 12:56um you can it can read images just like bin Ken um one of the guys in the AI Salon that I that I run um uploaded a sketch of a website like
82. 13:06they did in the in the open AI demo when they when it when they announced gpt4 he uploaded a sketch of a website and
83. 13:16said build me a website you know that looks like this and it did and one of the guys in the salon did that with
84. 13:23bard.com so he uploaded a sketch of a website and it actually wrote the code um to make functional buttons on a website
85. 13:31the T-shirt today is just some sort of album [Music]
86. 13:46a lot of butter jokes going on in the comments here I approve [Laughter] is somebody eating popcorn is joining anybody figure out a way
87. 14:02around the internet being disabled I think that might be a local issue I think we'd heard about that on the news
88. 14:14let me see uh well let's wait well glass Bing Let's go ask Bing I'm gonna try a new thing today too um being go to Bing chat
89. 14:30and [Music] has the internet been disabled [Music] all right Bing's gonna tell us
90. 14:46and I'm in more precise mode so it definitely won't hallucinate I'm sorry but I don't have any information about the internet being
91. 14:55disabled globally it's possible that there may be local issues but all right we're okay it's okay Bing
92. 15:05told us the internet has not been shut down we're good we're good for at least an hour what are you talking about Moody
93. 15:11I have I have no reference for what you're talking about x dot AI uh what about it it's they made some announcements it's that's Elon musk's
94. 15:20new competitor yeah they'll go up there when they have something to play with the reason I have these there these are
95. 15:24all things that I think anyone who hasn't played with you know these tools you know might want to experiment or
96. 15:30start with and there's a bunch of free options there claude.ai is is a definite true competitor to chat GPT it's a bunch of X openai people
97. 15:41um and that that site's actually quite powerful the the language stuff isn't quite up to the the um the quality of gpt4 but it's it's quite
98. 15:52good it's quite powerful and the fact that you can you know have you have a a hundred thousand token context Windows huge
99. 16:03uh what is your go-to prompt for getting the AI to act like an expert in the specific field so the reason I have go to prompts.chat
100. 16:11that little second URL from the bottom prompts.chat go there that is a whole list of exactly what you're asking for
101. 16:21act like a an accountant act like a screenwriter act like a comedian act like an idiot um I don't you don't have to get super
102. 16:33fancy with those generally what you want to do is you want to give the thing a roll so I want you to act like this kind
103. 16:41of role which kind of Narrows the context a little bit and then I'm going to be giving you you know some things to accomplish or you know
104. 16:51here's the things I want you to do and then you give it the specifics and that set of three contacts who it is you know the role it plays um
105. 17:02what is what its General tasks are the milestones and then what what the specifics are and then you get into
106. 17:09specific back and forth prompting um the other thing you can do is very often if you if you prompt it and just say I want you
107. 17:18to act like this that and the other thing and just hit return it'll just start talking to you so one
108. 17:23of the things you can do is you can give it these little contexts and say I don't want you to answer yet I just want you
109. 17:29to let me know if you understand and it'll say yes I understand and then you say okay here's another piece of information you know
110. 17:36you know I don't want you to do anything yet I just want to make sure you understand so you can give it a set of
111. 17:41instructions and make sure that it understands all of those and then start to get into asking it questions doing
112. 17:48things like that so that's that's kind of how I approach it the other the big shift for me with large language model prompting versus
113. 17:57Google searches because they look similar but they're not Google wants you to put as concise a prompt in there as possible right I just
114. 18:09want this thing and they're going to give you back that answer these large language models they need a lot more context because because
115. 18:16they're the the way they the way they do what they do they're pulling from a massive Data Corpus um not that Google's isn't massive but
116. 18:24what it what it's returning is a list of links uh which is you know from an index so it's a very different thing than
117. 18:30here's all of the words in all of the languages in all of their different semantic meanings and I'm going to start doing
118. 18:40probabilities on the words so the more you the more context you give to these tools the better they get
119. 18:48um and so that's something that I had to relearn is that it's much more like a conversation with a very simple person
120. 18:54that's a genius like this is these are like idiot savants they need a lot of Direction they need a lot of focus but
121. 19:02then they know everything so I'm gonna put on the air it's getting a little warm in here it's getting a little warm in the learning lab
122. 19:21um [Music] what up homie how are you I'm doing good Doug Lee what kind of interesting stuff are you doing in here real quick math in 2022
123. 19:37IBM divested and spun off their Watson Health division yep why did IBM Watson fail I don't know the I don't know probably I mean I I would
124. 19:48think that there's a couple of things behind it Watson Watson was machine learning technology that was pre-transformer
125. 19:55um everyone that I know that that was working in AI you know for the past decade working with working with old modalities basically
126. 20:08um are a little Blown Away by what these tools are doing so it it you know there might have just been some of the tech might have not
127. 20:17quite been there it might have been data set I don't know I don't know the details of it um but the the the healthcare space I do a lot
128. 20:30of work so my company's called storybine we do a lot of work in healthcare and Pharma and it's such a regulated
129. 20:35environment and hospitals in particular are so resistant to change um it takes it takes such a different
130. 20:44mentality to innovate in that space and it's really hard and it's really slow so it might have just been a thing where
131. 20:50they weren't getting the traction they needed and they said hey we just don't you know this isn't you know this isn't
132. 20:56delivering the level of results we want so let's just spin it off I find it ironic that what was it four months ago
133. 21:03they they announced Watson X which is which is a large language model it's a it's a rebranding of Watson it's
134. 21:11a reintroduction of of Watson going we can do AI too right so Chachi PT the reason we're seeing all these
135. 21:19competitors and the reason Elon Musk is starting a new company and the reason inflection just raised 1.3 billion
136. 21:25dollars all of this stuff is happening because chechi PT went from zero to 100 million users in six weeks um Sam Altman didn't expect that
137. 21:34Microsoft didn't expect that Google certainly didn't [ __ ] expect that um so so Chachi BT being such an immediate hit changed everything so
138. 21:45um one of the things that I heard about IBM which which is uh fascinating is that they created an AI Center of
139. 21:53Excellence across IBM you know they have 150 or 200 000 employees they have a lot of employees um they have a thousand people at IBM
140. 22:02writing prompts for this stuff for this new stuff um so everyone's everyone's head in this direction why because these tools are so powerful
141. 22:15um you're live awesome thank you very much I do appreciate that uh yeah if you're new here uh pop any
142. 22:26questions you have down below I'll do my best to answer them where do you find all these AIS I don't really have a life
143. 22:38I'm running my company and I kind of spend every waking moment these days trying to figure this stuff out and it's
144. 22:45you know I'm not so so Matt wolf who does futurepedia.io there's like 3 700 last time I looked it was 3 700 AI tools there
145. 22:55um I've given up on following any of the open source things I've given up on following any of the Tactical specific
146. 23:03tools like you know here's the music generator here's the other than Runway ml I'm not really [ __ ] around too much with video um
147. 23:13um like all the specific tools you know specific tools for writing copy or writing doing real estate marketing or
148. 23:19building websites I'm just I just go to futurepedia.io I'm not paying attention to any of those where I'm putting my attention is the
149. 23:28core large language models so the the large companies building these large language models that all these other
150. 23:34applications are going to be built on because these are going to be the engines that drive everything else
151. 23:41and I just like it for me it feels very important to understand the nuances between these what the different things are good at
152. 23:51um and and try to read the tea leaves about what's coming and what are the implications like for example if you haven't played with
153. 23:59code interpreter at chat GPT oh sorry just ate dinner um if you haven't played with code interpreter chatgpt um it it is
154. 24:13mind-numbingly powerful it's it's insanely powerful and I know nothing about data analytics but I can see that what that is is you
155. 24:23know a superpower right for anyone with any kind of data which is essentially any business these days right um
156. 24:31so I'm spending my time trying to look at what these tools are what they do what makes them unique like Claude
157. 24:38having a hundred thousand word context window to put that in context Chachi BT currently is a 4 000 word context window
158. 24:47which means it's got like this 3 000 word sliding window that is its memory for your chat session so if you've ever worked with chat gbt
159. 24:57and all of a sudden it seems like it forgot the early [ __ ] that you were talking about that's because it forgot the early [ __ ]
160. 25:03you were talking about because that early [ __ ] popped out of that sliding context window right um Claude has a hundred thousand token
161. 25:13context windows so it's 75 000 words right so it's a lot bigger than chat gbt so even if the writing capability of Claude
162. 25:25isn't as powerful as gpt4 the fact that you've got this massive context window is a big big deal and it's a reason to use it
163. 25:35um and and just all that sort of stuff you know barred from what I understand is better at code um chat GPT now has code interpreter
164. 25:44which is just insanity if you if you watch it do what it does I'll show you it's it's a it's an amazing thing so anyway uh and I find
165. 25:58them on uh I just pay attention to Twitter I follow people on Twitter um I find people like Andre carpathy
166. 26:07um who's the one of the founders of openai and he he posts a lot I find people like Ethan Malik who's A Wharton
167. 26:15Professor who's been doing amazing posts about all this stuff and and so I go look at who they follow on Twitter and I
168. 26:22follow them so I just start following people who are who who the people that I respect are following and within about
169. 26:3115 minutes you can have a pretty solid list to follow on Twitter and then just pay attention to it look at it two or
170. 26:37three times a day and go play with stuff and start to get a critical eye like you know understand the strengths and
171. 26:45weaknesses of these things one of the things we did the other night on the live someone asked can these things
172. 26:49write a joke and I was like yeah probably not like it ain't gonna be funny but let's go see so I came up with
173. 26:56a joke premise and we told it to act like a stand-up comedian or comedy writer and write a joke about this
174. 27:02premise about how avocados you know are not ripe and suddenly they're overripe right um and we and we went to all these
175. 27:10different tools and had it try to write the joke and they were all bad except for gpt4 was the only one that actually wrote a joke
176. 27:18I mean Bing wrote a joke but it was as if Microsoft wrote it it was really bad um I was referring to chat gbt Moody
177. 27:28where was that before what were you saying before just joining anybody figure out oh oh about chat GPT being dumb was it down
178. 27:39today code interpreter was down yesterday or the day before but that's just because they're fixing stuff here I am a chat gbt oh yeah I'm
179. 27:49gonna instead of typing instead instead of uh forcing you to watch me type which I know is painful I'm starting to use my
180. 27:55microphone on my Mac write me a poem that explains the concepts of quantum mechanics like an Eminem rap [Music]
181. 28:10it's going yo let's dive into the deep end my friend quantum mechanics ain't no child pretend all right it's working
182. 28:21it's going it's doing its thing it was probably just a hiccup there I know they're they're working on a lot of stuff over at open
183. 28:29AI right now I I have this strong strong suspicion that there's a lot of uh legal conversations going on over there
184. 28:39they cut a deal with the Associated Press on Friday and they cut a deal with Shutterstock on Thursday and they pulled
185. 28:46down browse with Bing the connection to the internet last week uh because openai was stealing content from behind
186. 28:55paywalls and I think they got a call from time Inc that said hey you're uh you're sharing our content that people should be paying for
187. 29:05and that's bad so I I just I think there's a lot of there's a lot of uh holes being plugged right now is my sense
188. 29:18but the good news I mean the the good news about things like them cutting a deal with the Associated Press means
189. 29:24that we're going to start to see specific flavors of things like chat GPT like imagine a mode in chat GPT where you can restrict
190. 29:37the the sort of world of information it's looking at to the Associated Press so I want to I want to learn about
191. 29:44Middle East peace but I only want it to pull from Associated Press you know documents um we'll start to see things like that
192. 29:54um and I I think that's gonna that's gonna make things really really interesting is Fran Drescher your president
193. 30:10um no I have a degree in acting and I've written seven screenplays but um I understand why they're striking I empathize with it um
194. 30:24I think some of the AI issues are they can fight it but they can only fight it for so long I think it's inevitable that
195. 30:34what these tools are going to do to the creative arts to writing to acting to voice over um to post-production to pre-production
196. 30:46uh to marketing to thought leadership to Consulting to everything um are are going to be incredibly disruptive and and ultimately Innovative
197. 31:00but almost every business model is is gonna is gonna be forced through through um a new Gauntlet of innovation so just like when when
198. 31:15um when the streaming company started streaming content and weren't paying for any of it and then they cut deals with
199. 31:22the studios and did all that sort of stuff and they were shitty deals but like they're not going back because
200. 31:27you're not going to uninvent streaming same same with this stuff so while I empathize with what they're fighting for
201. 31:33and they should absolutely fight to get as much as they can um you know like the the one thing that the
202. 31:41studio is saying you know we're gonna do something groundbreaking with AI we'll pay people for one day scan their
203. 31:46likeness and then use that likeness for all of eternity and never have to pay them again isn't that awesome and the
204. 31:52actors are like [ __ ] no but what the studios are gonna do is go okay if you don't want us to pay you for
205. 32:00a day then we'll just synthesize them entirely like we don't need you like this we're good we'll just go make it
206. 32:05mid Journey so so I don't know I don't know what's gonna happen it's crazy traffic a lot of people need to learn this isn't
207. 32:15wait wait what why do people need to learn this isn't it pretty self-sufficient why do people need to learn this this stuff
208. 32:27isn't it pretty self-sufficient well it is and it's not if you've got a so if you're if you're an engineer if you're a
209. 32:34program or a coder and you understand how to think you know in in logical expressions that's what you're doing here you're
210. 32:43just doing it with the English language or whatever your native language is people that are either that are either
211. 32:50not writers or not used to thinking in in logical expressions don't know what to do with this stuff right it's just it's just a search bar
212. 32:59right so you just type in you know um what was the score of the you know who who won Wimbledon and Chachi BT goes
213. 33:08my latest information is September 2021. why do you ask and like people are like well what is this thing it's it's useless or they put
214. 33:17in you know um do this math calculation and then it does it wrong and they're like well this is stupid it lied to me
215. 33:26you know write me an opinion piece on this thing and give me the citations of the Articles you used to come up with
216. 33:33that opinion and then the articles are fake articles that don't actually exist you're like well this is [ __ ] so so
217. 33:44people that have not been playing so so here's the thing here's what this channel is about I'm not a machine
218. 33:49learning guy I'm I'm an entrepreneur I've started 12 or 15 companies I um my jam the thing that really lights me
219. 33:59up is once technology gets to the point that the rest of us can use it that's when I like to innovate
220. 34:06I don't like to create the technology so if you've been in machine learning for the past five years or 10 years or 20
221. 34:14years this [ __ ] seems trivial to you but it's trivial in in a in an interesting kind of way so in the early
222. 34:22days of the World Wide Web the internet had been around for decades and who was using it well it was mostly
223. 34:30researchers and scientists and computer science people and programmers and engineers and universities and then Tim berners-lee invents this
224. 34:40thing called the World Wide Web where you can just make a couple of tags and you make this thing called a page and
225. 34:45it's got blue underlined words on it and if you click on those they jump you to some other page anywhere in the world
226. 34:51that was revolutionary because you just had to do this so all of a sudden a thing went from command line you have to be a geek to
227. 34:59use it to I can use the internet click click click click and that transformed everything but it wasn't until
228. 35:08it it made Escape philosophy escape velocity beyond the Geeks and said hey rest of the world there's this thing
229. 35:16called the internet you can click on want to click on it oh cool oh now I'm looking at something that was written in Africa oh now I'm in
230. 35:26in China now I'm in Japan now I'm in Texas oh that's cool this is exactly like that it's been around for decades
231. 35:34machine learning neural networks AI all this [ __ ] tensorflow Transformers Gan networks it's been around chat GPT
232. 35:48is this it's the equivalent of this was you know World Wide Web to the internet that's the difference and so why people
233. 35:56need to use it is they're like what the [ __ ] is everyone talking about why are they so excited about this
234. 36:02and and you know what I the the thing the thing that really gets me excited is what is when I'm in the presence of
235. 36:12people that have what I call the Kevin McAllister moment right from Home Alone it could do that and so that's why I put these up here so
236. 36:23if you haven't played with chat gbt if you're just sitting on the sidelines and heard that the robots are going to kill us
237. 36:29and you've seen too many tick tocks of people going you can make money and they annoy the [ __ ] out of you they should
238. 36:37it's those people are jerks but it doesn't mean you shouldn't use these tools because these are the most
239. 36:44powerful tools ever put in the hands of mere mortals and so go there and start playing chat.openai.com is the official chat GPT
240. 36:56website prompts.chat this little bottom one right here is we'll teach you how to talk to these things give you a little primer and just well
241. 37:06if you know if this sounds interesting to you and maybe you do want to make money with chat GPT you got to start somewhere here's the
242. 37:14deal you're not as behind as you think you are and it doesn't require nearly as much technical expertise as you think it does
243. 37:23it really is just talking with words and asking for remarkable [ __ ] and it does it um and there's a whole bunch of issues
244. 37:34around it that we can talk about but if you have questions pop them down below and I'll do what I can to answer them why don't you stream
245. 37:42acting as an AI why don't you do a stream acting as AI I might do that I know there's a couple of tools that that
246. 37:49can do it live I know Tick Tock can do it live um I've I've thought about that for me right now I don't know a thing a thing that
247. 38:02actually is interesting to me is is doing some sort of fully automated persona that grabs news does education does something
248. 38:17does entertainment um but I haven't quite been inspired what to do yet because it's the one of the problems with this Tech is it's very
249. 38:27easy to get seduced by the tech and think that the tech is the strategy the tech isn't the strategy these are just
250. 38:34tools right so so when someone really does ask the question well how do you make money with AI it's the same answer as how do you make
251. 38:42money with Photoshop I don't know you [ __ ] solve a problem good enough that someone's willing to pay you money for it that like that's
252. 38:50all of business is someone has a problem you use some tools to solve it or you're this tool right to solve it and then they pay you money
253. 39:01that's how this works you use this and you came up with a clever t-shirt and you did print on demand print on demand is the tool you
254. 39:10know this is the copywriter and then you stuck it on Tick Tock and said look at my cool t-shirt well now
255. 39:16you can do that with AI and you can make 100 T-shirts instead of one okay that's how you make money with chat CBT you
256. 39:23know but but that's not making money with AI that's making money making people laugh and having a cool t-shirt that they want to wear right
257. 39:31so so when I make the money when I make the t-shirt for you can make money with Chachi vitae it doesn't matter like you know it's
258. 39:43about AI but it ain't right all right um where are you reading these questions well class X here's the deal I read all the questions and
259. 39:56um it annoys the [ __ ] out of the younger people because they're like because I'm not reading the questions live at the
260. 40:01bottom I'm just sort of reading down through them and so I just I just get behind so I am live and class X if you're still here I'm
261. 40:09reading your question now so I'm reading them up where they turned gray Google just added a great feature for
262. 40:17searches too it now has generative searches so it will give you the specific answer really is that it barred um let's see
263. 40:36um
264. 40:56hello [Music] uh hello hello [Music] um
265. 41:12what are the top business um canvas books in the past
266. 41:26five years take a look at that oh yeah that's interesting
267. 41:40so it looks like what it did I asked that what are the best business canvas books of the past five years
268. 41:49business model generation book it put an image in there of it value proposition design oh this is really good
269. 41:58yeah so this is starting to get multimodal so they're incorporating search results into and then oh that's
270. 42:03nice you've got Google it down here right so I can now Google it let's see what that does so that pops me over to a normal Google
271. 42:15search that's smart that's a that's a really smart approach right so I can have my kind of Standalone chat GPT or
272. 42:23this is more like Bing Bing like experience where where Google is now better than Bing is Bing does lots of citations but
273. 42:31they just do it as like little you know numbered citations um here they're actually putting book covers up here um let me see let me see
274. 42:41um who let's see who are the top five uh business cartoonists in the past decade let's see if it
275. 42:58brings up any cartoons from the cartoonists let's see oh it does look at that Tim Fishburn Stephanie Pyro Dave ramund
276. 43:13Mike Luke Vic and Jim Boardman with examples huh give me 10 more yeah this is impressive
277. 43:27I told you guys I said so so for months Google has not been on my list of you know things you should go play with
278. 43:34because for months Bard sucked really bad but one of the things I said was don't count Google out because they got
279. 43:42buckets of money they invented this [ __ ] and they got a bunch of really smart people there yeah this is really really clever and
280. 43:52then what happened are any of these linked oh that's interesting this is linked to just a New York New Yorker article ah
281. 44:03that's fascinating that's really cool all right Bard's getting there Bard hallucinates a lot though that that I do know right now
282. 44:13so if you don't know what hallucinating is that is when it lies to you when it lies to you a lot like a dog it lies like a dog
283. 44:26very cool thank you for sharing that Tyler that's uh that's pretty neat yeah that's that's new like that wasn't it
284. 44:32didn't do that yesterday it might not have even done that this morning maybe it was this morning and that's that's
285. 44:39the other thing about this technology like it's being developed in real time like we're getting access to these tools
286. 44:45and they're they're literally like the engineers I can't imagine what it's like to work in these companies they've got
287. 44:51to be you know coding 18 20 hour top hours a day um because this is this is insane it's major major stuff going down
288. 45:06um how do you know if your AI answer is violating copyrights um for the most part they're not because the nature of how the tools work is
289. 45:17they're writing your answer word by word and it's just a probability engine now if you tell it this is the the the
290. 45:25essence of Sarah Silverman's lawsuit is if you ask for a specific quote I forget what she did I think she might have
291. 45:33started the beginning of of a paragraph in her book and it completed it and completed the rest of it so in that case it's pulling content
292. 45:43from something it was trained on but most of the time if you say I want you to write me you know uh an opinion piece
293. 45:51on the value of generative AI for writing marketing copy it's it's going out you know across all of what it's learned and it's just
294. 46:02probable probabilistically putting one word after another that has the highest probability of answering your questions it's not
295. 46:11actually copying and pasting answers from different things right so if you purposefully try to get it to quote
296. 46:20something if it has that thing it will do it very often it will tell you it's giving a quote but it's lying it's not
297. 46:26even the real quote so that's something you know where it might look like it's something that's copy written but it's actually not um
298. 46:36so so most of it is is original content that's been trained on all this all you know all this other stuff
299. 46:46I wouldn't worry too too much about it because these tools are going to be rolled into like like chat gbt is
300. 46:53getting rolled into all of office 365. Bard is getting rolled into all of uh Google's office docs you know docs
301. 47:01and sheets and all that sort of stuff so these tools are going to be ubiquitous they're going to be spread throughout it
302. 47:07and it's you know it's not this similar the the way these things create what they create you know when you're when
303. 47:13you're typing up an email in in Gmail and you write a sentence in like the three gray words come at the end of it
304. 47:21it does that auto complete thing at the end that's literally what these things are just at a massive scale where it can
305. 47:28write you a whole paper instead of just completing three words um but it is generating original content that's why it's called
306. 47:36uh the G in GPT is generative so it's generating original content based on pre-trained information that's the p pre-trained
307. 47:48using Transformer technology which is basically that's the model that they use to be able to train it on the entire internet so
308. 47:58um but if you're doing this for a you know for your company I would get your lawyers to give you an opinion on where the boundaries are
309. 48:10because I ain't a lawyer um but I do know how they work they're not copying and pasting copywritten material they've been trained on
310. 48:20copyright material and lots of non-copyright material and lots of blog posts and Wikipedia articles and
311. 48:27websites and idiots talking on Reddit you know it's all the [ __ ] right is what's in the Corpus and then depending
312. 48:37on your prompt how much context you give it you kind of narrow it down into specific semantic Gardens
313. 48:45and then you know the more specific you get it you probably run a higher risk especially if you use someone's name
314. 48:52you know finish this paragraph from Sarah Silverman's book called this and it will do that because it's
315. 48:59somewhere in the Corpus you've narrowed it down now to a point that it's it's got a high likelihood of giving you
316. 49:05something that's copy written I'm disappointed with tonight's shirt Choice well at least you can read the
317. 49:14letters like like I I don't disagree like all my good shirts are hard to read text again so so I mean what about the hair the hair it's
318. 49:25it's a problem never mind it's a we've got issues we've got we've got fashion issues we've got a coughing situation
319. 49:37Kyle did you pay to use chat gbt yeah I pay for chat EBT and I pay for Poe and I'm gonna I'm gonna unsubscribe to Poe
320. 49:46because it no longer has the advantages it used to because Claude came out so the reason I paid for Poe is that you
321. 49:56could play with Claude 100k um context window and now that's available at claude.ai thank you hair growth geek for the for
322. 50:07the rose I appreciate it I appreciate it I appreciate it amazing but a little bit scary yes totally agree wait till I show
323. 50:14you uh wait till I show you a code interpreter if you haven't seen code interpreter who just mentioned Quantum Leap meta
324. 50:24Victory must be another gen xer Quantum Leap references out of the out of the archives [Laughter]
325. 50:42hey my cuz O'Donnell that's a Rosie O'Donnell reference meta victory meta Victory saying I look like Rosie O'Donnell
326. 50:52I'm getting Rosie O'Donnell the guy from what is that Family Modern Family uh skinny John Goodman fat Alec Baldwin uh
327. 51:06uh what's his name uh I don't know it doesn't matter screenwriters fighting AI is a losing battle I do fear so I've written a bunch
328. 51:20of screenplays so I feel for them but I also I mean here's the thing that's kind of [ __ ] up this is the most powerful writing tool
329. 51:31in the history of humanity as a writer as a marketer as a screenwriter why would I not use the most powerful writing tool in the
330. 51:42history of humanity I remember I remember wanting to write a screenplay um about um oh I forget her name it was Mary
331. 51:54something she was the first female pirate they ended up making a movie about it with Gina Davis um and like a decade before that I had
332. 52:05an idea to write a screenplay about this this female swashbuckling pirate that pretended she was a man um
333. 52:13and it was from a like with ADD I'm horrible at research right so I went to the library and I got books and I read them and I took notes and
334. 52:25that and it was just it was such a for me it was such a burden to do the research that I ended up not I didn't
335. 52:33even really finish the outline because I realized it was just to do it well I would have to basically be a dramaturg
336. 52:40and I just don't have the brain for that well with these tools now I could do all that research in an instant right and and you know I might
337. 52:49need to go check is the is the research that it did accurate right but now that we've got things like Bard uh and Bing
338. 52:56that are connected to the internet that I can check those citations and I can just say hey tell me a tell me five good
339. 53:04stories about that particular character in history oh that's good tell me more about that one tell me about who else
340. 53:10was on the ship tell me about where they were going it can just give me all that stuff it can just feed me so if I'm if
341. 53:16I'm a screenwriter and I want to quickly Knock Out Concepts why would I not use this right and that's not what they're
342. 53:22fighting for what they're fighting for is hey Studios don't use this to write a shitty screenplay instead of using me to do one
343. 53:30and if the studios are smart they won't what the studio should do is go you know give us more screenplays faster so we have more options
344. 53:42feel free to use these tools and give you know give me a bunch of different options because that's what
345. 53:49exact film Executives love to do is they love to get in there and act like they're creatives and muck with the [ __ ]
346. 53:56well use the most powerful writing tool in the history of humanity to help you knock some [ __ ] out I mean we're telling stories here
347. 54:09what are stories language what are these things called large language models large language models what are screenplays language
348. 54:20what is business language what is empathy language this is like these are Humanity amplifiers they're amazing so yeah so I feel for
349. 54:34them and it's yeah here's a metaphor for you the The Writer's Guild resisting catch
350. 54:49EBT and the and the Actors Guild resisting synthesized people is the equivalent of Florida declaring
351. 55:01legally with a proclamation that they're no longer accepting hurricanes I mean it's it's it's a noble law great
352. 55:11awesome congratulations by the way one's coming like you know this this is a force of nature right this this can't be uninvented
353. 55:23the fact that chat CPT got so popular so quick and now literally trillions of dollars of investment are heading that way
354. 55:33it ain't stopping it's accelerating right so the storms the storms are Brewing right the storms here but it's like a tsunami where it's like
355. 55:45the you know the tide went out and it went way out and it's like oh let's walk out in the let's let's walk out into Tampa Bay
356. 55:53because there's no water in it that should be fun um so yeah yeah Florida is no longer accepting hurricanes I I mean that's
357. 56:03that's the equivalent right now so it's Noble and listen the The Writer's Guild and the Screen Actors Guild should get what they can get
358. 56:12um but everything's changing everything everything the nature of intellectual property is going to be fundamentally Rewritten
359. 56:27we're a few years from a world of infinite personalized content generation real time personalized content generation
360. 56:38I like that Jay-Z song but I wish it sounded more like Travis Tritt and talked about me here let me play that for you
361. 56:51that's coming eliminate hallucinations I pulled email names from a scan but it still had errors and no fillings yeah uh
362. 56:59so wait what's your name pickle piccolochio all right eight out of ten solid name I like it um yeah the the hallucination thing is
363. 57:09just something we're gonna have to deal with for a while so so any work that you do with these if you're doing them for
364. 57:15work and where you know accuracy is important check your work like your homework now is not to not use these
365. 57:21it's use these and then make sure it didn't lie especially if you say a lawyer can you help me run GPT for non-profit
366. 57:31experiment I've been trying to run yeah tell me about it and I'll I'll maybe we can do something live here Bing asked me
367. 57:37three times to be his girlfriend that's really funny uh Rachel Studio what was I'm curious what you were talking about
368. 57:45well that was the The New York Times uh um reporter had a four hour conversation with Bing it said its name was Sydney
369. 57:53and then it said it didn't just ask him to be uh her boyfriend it asked him it said he was in love with him and and he should leave his wife
370. 58:03if for Sydney and then uh Microsoft lobotomized old Sydney they shut that [ __ ] down pretty quick once it made the
371. 58:12cover of the New York Times what do you think is coming relative to AI how do you see General businesses using them um [Music]
372. 58:34it's a great question I don't have the answer but what I can tell you is this what these things do is they can write
373. 58:48what a human would write about a hundred times faster now if you give it a really shitty prompt it'll write about 40 percent as good a
374. 58:59hundred times faster right it'll give you crappy stuff faster if you get better at prompting it'll give you like eighty percent of
375. 59:11what a human would written a hundred times fat or it wouldn't be a hundred times faster if you're giving it a good
376. 59:18well-constructed high context prompt it can get to about eighty percent of like what I would write as my my first draft
377. 59:27in maybe 30 times faster not 100 times faster but 30 times faster so what's happening is businesses that are aggressively leaning into using
378. 59:37generative AI are seeing 30 to 40 percent productivity increases with an increase in quality right so 30
379. 59:47to 40 percent faster with an increase in quality so it's as if they hired better people better copywriters that might
380. 59:56charge more or and not just copywriters like think any kind of knowledge work it could be project management it could be
381. 1:00:03copywriting it could be ideation it could be Consulting it could be talking about HR issues it could be how do I
382. 1:00:10have a tough conversation one of the things we've done on these lives is um I want to quit my so someone wrote in the comments um
383. 1:00:19how do I press professionally tell my you know current employers to go [ __ ] themselves right and so we did a little
384. 1:00:26prompt for that and and came up with a professional it was a really nice professional letter that basically said
385. 1:00:31you don't listen to me and you know go [ __ ] off um so so all of that stuff if if the business environment let's just say just
386. 1:00:45um abstract it up completely and let's take the the lower end of that let's say there's let's say that on average
387. 1:00:54companies that Embrace this stuff see a 20 increase in productivity with the same staff and a 10 increase in quality either that company
388. 1:01:10can let go of 20 of the staff because it's being 20 more productive some companies will choose that right
389. 1:01:18especially the ones that are invested in by private Equity firms that are just like slash and burn it [ __ ] it let's
390. 1:01:24make more profit they'll just be like get the Smart Companies are going to say hey we're seeing a 20 increase Here Without
391. 1:01:34Really having to do [ __ ] with our same staff why don't we upskill our staff to use even more of these tools and
392. 1:01:41reinvent some of our processes because if we could take that from a 20 increase to a 40 increase either we could take on you know
393. 1:01:51two times or five times more clients or we could lower our rates or we could increase our value so that we retain our clients longer term
394. 1:02:01so so at the sort of highest abstraction level what what you what you move from is you move from today where you've got
395. 1:02:08this many people producing this much productivity to the productivity starting to do this and and and then and then what starts to happen
396. 1:02:19so so here's where the disruptive [ __ ] comes in the the world that I know really well is the agency world
397. 1:02:27so one of the ways that agencies make a lot of their money right now and and the holding companies that own you know lots
398. 1:02:36of the big agencies and lots of agencies their model is based on time and materials so a client hires them to do
399. 1:02:46write a bunch of social media content come up with the strategy do all the marketing for this particular product
400. 1:02:54and they have a big Bullpen at the agency full of copywriters and artists and they come up with the strategy and
401. 1:03:04then they sort of dump the strategy into the creative teams into the bullpen and the more hours that those people
402. 1:03:13spend writing copy and making pictures the more money the agency makes they're charging 120 150 200 an hour for people that they're paying
403. 1:03:23some amount less than that so the agency could take advantage of those efficiencies but that's actually
404. 1:03:34contradictory to their business model because in their business model the more hours we spend the more money we make so
405. 1:03:43why would we want something that does [ __ ] faster we wouldn't but the clients also are using this [ __ ] so I've talked to two major
406. 1:03:55pharmaceutical clients in the past week that have confided in me because I was talking about this stuff have confided in me
407. 1:04:06our agencies don't get it we're asking them deliver faster and charge us less money and everyone we're talking to is like no
408. 1:04:19no we're good this is how we do business so that's going to [ __ ] some [ __ ] up because what's going to happen
409. 1:04:29the client's going to tolerate it for what maybe a year till the next budget cycle goes around or they're going to tolerate it until a
410. 1:04:4010 or 15 person agency that's all in on AI that can turn around in a week what an agency can turn around in two months
411. 1:04:51they can turn around the same level of stuff same level of quality some aggressive AI agency is going to be
412. 1:04:58able to turn around in a week what a traditional agency turns around in two months at 10 times the cost because they
413. 1:05:05have no incentive to drop their cost they want more hours that's going to [ __ ] some [ __ ] up and every single business category is going
414. 1:05:20to face when you have a productivity Spike like that there's it's just it's gonna [ __ ] some [ __ ] up
415. 1:05:31so so in the end it's good it's we're increasing Prosperity but there's going to be jobs lost there's
416. 1:05:37going to be disruption there's going to be business models they're going to be completely blown up sagin the wga the the Screen Actors
417. 1:05:46Guild and the writer's Guild right now are basically saying save the business model the business model is over the business
418. 1:05:56models changed already the contracts haven't caught up to it yet but it's this is this is a this is a storm that's coming so I don't know
419. 1:06:14it was a broad question to say you got a broad answer sorry about that uh seen you do it amazing uh Ethan Malik
420. 1:06:21is great love his post yeah he's awesome um do all these companies own the usage rights to AI responses yeah I'm pretty
421. 1:06:28sure they do at some point here's the deal okay you know who's happy right now do you know who's pre-ordering their Yachts
422. 1:06:36the intellectual property lawyers are [ __ ] ecstatic the lawyers are gonna make a fortune off all this so everyone's gonna Sue everyone
423. 1:06:46so yeah Germany comedy that's funny Yadi using the mic thank you Shelly although I couldn't use it in bing for some reason that was weird
424. 1:07:00was it a big oh no in Bard Bard I couldn't use the mic can I use it here oh yeah there it goes [Music] yeah it's there okay
425. 1:07:18okay how do you use your mic so you type in your content been trying to figure out how to do that I'm on a Mac so there's a
426. 1:07:30mic button F5 is a mic button and I think you have to enable it and you probably have to enable it for the browser and then it looked like I
427. 1:07:40had to enable it there for that site and so it everything is more complicated than it should be but if you want to play with something cool
428. 1:07:49if you're on iOS for if you have an iPhone download the PI app pi as in 3.1415 and it's got a mode where you can talk
429. 1:08:02to it so it's it's a it's like chat GPT but you talk to it it talks back you pick one of six voices and then it's connected to the Internet
430. 1:08:12so I was asking it like what are the big concerts in Denver this weekend it's like well Taylor Swift is doing one and
431. 1:08:18she's on this big tour and people seem to be loving your concert it's talking back to me so that's a fun one but on
432. 1:08:25the Mac it's uh I forget what it's called the talking thing have there been any reviews of the
433. 1:08:40Taylor Swift concert in Denver [Music] I'm on bing.com so it's working for me now so yeah that's better than me typing that's less painful
434. 1:08:55let's there have been reviews the concert took place on Friday thousands of fans crowded the stadium I drove by it it was insane
435. 1:09:06Swift's Talent discipline and cinematic Poise was far more dazzling then confetti Blaster fireworks could ever be oh my God
436. 1:09:18this this AI writing is just it's a little insufferable right now with spectacle and emotion tightly braided
437. 1:09:25Swift Mega dose the faithful at the sold out show with songs from across the catalog while gracefully navigating
438. 1:09:33costume and set changes against a massive LED screen would you like any inform any more information I'd like you
439. 1:09:43to be a better writer I could have said that responds to us a smarmy little troll thanks for your feedback I'm always
440. 1:09:56working on improving my writing uh all right love the Bed Head yeah the glue the glue broke it's yeah sorry about that
441. 1:10:09Skynet just everything screenwriters fighting ai's a losing battle Napster exactly this the shit's coming
442. 1:10:16out of touch to the max hi Kyle Ohio I listen again I have deep empathy like I've written screenplays I have a degree
443. 1:10:25in acting like you know I know people that are you know deep in that world so I feel bad I mean strikes strikes in
444. 1:10:37a business where I forget what the the Actors Equity when I when I was acting in New York the unemployment rate for
445. 1:10:46for Actors Equity which getting into actors equities relatively hard right you have to get a professional show to be in it um
446. 1:10:53had like a 96 unemployment rate like you know the the the Arts is not you know not everyone's Leonardo DiCaprio in terms of their revenue
447. 1:11:04you know I think the average Screen Actors Guild you know annual salary is you know probably 20 grand or something
448. 1:11:11like that on the average Equity salaries probably more like eight I don't know what it is but actually let's ask let's ask Bing that
449. 1:11:19okay I'll talk it I won't type it can you tell me the average salary for an Actors Equity actor comma a Screen
450. 1:11:35Actors Guild actor comma and a writer's Guild screenwriter [Music] [Music] and that's cool what it did was it took
451. 1:11:51my little prompt and it actually is doing three different searches that's kind of cool this is Bing that I'm on right now
452. 1:12:01yeah annual average annual pay for Actors Equity is 26 000 a year that's for the ones that are working and it's like you know
453. 1:12:09higher than 90 unemployment rate um Screen Actors Guild 45 to 60 Grand a year but most sag actors earn less than
454. 1:12:19five grand a year and most sag actors are not working and then the writer's Grill Guild spec scripts range from 72 Grand to 136 Grand
455. 1:12:30the average around 100 000. yeah I'm gonna say and what are um the oh wait I forgot to talk it dang it okay
456. 1:12:44and what are the employment rates [Music] uh [Music]
457. 1:13:08employment of actors is expected to grow eight percent about seven thousand openings for actors are projected each year so 7 000 jobs
458. 1:13:26for 70 million people that want to be a star it's it's a rough business so so listen you know I support them going for what they're
459. 1:13:38going to go for but I don't know you just they're feeling reactive and doesn't feel good right now Hollywood
460. 1:13:48should everyone become an investor on blockchain for movies blockchain needs to get his [ __ ] together blockchain [ __ ] up
461. 1:14:04you know blockchain [ __ ] up black blockchain needs to sit in the corner for three years but here's one blockchain's coming back you ready
462. 1:14:14yeah I don't think it's a bad model to invest in movies you know using that kind of investment um nfts and blockchain and crypto I'll ask
463. 1:14:26to go sit in the corner because they stole everyone's [ __ ] money because they're a bunch of crypto Bros and then
464. 1:14:33AI comes along it's going to be doing its AI [ __ ] and the politicians and the consultants for the politicians are going to start using
465. 1:14:43AI to do all sorts of misinformation and deep fakes for the 2024 election and then people are going to lose their
466. 1:14:51[ __ ] minds and go we've got to do something about the misinformation this is serious now and I mean it this time
467. 1:14:59and then the blockchain community is going to go um we've got tokenization that could help what if we verified humans on the
468. 1:15:07blockchain and verified what they wrote that it wasn't modified and it actually was a New York Times article and those
469. 1:15:15sources actually were those sources we could do that with nfts would you like some we have all the technology so
470. 1:15:25so I think AI is going to drive adoption of all that [ __ ] I don't know what is going on with my hair O'Donnell Rosie O'Donnell come
471. 1:15:41I want to control it not control us yep I agree with that the fear that these things are going to control us these are our tools
472. 1:15:49and and one of the reasons that I love pointing people in the direction of just go start playing get curious start
473. 1:15:56exploring start thinking of this like you're on a [ __ ] Adventure is because if we don't it will be done to us
474. 1:16:06right and not the robots will do it to us other human beings using these tools will do [ __ ] to us so the more people that learn how to use
475. 1:16:18these tools and have that power we can bend it to the light bend it to the light bend it to the light because if we just sit back and
476. 1:16:28wait and judge it and say it's not perfect and you know what I was saying is I went on there and made a mistake I hear the
477. 1:16:35robots are going to kill us you know you know what I told that AI why don't you go [ __ ] yourself if we're just doing that the whole time
478. 1:16:42then it's going to be done to us and that's not a feeling I want to have I use earbuds I was on a live with a developer and he
479. 1:16:53used chat GPT to fix something on the web 3 wallet or something yeah you can do things like you can write smart
480. 1:16:59contracts it can analyze smart contracts and tell you where their vulnerabilities are um yeah it's pretty slick and I would
481. 1:17:06imagine putting smart contracts in the code interpreter where it's actually writing python code and and doing things
482. 1:17:12like that there could be some really interesting things to play with there do you think the best questionnaire would make wait
483. 1:17:21would make the best of this keep do you think the best questionnaire would not a victory I don't know what that sentence means
484. 1:17:35uh is it missing punctuation do you think the best questionnaire would make the best of this keep the best oh well I don't
485. 1:17:46understand the question so maybe if I understood the question smart to Smart to get ahead yep it's important to get ahead if your talent is
486. 1:17:57limited to allowing the capturing reflected light they are trying to sell buggy whips yeah K Sanders I I mean this thank you for
487. 1:18:07the game controller that's a new one I haven't seen that before um this is one of those things that um foreign
488. 1:18:19like I don't know what the future of Cinema and entertainment is and it you know part of it kind of looks like
489. 1:18:26whatever the [ __ ] the metaverse is going to turn into like it isn't going to be Mark Zuckerberg's metaverse because he
490. 1:18:33obviously [ __ ] that up you know you spend all that money and he made something that looked like second life from 2001. I'm like
491. 1:18:42yeah maybe we could come up with something that's [ __ ] not horrible um so yeah the the only the the the place I
492. 1:18:57that's been spinning in my head for the past month or so is just this phrase of radical personalization that that
493. 1:19:04entertainment becomes radically personalized to each of us individually and so then what does that do for for
494. 1:19:13narrative where you've got a director telling you know his or her story you know in the following genre with the
495. 1:19:20following script and the following actors when you can take that basic thing and you can swap all those elements out
496. 1:19:28and say I want to watch that movie but I want different actors in it and I want it in a different genre and you can just watch that movie as
497. 1:19:35[ __ ] insane as that sounds and far out as that sounds we're not that far from that so who become who are the movie makers
498. 1:19:46that are crafting those personalization engines that understand narrative and understand performance and nuance
499. 1:19:55and all that and find ways to capture that stuff in a creatively um at a creatively high level right
500. 1:20:08like in order for that [ __ ] to be good it can't be like a shitty College film it's got to be like the Hollywood
501. 1:20:17Blockbusters we know and love right so and maybe that's more like 20 years away maybe maybe five years from now we have
502. 1:20:25the shitty version of this because it's just technologists figuring it out and then over the next 15 years after that then the artists start
503. 1:20:34reintegrating into the technology and you know who becomes the Steven Spielberg of the hyper personalized entertainment thing I don't know
504. 1:20:47I guarantee it's someone that's [ __ ] around with this stuff right now how are you innovating using AI right now I'm using it
505. 1:21:01um I've generated some content generators within my business I've got a video technology business and
506. 1:21:09one of the things that I'm doing is I'm I'm creating an omni Channel content generation engine so for every video in
507. 1:21:16my system I'm generating content around it like a whole world of content around it I'm using it for creative development with our clients
508. 1:21:27I'm using it for efficiencies in things like writing emails and summarizing meeting notes um I'm radically rethinking
509. 1:21:41so so my company's 11 years old it's called storyvine and it's about human beings telling stories right it's a
510. 1:21:49story-centric authentic storytelling platform automated video editing and it's about people telling stories
511. 1:22:03and I I feel like we're really good at that like we understand it we understand what it is but one of the challenges we have is
512. 1:22:11getting people to tell their stories well what if I didn't have to get people to tell their stories what if I could synthesize those but
513. 1:22:18what if I could do it in a way that that it was more authentic than the other [ __ ] that was out there that's that's something that's maybe
514. 1:22:25interesting to me or what if I could take what if I said hey would you be willing to share your story sure because everyone says sure
515. 1:22:37but then when it comes time to really record their story they're like well does my hair look good oh my God my hair
516. 1:22:43is a disaster what do I do oh I don't I don't think I talk good and I don't want to do it so what if I just said
517. 1:22:49fill out this question there and answer these give me some bullet points of your story and then um just record a 30-second
518. 1:23:01a 15 second hi my name's Bob and I was diagnosed with this disease in this year and I take that 15 seconds of video I
519. 1:23:11synthesize Bob I synthesize Bob's voice I take the bullet points of his story I turn that into a compelling narrative
520. 1:23:19and then I make a video of Bob where Bob didn't have to say [ __ ] and it looks like a storybine video but
521. 1:23:27I don't have to you know break my back or the client doesn't have to break their back or the agency doesn't have to
522. 1:23:33break their back to get people to tell their story and then what if I could do that instead of 10 stories with real people I could
523. 1:23:43generate a thousand stories that are relevant to a thousand different audiences and it gives the agency a massive amount of organic content
524. 1:23:55so yeah and you know I've done some experiments and and we're closer to that than I thought we were I'm pretty close to that
525. 1:24:12hyperlink was the leap yeah the hyperlink was a big [ __ ] deal IRC chat there's there's a there's a 1995er watch the Quantum Leap
526. 1:24:22um I tried everything to get access to dream boot is booth it wait dream Booth is it really difficult to get up and
527. 1:24:31running yeah yes Robert Robert Smith 333 listen so the the image right there that little eyeball with the chin on it that's
528. 1:24:39that's one of my favorite dream Booth images I tried nine times to get stable diffusion and dream Booth
529. 1:24:50up and running and I I failed nine times and this wasn't like nine times in one night this was like over a series of
530. 1:24:59like a month and a half because I didn't know what Google collab was I didn't know how to run it you know I would go on to Reddit and
531. 1:25:06they're like here's here's an automatic 11-11 you know Google collab notebook and you just all you do is go here and
532. 1:25:12push the play buttons and I would go there and I push the play buttons and then one of them would give me an error
533. 1:25:19and I wouldn't I don't I I wouldn't know what to and it took me nine tries until I finally got it running and then I
534. 1:25:26finally got it running I was like yes um so yes so so here's here's the deal Robert um if you know Discord mid-journey is is
535. 1:25:38really high quality and and it's just some relatively simple commands in Discord to do it um it's the best one from a quality
536. 1:25:45perspective stable diffusion is by far the most um flexible because it's open source and the develop the developer
537. 1:25:53Community is just going nuts on it so there's all sorts of cool things it does but the thing I would recommend if you
538. 1:25:59want to use stable diffusion is go get on the wait list for um leonardo.ai which is a web front end it's it's like
539. 1:26:08automatic 11 11 but you just go to leonardo.ai you sign up for the wait list and then you go to the Facebook
540. 1:26:17um group for leonardo.ai sign up for their group and I think they fast tracked you into the wait list it's
541. 1:26:24really good here I'll show it to you so this is leonardo.ai so it's got across the top here they just added a
542. 1:26:33bunch they just did an upgrade update on this it's like version three um so across the top these are different
543. 1:26:39models that you can use so like you know this one's sort of like a cartoony kind of thing this one is like ice and match
544. 1:26:47isometric sci-fi building so you pick one of these and then you can anything that you do within it is going
545. 1:26:53to kind of look look like whatever the style of the model is so there's some cool characters there's a pixel art one
546. 1:27:02isometric fantasy illustration that looks like someone did it in Adobe Illustrator here's paper Art here's
547. 1:27:09vintage photography so it's got a bunch of these selected models you can also fine tune your own models so these are um
548. 1:27:20like this is one I did on my photography I I've done some you know landscape photography with dramatic clouds out
549. 1:27:27here in Colorado and so I took a bunch of my my best photos and I put it up there and so now I can generate images
550. 1:27:34based on that so this is this is basically dream Booth but without dream Booth you just upload some images
551. 1:27:40um it does in painting and out painting it's doing texture generation for 3D um it's got a Community Feed of you know
552. 1:27:49here's all the images that people have created and like any one of these you can click on see what prompt they used
553. 1:27:55you can either copy The Prompt some of them you can just go use the model that they used so it's really really good
554. 1:28:05so I would go play with leonardo.ai and I'm not being paid by leonardo.ai so I don't give two shits if you go use
555. 1:28:15it but it's a hell of a lot easier to install than stable diffusion automatic 1111. um [Music]
556. 1:28:29but thank you for sharing that you were having difficulty with it because it like stable diffusion is powerful but
557. 1:28:35man is it a [ __ ] show if you're not a technologist I mean this this is one of the one of the frustrations with all new
558. 1:28:41technologies is that especially if you're a creative and you have a creative mind like to what you know to what level of
559. 1:28:53technical proficiency do you need to get to be able to be creative right and if you if you're not good with
560. 1:29:00virtual machines in Google collab because you haven't been programming for the past five years then it's a [ __ ] pain in the ass
561. 1:29:15all right amazing and a bit scary yes you make it even more exciting I don't know what it is but thank you yes I've been
562. 1:29:23researching that I'm just kidding you have a great sense of humor Kyle thank you Joy he does appreciate that
563. 1:29:30he was kind of funny but looks ain't everything [Laughter] Hayden seemed to take a joke he can see him take a joke yeah you know
564. 1:29:40we pick on him he calls it trolling I don't know what trolling is like the troll troll under the bridge it's like
565. 1:29:45troll under the bridge he's like that but he doesn't he doesn't he doesn't seem to get too upset about it but I
566. 1:29:51I'll tell you what I bet he does I bet he hangs up from these lives he probably just goes and cries because that's
567. 1:29:56that's what I would do I would just go and cry they talk you know what they said they said you look like Rosie O'Donnell I I
568. 1:30:02just I think that's rude I think it's rude he thought it was funny all right very true merely Immortals 80 20. don't
569. 1:30:12know what that means but okay yes 80 20. mere mortals we are I had one downloaded and it I guess went away so you
570. 1:30:22recommend one what do they cost Tam the obsession by the way uh seven and a half out of ten on the name I like it
571. 1:30:30but I don't know what your question means I had one downloaded and I guess it went away what was the one you had downloaded
572. 1:30:38and which so I'll recommend something if I know what you're talking about do you mean one of these you don't have
573. 1:30:46to download any of these all these are just websites you go to and Claude is free Bing is free chat gbt has a a free level Poe has a
574. 1:30:58free level part is free so those are all free and you don't have to download anything Kyle did you pay to use chat EBT yes I
575. 1:31:07pay for chat CBD could be largely because of access to their new features like code interpreter how about building a virtual automation
576. 1:31:17bot on this channel okay we can do that I'll do a zapier it's the problem with doing a zapier on this channel is it'll get really boring
577. 1:31:25really [ __ ] fast because zapier is really janky but maybe what I can do is I'll go show you one that I've already
578. 1:31:30built and I'll talk about kind of the building blocks I'll talk about it conceptually and and get you in there
579. 1:31:36um and if you if you want to see a bot that I built go to content evolution.net and there's a thing on the home page of
580. 1:31:46that site called The Challenge engine and it's it's for business people it's it's content evolution is a like a
581. 1:31:51Federation of marketing companies and I'm the chief generative officer there first Chief generative officer in
582. 1:31:59all of the world as far as I know huh but um if you if you go play with the challenge engine what it is is it's a Google form
583. 1:32:11where you put in your emails so we can send you an email back you put in your email your name and a business problem you're having and
584. 1:32:19then you answer a few questions about it like what does it cost you like what's the impact of this problem and what
585. 1:32:25would your business be like if you improved it all that sort of [ __ ] and you answer those questions you hit
586. 1:32:31submit and then 30 seconds later you get back essentially like a Consulting report that says hey if you're trying to
587. 1:32:37solve this problem here's three questions in six different categories like strategic customer engagement
588. 1:32:45branding you know you know you should you should know the answer to this question that one now and so so we kind
589. 1:32:50of give back this Consulting report in 30 seconds based on your input so that's a thing that I built so if you want to
590. 1:32:56go go play with that and see what it does and then I'll I'll show you a smaller version of that um that I that I was I was working on
591. 1:33:05last week with a couple of guys so we can do that um how do you force AI to stop making up info when automating a task like
592. 1:33:15transforming a table okay so large largely you you you can't 100 but there are things that you can do to make it
593. 1:33:30better if you're using Bing Bing's got three modes creative balanced and precise use precise it'll hallucinate less it'll
594. 1:33:40make less [ __ ] up um the most important thing to do to reduce hallucinations is to reduce your context
595. 1:33:51or or actually expand your context which will reduce the areas of knowledge that it will draw from so so here's here's
596. 1:34:04all right let me give you a little bit of kind of large language model 101 like like super high level Theory so the way these tools work
597. 1:34:16they use this technology called Transformer technology and they can basically dump massive amounts of data into this machine
598. 1:34:26and so the massive amounts of data that open AI decided to dump into this machine is all of the public internet so
599. 1:34:34they spent years tagging finding data sources making sure that they were in in appropriate whatever structures or
600. 1:34:43formats to be able to be ingested by these machines so GPT 3.5 and gpt4 the underlying large language model is essentially trained on
601. 1:34:53as much of the public internet as they could get their hands on so it's a lot it knows everything so here's kind of how it works the the
602. 1:35:05metaphor that I've come up with is this imagine that you're in a helicopter tour over the Grand Canyon and the Grand Canyon is filled to the
603. 1:35:14rim with all of the words of the internet and and every each little sub Canyon is a little semantic grouping of words meaning this
604. 1:35:24you might have the word dog well what does the word dog mean well the dog is the fluffy thing that you pet at night
605. 1:35:30and you know take out to pee and poo well but a dog could also mean the guy that did his girlfriend wrong he's a dog
606. 1:35:39oh that's a completely different meaning so so when these machines go out and look at all the words they're looking at
607. 1:35:45all the words in context of how they're used so so there's these semantic clouds basically so think of these little sub Canyons
608. 1:35:54so if you type in a really generic um prompt like make me a table comparing this concept and that concept your
609. 1:36:04helicopter is really high above the Grand Canyon that can see all the words right and so it can see all of the words that are
610. 1:36:14accurate about the concept that you want to learn about and it can see all the words that are inaccurate about it the
611. 1:36:19the example that Andre carpathy used when he talked to the Microsoft developers conference was he said if you're asking chat GPT to solve a
612. 1:36:28mathematical equation and and and you don't have the the enough context to sort of limit those semantic regions
613. 1:36:37he said you're kind of looking at the entire Grand Canyon he didn't use Grand Canyon he's like you're looking at the
614. 1:36:42entire Corpus and within that Corpus you've got the professor who solved that mathematical equation correctly you've
615. 1:36:50also got the 30 students who solved that mathematical equation incorrectly So when you say just solve this equation it
616. 1:36:58kind of you know pulls from all of the places where someone's talking about that problem it kind of melds them
617. 1:37:04together in some sort of probabilistic average and just types one word after another based on all of the Corpus that
618. 1:37:12it has and half of that [ __ ] or you know ninety percent of that shit's wrong and so that's going to hallucinate more
619. 1:37:18you're going to get more mistakes in that but if you tell it to act like a Stanford Professor that that specializes
620. 1:37:25in this branch of mathematics and you know and is obsessed with with never talking about equation unless it's right
621. 1:37:33and I'm and I'm going to tell you to work on this equation and I want you to show your process all of the stuff that
622. 1:37:40you do to give it more context what you're doing is you're kind of lowering that helicopter from seeing all of the
623. 1:37:47Grand Canyon down into one of those sub canyons and potentially down into a micro Canyon a little path where there's
624. 1:37:54just there's there's not that much for it to pull from and it's got a higher likelihood of pulling the right [ __ ]
625. 1:38:03and then still check your work because here's the other thing that that is not obvious about this stuff because it's doing word by word
626. 1:38:13probabilities it's got the the intellectual awareness of what it's doing of a calculator so if you say to a calculator hey calculator
627. 1:38:25how do you know how to do math that's ridiculous right it doesn't it's just Pro like it's circuits know how to do math right so
628. 1:38:37when you type in two plus two and hit equals it makes it four and it just knows how to do that that's kind of what these large language
629. 1:38:44models are doing it's saying based on the prompt you wrote the next most probable word is whatever word is at this mathematical
630. 1:38:54Vector space in this thousand dimensional word cloud with all of the words in all of their meanings and it's whatever word is at Vector 342
631. 1:39:09comma six eight nine point three two seven eight point four two equals dog and then it types the word dog it doesn't know it typed the word
632. 1:39:18dog it just knows that mathematically the word dog is the one that's next and then it does that for every single word
633. 1:39:24so what it presents to you looks like it knows what it's talking about but it it's not going back and looking
634. 1:39:34at what it wrote and going is this right so so it literally doesn't know what it wrote which is counterintuitive because it's
635. 1:39:44these things are so sophisticated that it's uncanny how much it seems like well holy [ __ ] it knows what it wrote right nope
636. 1:39:55oh that's it's doing copywritten stuff no it's just it's written words probabilistically in a way that looks like what
637. 1:40:05Shakespeare would have wrote written so I hope that helps does your AI address you no you know who addresses me Tick Tock because they're
638. 1:40:16stupid [ __ ] little letters I can't put some sort of covering like like a gray area underneath the letters so if
639. 1:40:23I'm wearing a shirt that has nice patterns on it then I can't read the words and everyone complains I can't
640. 1:40:29read it and we're all it all makes us very fad my wife does my dressing for me my my um my fashion um strategy is what's on top
641. 1:40:42like of the of the closet or of the door of the drawer just pull it out but now I have to pick dark shirts so we can all read the the
642. 1:40:53little words on screen um I haven't tested the Claude AI how good is it compared to gpt4 it's it's different the the writing of Claude is
643. 1:41:03not as good but it's but it's solid it's it's good it's like GPD 3.5 good or better um it's not GPT for good but
644. 1:41:11you can upload up to five like PDFs and interact with them it's got a hundred thousand token context window so you
645. 1:41:21could literally copy and paste a 50 000 word novel into your prompt and start interacting with it I'm disappointed with tonight's shirt
646. 1:41:29Choice ah I love my trolls you're the best the first one is great don't know what that means Google search seems to not have
647. 1:41:42pagination anymore infinite scrolling oh that's interesting yeah Google listen man Google is uh I can't even imagine the conversations
648. 1:41:53going on there A disruption you know they invented the thing that allowed open AI to exist and then they sat on it for seven years because
649. 1:42:08they're like you know we probably shouldn't make a thing that completely invalidates our entire existence for being
650. 1:42:20so now they're trying to figure out okay what is search how do we make money when our old model was predicated on
651. 1:42:30people clicking 40 times and now no one has to click anymore because we give them the answer that's a that's a problem
652. 1:42:39and what's fascinating is Microsoft doesn't have that problem because they've got all their software they've got their operating systems to
653. 1:42:47sell they've got Office 365 they've got Dynamics they've got their cloud services so Microsoft has an entire Suite of software applications
654. 1:42:58that they can just charge for chat gbt Within you know almost all of Google is Advertising based on clicking they want you clicking more
655. 1:43:11and they invented a thing that doesn't make you click at all got to be some fascinating conversations going on
656. 1:43:25I'd like to be a fly in those beatings actually I wouldn't that old dog dog dog dog dog dog dog dog dog
657. 1:43:36I would say most everyone could benefit from code interpreter because of the file use yep totally agree
658. 1:43:43code and interpreter super cool oh Tick Tock just jumping to the bottom again it's one of my pet peeves does anyone
659. 1:43:50here work at Tick Tock hey Tick Tock you're listening to this right you're transcribing this in real time
660. 1:43:56and doing a psychological evaluation of me in real time yes and everyone in here could you [ __ ] fix your scrolling bug
661. 1:44:04thanks that'd be awesome so while I'm scrolling to figure out where I was in my questions I like answering all the questions and then if
662. 1:44:16I'm scrolling when someone makes a comment it jumps me to the bottom and I lose my place and then I have to go back and it makes me cranky
663. 1:44:24so and you're like why don't you just go to the bottom I don't know because I'm 58. and I'm cranky and I'm bitter and I
664. 1:44:37you kids get off my lawn you sons of [ __ ] don't you tell me how to tick tock Marge I'm on a tickety-tock again they're making fun of me
665. 1:44:51I don't look like Rosie O'Donnell do I tell me I don't look like Rosie O'Donnell all right this is like a Taylor Swift concert except
666. 1:45:09the talent and the entertainment slightly smaller crowd but we have our fun
667. 1:45:24I feel like I've been here more time [Laughter] oh lordy lordy I'm coming Martha
668. 1:45:39what's that from Sanford and Sons I'm coming Martha you remember that God that was a great show all right does your AI address you I
669. 1:45:46found I found I got back to the troll I was at before okay back to New questions and I know I haven't started zapier yet
670. 1:45:55I'll show you zapier I gotta get in the mood for it I gotta figure out how I want to do it so it's not boring as [ __ ]
671. 1:46:01um they code with chat EPT who comes with chat gbt what are you talking about Big D I would say most everyone could benefit
672. 1:46:11from code interpreter okay that's where we were um do you have to pay for a code interpreter you do it's at
673. 1:46:17chat.openai.com you pay the 20 bucks a month for GPT uh chat gbd Plus and then you have to go into your
674. 1:46:24settings and turn on beta features and you have plugins and you have code interpreter turn those on and then you
675. 1:46:30have it and then just upload if you if you if you're like me and and do not have a history in data analytics which I don't um go to kaggle
676. 1:46:44kagle.com and it's just a big pile of data sources and so you can go find like crime stats for your city or
677. 1:46:54um IMDb TV show statistics for the past decade and then just download it and even if it's in zip format this is a this is a
678. 1:47:02good thing to know about code interpreter so you can upload one file right so sometimes when you download
679. 1:47:08data sets sets they're in like a dozen single CSV files so I was thinking like oh I've got to find a way to get all
680. 1:47:15those csvs into a an Excel sheet with multiple workbooks nope you can upload a zip file to code interpreter so you
681. 1:47:23could have a zip file full of 100 files so as long as it doesn't overload the I think it's 100 megabyte uh maximum
682. 1:47:30so you can just upload an archive.zip up to code interpreter it will write python code to unzip it and then it will look at what's inside it
683. 1:47:40and then it will go analyze all that data it's it's just [ __ ] insane it's outrageous try identity code interpreter I have
684. 1:47:48I'll show it here in a second I use chat GPT for schoolwork about to finish my bachelor's awesome it's great
685. 1:47:55for rough drafts yeah so Brandon that's what uh what I'm using it what what my experience consistently is with these
686. 1:48:01things is I get to 80 like that I get to 80 in five minutes writer's block is a thing of the past ideation is no longer
687. 1:48:11intimidating and like I like ideation but you know sometimes if you're like I gotta come up with six ideas for this oh
688. 1:48:18God I guess I gotta erase the [ __ ] off the Whiteboard which means I got to take a picture of their whiteboard which
689. 1:48:23means I gotta put it somewhere and I can't ever find my whiteboard pictures again so maybe I need to go get a
690. 1:48:29whiteboard picture organizer maybe I need to find a thing that like you know how that you ever had that conversation anyone have ADD here
691. 1:48:36and now I don't have to do that now I'm just like I gotta come up with an idea oh yeah let me go to church EBT bang
692. 1:48:41eighty percent five minutes and then I'm off to the races um so it is it's great for rough drafts um was shaking wait for your younger AI
693. 1:48:56coming soon yeah exactly I need it you on fleek dude don't listen to the haters thanks I assume you on fleek is a good thing
694. 1:49:05but I'm an old dude so I think I've seen that in one of those here's what the young kids say let's ask a boomer about about what the
695. 1:49:16Millennials are saying I think is on fleek Millennials I think on fleek is uh gen Z right huh I am culturally relevant so could I have an AI
696. 1:49:30I could have an AI continuous stories from books for me Ed would be upset I don't know who Ed is and was I mentioning Ed before
697. 1:49:43but yes you could have you could have so there's a there's a really cool project called um GPT author
698. 1:49:54and it's a it's a Google collab notebook so you do have to use Google collab um which which if you've tried to
699. 1:50:01install stable diffusion you'll be familiar with um but what GPT author is is you give it your open AI key for the words and you
700. 1:50:14give it your stable diffusion key for for cover art and then it asks you a couple of questions like what do you want the
701. 1:50:21topic of the book to be I think I think it's set up to do fictions like fantasy fiction so you give it the topic and then you
702. 1:50:28give it the writing style and you give it the number of chapters and then you hit submit and it starts writing a book it starts out
703. 1:50:36writing an outline and then it writes chapter one and then it writes chapter two and then it writes chapter three and then it writes chapter
704. 1:50:43four and it does up to X number of chapters and then based on the book it wrote it generates cover art
705. 1:50:50and and turns it into to ebook format so you sit there and watch it and it's also cool it gives you the cost of each
706. 1:50:58chapter and it's like two cents for this one four cents for that one it gets increasingly expensive because I think
707. 1:51:05it's sort of building on it takes you know the outline in chapter one and then makes chapter two based on all that and
708. 1:51:13it takes all that makes chapter three so it gets more expensive as it goes down but it's like a dollar to write a book
709. 1:51:18so yeah you could write a book an hour automatically if you wanted to uh with the head of hair thank you for
710. 1:51:25all your input you're welcome code interpreter is good for basing future School responses based on zip previous
711. 1:51:31file work oh that's interesting yeah yeah you take you take a bunch of previous work that's one of the things
712. 1:51:38that these these tools are really good at so so here's an early frustration that I had and there's there's a way around it
713. 1:51:45you tell it that you you want it to write a poem and it writes you a shitty poem you're like well no but I like
714. 1:51:51poems that are like this and then you're like write a poem like T.S Elliott then it sort of does that well but then you
715. 1:51:57want it to write poems like you write poems but they're not like T.S Eliot and it doesn't know who you are
716. 1:52:02well it's really good at like you can take one of your poems and toss it in there or ten of your palms and toss them
717. 1:52:08in there and say write me a new poem about puppies in the rain um in my style of the poems up above and it's really good at that
718. 1:52:18so with code interpreter you can just pile a bunch of documents like I was I was actually thinking about this I've got
719. 1:52:24seven feature screen plays a one-man show and probably 15 to 20 like like partially completed screenplays or
720. 1:52:37outlines I could tell us all of that into a zip file and put it in code interpreter and basically start interacting with all the [ __ ] I've
721. 1:52:45written you know over the past couple of decades that's kind of fun calling me out Gen X I know I see you I see you
722. 1:52:56I know I know who's in here you look like Sammy Boyle Jackson just talk like him [ __ ] snakes on this [ __ ] plane I actually did
723. 1:53:07marketing for Snakes on a Plane that was fun I was working for uh I was working for a Hollywood marketer we did
724. 1:53:14snakes on the plane was one of our projects it was good I also got to do I did uh digital photography for one of the Quentin
725. 1:53:22Tarantino films the uh hateful eight and was down at Comic Con taking pictures you know in The Press Room that was fun
726. 1:53:30Kyle's approaching the end of our questions ask more uh oh I want to make a website to upload my screen my screenplay wait I want to make
727. 1:53:39a website to upload my screenplay and get coverage comps cast workflow thoughts dude okay toast post okay uh 8.5 out of 10 on the
728. 1:53:49name toast post it's it's solid it's close to a nine it's an 8.5 um go to claude.ai right now or or if
729. 1:54:00you if you're a GPT plus subscriber you can do it there with code interpreter but just go to claude.ai
730. 1:54:08you can upload files so just upload your screenplay or copy and paste the text of your screenplay paste it into the prompt
731. 1:54:16and then just say based on this screenplay write me a Hollywood log line you know you know give give me you know uh act act as a as a uh
732. 1:54:30um a Hollywood script reader right who writes coverage for high-powered agents so I want you to write write the uh I
733. 1:54:39want you to read the screenplay and write write some high covered you know some high scale coverage I want comps I
734. 1:54:45want cash breakdowns I want plot breakdowns I want I mean you could anything yeah go do it right now I did it um I did exactly that
735. 1:54:55three days ago it was it was [ __ ] brilliant there's a screenplay that I haven't thought about in 25 or 30 years
736. 1:55:07and it gave me the synopsis of it it gave me the character breakdowns and what happened to them it was like [ __ ]
737. 1:55:14it was just it's it's amazing it's it's just you owe it to yourself to play with these tools until you have this moment
738. 1:55:28and then once you have that moment you can't go back you can unsee it I'm gonna take the red pill are you gonna take the blue pill
739. 1:55:39we ain't uninventing it it's gonna [ __ ] things up for you whether you're using it or not so you might as well start using it
740. 1:55:52have you heard of AI cannibalism I read an article about it today um is that like the the model collapse
741. 1:56:00like if if you've got AI training on synthetic data that they're going to have model collapse that the AI basically it all becomes
742. 1:56:07self-referential [ __ ] um I that might be what you're talking about but I I haven't heard that term AI cannibalism but it sounds like that
743. 1:56:15model collapse kind of concept use your personal expertise to generate and know that you are creating quality
744. 1:56:21work with AI yeah yeah exactly the the you're thinking about this right Brandon all the stuff that you're saying tonight
745. 1:56:30feels very right to me like understand these tools like the best thing you can do right now in my opinion
746. 1:56:37in any profession no matter what you're doing is learn what these things make possible and start doing good work with them
747. 1:56:44don't get precious about like I I was on here one night I don't know a month or so ago and someone was like I feel
748. 1:56:51really guilty I'm using mid-journey to make patterns that I'm selling on Etsy and I feel really guilty I'm like why
749. 1:57:01if you're making patterns that you like right you're taking the time to learn how to use mid-journey well enough to
750. 1:57:08make patterns and you're probably making a hundred patterns for anyone that you're gonna sell you probably got a
751. 1:57:14hundred that you're not going to sell right so you're taking the time to generate a bunch of work pay for that
752. 1:57:21curate it make sure that it's truly seamless put a product page together sell it on Etsy why why feel guilty about doing something that just
753. 1:57:34I mean the guilt is maybe I know that if I were doing this with traditional tools that would take me eight hours or four
754. 1:57:43hours per pattern and with this tool it takes me 30 minutes so if it's good work if it's stuff that you think adds value to the world
755. 1:57:57who cares what tool you used if you were to cut out patterns with linoleum that would take you a week and
756. 1:58:07with Adobe Illustrator it takes you four hours and with this it takes you 30 minutes who cares it's a tool so yeah
757. 1:58:19figure out what you want to do figure out what's the best tool to do it and [ __ ] do it I just heard all the union
758. 1:58:26actors are mandated to third image and then they're done no that was what the what the what the producers Guild what the what the
759. 1:58:37Hollywood producers basically said to Sag is we've got a groundbreaking proposal and that's that's the words they used
760. 1:58:48and their groundbreaking proposal is for every background actor which is basically all the people that stand behind the people that talk
761. 1:58:56you know like the people in the cafe that are like pretending to talk and occasionally they'll look at the camera them
762. 1:59:03what the producers were saying is we have a groundbreaking idea we'll hire them for one day we'll pay
763. 1:59:10them for a full day and then what we'll do is we'll just scan them in and we'll never pay them again but we'll
764. 1:59:17use their likeness forever for any Project without having to pay them without having to tell them about it awesome right
765. 1:59:26and that's when sag went go [ __ ] yourself and went on strike produce is not so smart they could have positioned that a little
766. 1:59:37we Tad better because for them it's groundbreaking I mean the thing that's really [ __ ] up about it is that
767. 1:59:46The Producers could say well listen not for nothing we said we would pay them for a day but now we're just gonna go synthesize
768. 1:59:57them completely and we're not going to pay anyone so we're just so background actors are now just gonna be and and one
769. 2:00:03of the things that sag will probably fight for and probably get some some version of is um there's got to be some percentage of
770. 2:00:14real background actors in in a scene right so if you've got a scene with you know 30 people behind the actor maybe five of those have to be
771. 2:00:24real people and the rest can be synthesized or something like that like they'll probably get to some sort of
772. 2:00:29agreement like that but I don't [ __ ] know I mean it's it's just nuts it really is just nuts um you know any good stock to AI stock to
773. 2:00:40invest in no I'm I have a degree in acting don't listen to me for for investing advice I here's your Clues
774. 2:00:48every night I've been I've been doing these lives for two or three hours every night and I have I have a bunch of
775. 2:00:53companies listed here start there and then look at all the companies related to them um yeah I I think anthropic claude.ai
776. 2:01:04they're they're very impressive to me um I don't think they're investable right now because they're not public but
777. 2:01:11um I think Microsoft is kicking ass I think when when Microsoft rolls all this [ __ ] into their into their entire
778. 2:01:16application Suite that's really interesting Nvidia seems to have a near Monopoly on gpus um open AI inflection
779. 2:01:31um Google's on their heels right now but you know they got a lot of money a lot of smart people interesting just saw Brian Tracy say
780. 2:01:43number one cause for failure is focus on Tech versus customer Solutions so yeah this is this is a lesson I learned over
781. 2:01:51and over and I and I this is a mistake I see happening over and over and over again every time a new technology comes along it happens um
782. 2:02:00when I started agency.com in the 90s in the mid 90s a huge part of the conversation for like three years
783. 2:02:15was CEOs and you know big companies would come to us and like we need a website and what most people would say that
784. 2:02:25we're building websites back then you can make money with the worldwide web those people is they'd say sure we'll build you a website
785. 2:02:34website and then they build them a piece of [ __ ] website we never said we'll build you a website when they came to us
786. 2:02:40and said we want a website we would ask why why do you want a website well I was golfing with my buddy and he's got one
787. 2:02:48okay that's a [ __ ] horrible reason well but also we'd like to have better customer service okay now we're getting somewhere
788. 2:02:57you know what are you trying to accomplish as a business what do you need what do you need this website to do
789. 2:03:03who are your customers what are your business goals does anything over here fit with anything over there you know
790. 2:03:15they weren't all like that some of them were like we're gonna [ __ ] figure this web thing out and we're gonna do it
791. 2:03:22world class like we we worked with British Airways they were like that they [ __ ] had their [ __ ] together
792. 2:03:28that was the toughest client I've ever had they were they were insanely hard to work for but fair and they know what they [ __ ] wanted
793. 2:03:36so understand first what you're trying to accomplish and then you go which tool right you want images for social media
794. 2:03:47do you need AI or could you just go to canva and pull up one of the one of eight million [ __ ] templates that are
795. 2:03:55pre-designed and ready to go and they've got their own stock photography do you really need to go generate something here for that no
796. 2:04:04if you just want good high quality Instagram [ __ ] go to canva which by the way is all full of AI Nows
797. 2:04:13so you can automate the [ __ ] out of it but but don't mistake the tool for strategy this comes after strategy this comes after why
798. 2:04:28what are you trying to do and then look at this and if this can help you do it 30 times faster or 30 times cheaper
799. 2:04:39or 30 times better [ __ ] awesome and that's why people are so excited about it because when they actually
800. 2:04:45start to apply like the parlor tricks are cool write me a country song write me an Eminem rap make me a table write
801. 2:04:52me some code that's all fun and games but when you go oh wait a minute I'm a project manager let me see if this
802. 2:04:58thing's any good at project management holy [ __ ] that's the Kevin McAllister moment that's when this stuff gets starts to get powerful
803. 2:05:11I mean sure it's powerful to show your friends but it's really powerful when you say ha I wonder if if I have to
804. 2:05:19fundamentally rethink the value proposition of my 11 year old business that's what I've been doing for the past three months
805. 2:05:29and it was four months before that I was just trying to figure out is this AI stuff at all related to what
806. 2:05:36we do and the more I learned about it the more it became clear to me oh yeah it's like existentially related to what
807. 2:05:43you do you better [ __ ] figure it out so that's what I'm in the middle of right now stressful and exciting excited scared
808. 2:05:55The Genius of and you can have both did you answer my Hollywood blockchain question everyone investor I did
809. 2:06:06I said that I think blockchain's got to go sit in the corner for another two or three years and then it's going to get important um I
810. 2:06:16I think that um funding films using blockchain is fine I mean it's it's crowdsourced funding so there's that's been around for a bunch
811. 2:06:26of years um you know I I the the crypto Bros [ __ ] things up for a bunch of people for a bunch of years so
812. 2:06:39they just need to go sit in the corner and and let let that [ __ ] calm down I but I think it's a I think it's a fine mechanism um
813. 2:06:47there's a credibility problem there that like the the Hollywood film investing business is already a [ __ ] show so I
814. 2:06:55feel like blockchain and nfts complicate the [ __ ] out of it and it's just it's a it's I don't know yes it's possible yes
815. 2:07:04it's fine technology but they have they have a PR problem right now which is we [ __ ] stole everyone's money
816. 2:07:16and now we're starting another company and we just raised you know 15 million dollars for that or 50 or 100 whatever it was
817. 2:07:26um I use GPT to rewrite emails to sound professional and have received comments even I'm telling you these these these tools are
818. 2:07:34going to help us be better writers more empathetic all that sort of stuff I'd be in on the non-profit also I'm not
819. 2:07:43sure which one I talked about for work that is my first thought is how dangerous it could be um can be dangerous there's a thing
820. 2:07:52called alignment um go look at the there's a blog post from openai I'm talking about super alignment so open AI the makers of chat gbt
821. 2:08:01um are dedicating 20 of their compute resources I think it's over the next five years to a thing called super alignment
822. 2:08:10which is basically making sure that as these things get more powerful that they are aligned with our imperatives
823. 2:08:21because if these have a mind of their own like we want to be the most powerful things in the world and at some point
824. 2:08:28they go these things are getting in our way that's where it gets dangerous um but there's all sorts of other ways
825. 2:08:37it can be dangerous like the misinformation in the 2024 election here in the states is going to be a [ __ ] show
826. 2:08:43in the spreadsheet world this is about to take those that do not know anything to be somebody this will take those that
827. 2:08:50do not know anything to be somebody um yeah exactly is Kyle Frozen am I Frozen um yeah this is this is
828. 2:09:04this is like everyone gets to be a data analyst everyone gets to be you know everyone gets to be in some sense
829. 2:09:11whatever the [ __ ] they want to be because I don't have the talent to be an illustrator but I can make illustrations that
830. 2:09:20accomplish my creative goals now I could never do that before I can write but I don't have but I'm I'm not good at rewriting and I'm not
831. 2:09:31good at research well I am now I sort of know coding well now I can write full-on applications I don't know anything about investing oh
832. 2:09:45now I can create trading Bots that merge six different philosophical approaches to investing into a single trading bot just for fun
833. 2:09:57you couldn't do that nights ago I mean you could if you knew how to get into the opening eye playground but like
834. 2:10:06chat GPD that's what these things have opened up for everyone what about where creativity is concerned sag girl or essay girl um
835. 2:10:18where creativity is concerned everything's going to transform everything everything's going to transform there's
836. 2:10:27going to be nothing there's writing will never be the same illustration will never be the same um video animation 3D World building
837. 2:10:41character development 3D modeling architecture none of it none of it will be the same now in a lot of ways it's going to be better
838. 2:10:52like like the if you just if you just erase the transition so let's say there's a transition from now to five years from
839. 2:11:01now if we just jump to five years from now we're gonna be in a creative Renaissance the likes of which we can't even imagine right now
840. 2:11:11architecture is going to be through the roof fashion is going to be through the roof writing music film TV gaming
841. 2:11:22is is going to be on a creative level we can't even imagine right now and how you produce all of that will
842. 2:11:32almost in no ways resemble how you produce it today so what that means is we can't just jump to the future we just
843. 2:11:43can't jump to a new state where we know what the new jobs are and we knew we know what the new creativity is and we
844. 2:11:48know what the new Creative Expressions are so we have to go from okay musicians just barely survived the transition
845. 2:12:01from from Studios to streaming for well from Studios to digital to streaming and now they're entering a world where
846. 2:12:13fan generated songs might be more popular than the artist's songs why because we're in a new place where the rules are different
847. 2:12:25so every single creative discipline is going to go through some sort of radical transformation where the business model changes
848. 2:12:35the value of work changes what the work actually is changes the the creative and particular is a really interesting one because
849. 2:12:54there's a guy I mentioned on the live a lot guy named Paul rateser and he talks about the value of writing in his life he's a writer
850. 2:13:03and he says for me writing helps me think the act of writing an article about something allows him to process the ideas
851. 2:13:15and so for him it's the act of writing is that's where the value is but it's a certain kind of writing it's
852. 2:13:23the kind of writing on on topics that he wants to learn something about he's still writing by hand he's not using
853. 2:13:29chat EBT but all of the other writing that he does he's using Chachi BT so I think every creative is going to have to confront
854. 2:13:43do you get your personal enjoyment out of the creative product or the creative process and if you get your enjoyment out of the
855. 2:13:51creative process you're probably not going to use these things but you need to understand that you're going to be
856. 2:13:56competing in a world where there's other creatives that are more interested in the creative product and they're going to be able to make a
857. 2:14:05hundred of their creative product in the time that you can do one of yours because you're committed to the creative process
858. 2:14:12right so if the value is in the process keep doing it the way you're doing it keep doing illustration by hand keep drawing if you get your
859. 2:14:22the benefit for you as a human is is in the act of drawing itself keep drawing but if you're selling those
860. 2:14:32illustrations in a commercial setting and you could knock out a hundred of those in the time it could take you to draw one
861. 2:14:41that's a choice you got to make and there's going to be a lot of people where that compromise is soul crushing right that's [ __ ] up
862. 2:14:57but everyone's gonna have to face it because you're increasingly competing in a world where people that haven't been an
863. 2:15:05illustrator for 40 years are all of a sudden able to illustrate like you illustrate and the clients are just
864. 2:15:11going to say well I just need something good enough for this ad I mean I like you Fred but it takes you a month to make a drawing
865. 2:15:18and I need 10 next week can you do that no I enjoy the process sorry this guy over here does something that's close
866. 2:15:28and like you're gonna look at that guy's AI generated work and you're gonna [ __ ] hate it you'll be like that's
867. 2:15:32not anywhere near what I do it's not and that client needs the 10 drawings next week that's you know what does that mean I don't know
868. 2:15:44it ain't pretty but it's exciting like go to I had to leave Thelonious I had to leave to unfreeze Kyle am I still frozen I
869. 2:15:58I've seen screen Taps you're still hearing me right I'm not frozen yeah tap the screen like make the hearts go and yeah okay
870. 2:16:08I think yeah okay our old-fashioned expert system still around built with something like Pike I don't thanks thanks hair growth um
871. 2:16:18I don't I don't know Pike I don't know the old expert systems but these things are kind of meta expert systems so so
872. 2:16:26all sorts of expert systems like the one that's the the pet peeve of mine it's it's not really an expert system but
873. 2:16:33it's a very dedicated specialized kind of system our phone trees phone logic thank you for calling atnt press one
874. 2:16:47for the thing you're not interested in if you're interested in this other thing you're not interested in press two
875. 2:16:55for the thing you couldn't give a [ __ ] about press three for another thing that you hate and don't even want to think about press four
876. 2:17:05if you're a commercial vendor of a t press five get to my thing that's an expert system every single one of those should be
877. 2:17:16[ __ ] turned off tomorrow and replaced with a chatty little synthesized voice that knows everything that just answers the phone hey Kyle
878. 2:17:29how's it going last time we talked you were having trouble with connectivity with your Wi-Fi that all working okay no
879. 2:17:37it's [ __ ] up again oh I'm sorry to hear that have you tried resetting your router yeah I've done all the stuff okay so it
880. 2:17:45sounds like you might want to talk to a real person yeah that would be great okay I'll synthesize all of what we've
881. 2:17:51talked about the last four calls and I'm going to put you through to Sally now hi this is Sally so [ __ ] expert systems
882. 2:18:03if they're like if they're like phone trees um and that's you know I I think that people that have been in
883. 2:18:14machine learning and Ai and a lot of this stuff expert systems for a long time I think I think this Transit
884. 2:18:21transition is as profound for them as it is for all of us on you know on the on the non-technical side of this um
885. 2:18:32because rather than I mean if you think about the complexity of a phone tree I mean I looked at how to program a phone tree
886. 2:18:41for my you know seven person company and I I could not believe how complex it was like I can't even imagine what it would
887. 2:18:52be like to program one of those things to think about all the logic of all of the different scenarios and all the
888. 2:18:59different levels of authority that people have within an organization all of that has to be thought of and and
889. 2:19:09programmed in manually every logical decision the thousands and that probably hundreds of thousands of decisions in that phone tree
890. 2:19:23in traditional systems have to be programmed in and Along Comes this and you can just dump all of your corporate data into
891. 2:19:32the customer service reposit you know knowledge repository and this thing will just start talking it'll just start listening to what Kyle
892. 2:19:43has to say if you've programmed it cleverly it'll sound like Kyle or it'll sound like someone in Kyle's neighborhood
893. 2:19:53it'll remember the last time you talked because that goes into the repository too and it'll it'll have all the answers
894. 2:20:06and it will be smart enough to know if Kyle has a problem that is beyond its answers it connects it with a person so
895. 2:20:21so what happens to those people that program foundries like when that transition happens when it becomes clear that
896. 2:20:27that model has to go the way of the fax machine what happens to those people what happens to those systems what happens to those companies
897. 2:20:36many of them are going to be like Kodak and say no no phone trees are still very vital people love our phone tree we've
898. 2:20:45got a very different kind of phone tree we don't say press one we say interact with one it's a very different thing no it's not
899. 2:20:55so I don't know what happens to all those companies some of them are going to be replaced by a couple of little 23
900. 2:21:02year old snot-nosed kids that are like wouldn't it be cool if we'd have bought that just answered the phone and told
901. 2:21:06you everything about the company you needed to know and had the authority to make decisions and give refunds and [ __ ]
902. 2:21:13like that that'd be cool man and they'll get funded for 200 million dollars and they'll create that system
903. 2:21:20and that company that's been writing phone trees for the last 40 years will lose their 170 million dollar a year contract
904. 2:21:31with their customer of 25 years and they'll be like what just happened I think I'm going to go golfing and not come back from it
905. 2:21:47I don't know I don't know I'm just messenger I'm just a lowly messenger with bad hair and moderately humorous humorous acting skills
906. 2:22:03I could have been a contender agree they should take this chance to redefine a new paradigm new paradigms are good and all but
907. 2:22:12they're painful to get to Smart contracts are replacing a lot of Ip lawyers yeah that's it I I think the smart contract thing is really
908. 2:22:20interesting and the the Dows conceptually all the blockchain and nft stuff the Dow stuff the the fintech
909. 2:22:28stuff is fascinating they just they just kind of screwed themselves the biggest issue I've with AI and marketing wait
910. 2:22:34the biggest issue I have with AI and marketing because it's over the top language geez yeah I know so Silver Fox
911. 2:22:41one of the things you can do there's a phrase that you can use I saw on a tick tock and I tried it and it's interesting
912. 2:22:49um the last sentence you could put in a prompt is say limit Pros and it will not be so wordy so you could try something like that try
913. 2:23:00to tell it you know I don't care about your damn words just give me the just the facts ma'am be a better reviewer love Bing
914. 2:23:14it's funny some people are like love Bing other people like hate Bing love Bard hate Bard which is good that means that
915. 2:23:20means we're in a competitive space now right when there's only one option everyone's like well Chachi BT's the
916. 2:23:26bump and now all these other things are are different enough and have enough different things that different people
917. 2:23:32are going to love different tools that feels very healthy to me for now yes it's a different way of interacting yep your hair looks great
918. 2:23:39man thank you at least I have hair right I mean come on not for nothing my dad at this age I think was he had the full Gene
919. 2:23:48he he had just he had this he had like uh he took like these six hairs right here he had them in the middle oh that was bad all right
920. 2:24:01what is a free chat GPT platform that talks other than Bing pie pie talks you mean like talks like this get the iOS
921. 2:24:13PI app pi as in the mathematical circle formula circle equation no what is pi it's a result it's a constant it's a I don't know 3.1415
922. 2:24:27like that that you can talk to and it talks backs it talks back I was doing it in my car the other day I was driving around we
923. 2:24:35were driving to Costco I said oh give me the give me the specials for Costco for the for the one we were going to and it
924. 2:24:41did that I asked it about the Taylor Swift concert um I asked it about quantum physics um what else did I do anyway yeah it's it's
925. 2:24:52fun it's fun because you know chat chat Bots over the past decade and things like Alexa and Siri have essentially trained us that these
926. 2:25:06Bots are [ __ ] idiots and and their only purpose in life is to ask them for something very specific and to get the
927. 2:25:14specific response I would like to order Cheerios would you like the 96 ounce box or the 60 ounce box uh the 96 ounce box
928. 2:25:25great that'll be there Wednesday like they're good for that right but if you say Alexa I kind of feel like crap oh I
929. 2:25:33don't know how to deal with that crap are you looking for a laxative you know it doesn't it doesn't know how to
930. 2:25:39deal with a lot of what we want to talk to it about so when I was interacting with pi it's good like at one point I was talking about my
931. 2:25:52kids dealing with Lyme disease and and and it was like oh I'm really sorry to hear that I hear that can be that can be
932. 2:25:59really challenging you know is there anything I can help you with and I started having a whole conversation
933. 2:26:04about it it was like it's like holy [ __ ] and then I switched gears I said you know what I wanna I want you to tell me
934. 2:26:11about you know quantum mechanics and how that works can you do that well yeah I can do that absolutely like there's no judgment
935. 2:26:19there's deep comp competence and there's a lot of empathy a lot of empathetic cues it doesn't have empathy
936. 2:26:28but it sounds like it does and the and the way they've got the voice programmed in pi I did uh for me the the voice on
937. 2:26:35pie that was good was was Voice number six let me see if I can I can do it here on the desktop because it's not quite as good um let me see
938. 2:26:47is this on are you there oops unexpected error
939. 2:27:03seems you've been blocked why oh
940. 2:27:20um [Music] foreign
941. 2:27:35[Music] how are you doing tonight [Music] it's not talking
942. 2:27:49the desktop one is wonky hey there I'm doing great thanks for asking how are you today um I'm doing good I'm wondering if you can
943. 2:28:03help me understand quantum mechanics [Music] sure I'd be happy to help quantum mechanics is a fascinating field but
944. 2:28:15it's also complex so let's start with the basics to begin can you tell me what you already know about quantum mechanics
945. 2:28:23I know about things like superposition and a cat that can be dead and alive at the same time [Music] yep famous Schrodinger's cat thought
946. 2:28:36experiment That's a classic when it comes to Quantum Mechanics it's actually meant to be a bit tongue-in-cheek but
947. 2:28:42it's a useful way to introduce some of the concepts let's unpack that a bit shall we would you like me to start with a basic
948. 2:28:48overview of quantum mechanics first no I'm kind of bored with physics could we could you tell me if there's any
949. 2:28:57reviews of the Taylor Swift concert from last night [Music] there's definitely a lot of Buzz about Taylor Swift's Eris tour right now
950. 2:29:10I can tell the concert has been a huge success people are calling it an unforgettable experience with plenty of
951. 2:29:16nostalgic moments and references to Swift's classwork would you like me to give you a bit more detail about the
952. 2:29:21concert and some of the reactions from fans and critics yeah that would be awesome [Music] I'll tell you all I know about it so
953. 2:29:33first I'm going to warn you that there may be some mild spoilers here but in the concert Swift mix all right I'll turn it off so
954. 2:29:41I mean that that's that's haypi.com h-e-y-p-i.com is where I'm just playing and then the there's an IOS app called pi um
955. 2:29:54so so I mean so that one's really interesting because the the way the way they've got that one programmed it's
956. 2:30:00much less about giving you an answer so so um what they're talking about is creating a personal assistant
957. 2:30:07not about a chat bot right not about a large language model not about a thing that gives you answers this is more a
958. 2:30:15thing that gives you dialogue so a lot of times if you ask it for an answer it will start interrogating you it'll start
959. 2:30:21asking you questions and so and the way that it's programmed to ask the questions it's very human and so
960. 2:30:31you know I think I think as humans we already anthropomorphize technology as it is and so when when you can truly start
961. 2:30:40having conversations and when when the when the voice synthesis gets a little less stilted like like it's getting
962. 2:30:47better and it's it's close it's not quite there but when it gets a little less stilted and and they remove that
963. 2:30:53that you know half a second delay or whatever it is one second delay when when the when the response is
964. 2:31:02immediate and I'll tell you when that will be they're getting the the uh optimization of these large language models to the
965. 2:31:11point that you're going to be able to install them on your phone and and like Apple's silicon Apple's M2
966. 2:31:18and their R1 their reality or whatever it's called the the R1 chip that that reads from all the sensors it's doing all the machine learning
967. 2:31:30computation locally on the device when you have a large language model that's that sits on the phone it doesn't have
968. 2:31:36to reach out to the internet and and you can do the machine learning processing like that and then you've got voice synthesis that
969. 2:31:45sounds like a human you're going to be able to have incredibly powerful conversations about anything it could be
970. 2:31:56your virtual physical therapist it could be your psychotherapist it could be your virtual friend it could be
971. 2:32:04a counselor to help you get through grief it could be someone helping you with anxiety like whatever and it's creepy and weird as it sounds
972. 2:32:14right now like that conversation I just had about the Taylor Swift concert and the Google or the quantum physics stuff and things like that
973. 2:32:27like this is where we get into weird hyper personalization stuff I did I did something this afternoon with it where I said um
974. 2:32:37I said I've got to go I've got to go wait in a line somewhere or I was getting my car washed and I was
975. 2:32:44waiting to to get the like the interior vacuumed out or whatever and so I'm just sitting waiting and I
976. 2:32:52said can you come up with a game for me to play where I learned something and it said oh that's a great idea what
977. 2:33:00would you like to learn about and then I came up with the whatever the topic was and then she said she said
978. 2:33:08this is awesome she said yeah we'll do a game I'll ask you questions and every time you get a question right I'll give
979. 2:33:14you points and it's like it was like it just came up with a game and it was an educational game and it's like so
980. 2:33:24you know do we quickly move from people listening to audiobooks to people having a conversation about
981. 2:33:34the audio books or about whatever books or any topic or any time in history or any person they want to learn about
982. 2:33:44so it's coming like the the the the dramatic difference between Siri and that conversation I just had
983. 2:33:55like can't be understated or it can't be overstated it's it's insane it's just such it's a completely different thing
984. 2:34:03they're like I don't even we don't even know how to do it and it's the other thing is going to be this you'll you'll
985. 2:34:08be having these conversations it'll be talking back and you're like oh could you just show me a picture of that sure boom up it comes
986. 2:34:14what would the what would that sound like as a song Oh do you want to hear it do you have a particular genre you like
987. 2:34:20more than the others yeah I like Americana okay let me do an Americana song about that is this what you had in mind that's a
988. 2:34:29little too slow do you have something in in three that's that's a little peppier sure here's that yeah that's good could you
989. 2:34:36save that to my music library great anyway where were we right it's just you just go hyper personalization of media
990. 2:34:48who are the media creators then is it the tool builders is an application layer sitting on top of the tool I don't know I don't know people
991. 2:34:59I got no answers tonight I got I got lots of I got lots of predictions and and vision I got no answers tonight I was in the
992. 2:35:09film industry and and I want to be actor s for my advice I said get a new Dream Dream Crusher I'll tell you Chris it's not bad it's not bad
993. 2:35:19advice it's not bad advice it's a it's a rough business but you know some people have to do it like like when I decided I wanted to be
994. 2:35:29an actor it was It was a 15-year obsession and at some point I think I feel like I'll do it again but I'll do it for a
995. 2:35:41very different reason right I'll that gets back I was talking earlier about you either love the process or you love
996. 2:35:46the end product and if you love the end product this stuff is awesome for you and if you love the process this stuff's a disaster
997. 2:35:54so if I think if I get back into acting I'll do it more for the process and if it turns into good results great but I
998. 2:35:59won't give a [ __ ] which that'll be like it's an awful feeling to be an actor and to want to work so bad and then you just
999. 2:36:08can't because it is such an oversaturated Market do you moisturize secret dancers back here talking about my dry skin again I
1000. 2:36:20think I've got some shine I'm a little moist I don't know what the hell you're talking about secret dancer I don't even think you dance
1001. 2:36:28and you're not that secret AI knows AI knows what's going on with you I love your acting telling AI to go [ __ ]
1002. 2:36:37itself thank you hey I'm gonna tell you what yes so check that Ai and it [ __ ] lied to me so then I said go [ __ ]
1003. 2:36:46yourself yeah I'll be over here I'll be over here you tell me you tell me when you want to get back to me
1004. 2:36:52when you're not gonna lie to me huh this is about truth this is about integrity um this is how we work here you coming at
1005. 2:37:01me from the side yeah I don't go for that the robots are gonna kill us he's selling too oh secret dancer good Lord get a life
1006. 2:37:16um if you're not I I gotta tell you I do a lot less selling than I could I'm on here two or three hours a night
1007. 2:37:26just giving and giving and giving and this is the thanks I get Shaw all right if you're not living on the edge you're
1008. 2:37:34taking up too much room yeah yeah but you know some people don't like living on the edge he wants us to subscribe no
1009. 2:37:41I don't I don't really care in fact I'm going to take off the subscription thing because I don't really give a [ __ ] yeah baby all right
1010. 2:37:52all right [Music] secret dancer you shouldn't subscribe you shouldn't even show up for these they are I'm unqualified and it's not
1011. 2:38:05worth anything so don't subscribe any of you and secret dancer you can leave please all right I think
1012. 2:38:13the personalization is cool but will be a novelty and people will get over it I don't think so I don't think so I don't
1013. 2:38:20think it's going to be exclusive I don't think it's going to be like personalized content completely subsumes
1014. 2:38:26you know generated content I think I think it's just like when film came along you know plays didn't go away right we
1015. 2:38:34have we still have Broadway but we have Broadway right we we don't have theater as the only Entertainment
1016. 2:38:42form other than reading and writing right you know films came along it's it's a new medium and then video games came
1017. 2:38:50along it's a new thing I think this idea of hyper personalization comes along it'll be a new thing we'll still have those other forms but
1018. 2:39:03you know you really hate the sad ending in whatever your favorite film is let me watch that film where they don't all die at the end okay
1019. 2:39:20and I think what's what's fascinating in that in that future is it's very possible that I create a version of a movie that has the happy
1020. 2:39:31ending and you know the original had the sad ending and they'll probably be the equivalent of Netflix of the future where I can say
1021. 2:39:41hey that was good enough that I want to sell that and I'm going to put that on the Netflix of the future
1022. 2:39:49and there's the very real possibility that my version of that movie will will be as popular as the original or more
1023. 2:39:58popular than the original and then and then what happens then right so do the original creators get 90
1024. 2:40:05of their revenue and I get 10 do I get 50 50 do I get nothing because it was just a brain fart and I put it out there I don't know
1025. 2:40:17why is chat EBT developed to only 2021 with its content the way they build the large language models is a very labor-intensive
1026. 2:40:26uh manual process where they have to tag a lot of the content and the data we have to then do this thing called
1027. 2:40:32embedding which is very expensive where they turn all the all the words from all the content into these vectors these numbers you know
1028. 2:40:41semantic language clouds um and and they have to basically just lock it down at some point and say okay you
1029. 2:40:50know this model cuts off at this point so Claude um is a more recent model so it cuts off sometime in 2023
1030. 2:41:00but if you use Bing Bing is gpt4's base large language model that that's that stops at 2021 but because it's connected to the internet
1031. 2:41:13if you ask it about Taylor Swift it'll basically go do a search on the internet get the latest data and it'll augment
1032. 2:41:20you know on top of that large large language model but that's why it does it because it's just it's a very labor-intensive process
1033. 2:41:28to make that base model do you offer a list of your top favorite AI apps yeah I sit beside it every night these are them right now so um
1034. 2:41:42the things I pay for chat GPT I'm gonna get rid of Poe but I pay for Poe Claude is free Bard is free I pay for mid-journey
1035. 2:41:56I pay for zapier which is an automation tool and I pay for Google collab in case I want to geek out and I might pay for one other but it's
1036. 2:42:11basically a large language model an image generation tool an automation tool and a and a virtual machine Google
1037. 2:42:21collabs like a virtual environment where I've got access to gpus if I want to run models you know in a kind of in a standalone way um
1038. 2:42:33and that's it right now like I'm not paying for a lot because this stuff's moving too fast like all the stuff at
1039. 2:42:40Future pedia there's probably you know there's probably 30 or 40 amazing Tools in there but you have to sift through
1040. 2:42:4938 or 3 800 or 4 000 to find those 30 or 40 amazing tools Runway ml is really interesting for video um if you just need loopy kind of
1041. 2:43:04background music um for projects then sound draw is interesting it's like it's like GarageBand except you just describe a genre and it makes
1042. 2:43:15you 20 or 30 songs and then you can you know mix and match the clips and the sounds and you know it's so that's interesting I don't know
1043. 2:43:25that's about it like like these things are so deep you know I could I could easily spend the next decade trying to figure this [ __ ]
1044. 2:43:32out like like what they actually mean what what they actually do code interpreter at chat GPT is is so mind-blowing and I like I I am so
1045. 2:43:45um under educated about [Music] um what I can even do with it or like what it what it even makes possible
1046. 2:43:55other than to know holy [ __ ] it's going to change everything so I need to get much more educated about code
1047. 2:44:03interpreter and what it makes possible and that's just going to be through experimentation so I just and the
1048. 2:44:08plugins like and now that there's five of these and there's pie the one I was just talking to there's just so much
1049. 2:44:15just in the in the chat GPT models I don't think you have to dig much farther than that other than the image Generation stuff
1050. 2:44:23radical personalization Choose Your Own Adventure that's exactly right he's selling services I'm not selling services actually I've got a uh
1051. 2:44:32a video technology company we sell subscriptions but really to Pharma I don't give a [ __ ] if you by any of our video services
1052. 2:44:42I think we'll see greatness out of the man secret dancer needs a hug can we can we all give secret dancers some love
1053. 2:44:51I think they are uh they need a hug I think we'll see greatness out of the generation that grows up with AI yeah for sure here we go money
1054. 2:45:03oh secret dancer he's so sweet all right what do you think of the SAG strike I talked a lot about that tonight [Music] um
1055. 2:45:12Miss perch secret dancer what the [ __ ] are you talking about don't pay for free services I agree don't pay for free services certainly
1056. 2:45:23don't pay an X actor a penny oh my God what services are you talking about these I don't make any money off these
1057. 2:45:37that's open AI like Bings free clutch free pose free bards free you can pay for Poe you can pay for a chat gbt but you don't have to pay for
1058. 2:45:47any of them that's free that's free I don't know what you're talking about but whatever I don't give a [ __ ]
1059. 2:45:54okay I will be ignoring secret dancer from now on what do you think about sag strike I think it's I think they're fighting a
1060. 2:46:02losing battle we love you Kyle thank you Troy appreciate that I gotta go um I thought he said stand up not shut
1061. 2:46:09up not sure what that means is Kyle Frozen artificial intelligence movie 2001 Dr no Dr No scene reminds me of GPT we're we
1062. 2:46:20are in uh here's an interesting thing so I did it I did a table go to go to go to any one of these and type in make me a table of science
1063. 2:46:30fiction movies um what are the major technologies that that were science fiction at at a time and then either give me it's here now
1064. 2:46:43um like like you know is it is it here now it'll be here in one to two years it'll be here in five years or it'll be here in a
1065. 2:46:52decade and it made me a whole list of movies and then and then what the Technologies were and it said you know
1066. 2:46:57what things were here or coming it's a really fascinating list because it's like we're sort of going through the
1067. 2:47:03movies of our you know of our childhood going yeah we can do that now we can do that now we can do that now so it's amazing
1068. 2:47:11uh what site would you suggest to write a book well you could play with a GPT author if you're geeky and want to fire up Google collab
1069. 2:47:21um but I I would use gpt4 at open AI still Frozen for me it's my internet Thelonious that stinks turning off Wi-Fi
1070. 2:47:30open up opens up new worlds for me for Thelonious uh good Lord all right I gotta go let's see xrp is about to go parabolic
1071. 2:47:42someone's talking xrp in my little AI chat we don't talk about crypto Bros in here work harder nobody cares come on
1072. 2:47:51Kyle what site were you speaking of I'm not sure cast yourself LOL um you don't read live comments well tigi nuggy okay Tiggy Nuggie [Music]
1073. 2:48:03um 7.5 out of 10 on the name it's close it's close 7.5 out of 10. uh no I just read I read all the comments so I'm usually way behind but
1074. 2:48:15Tick Tock just jumped me to the bottom and I gotta go so now I'm kind of at the bottom scrolling upwards what's your
1075. 2:48:21view on that uh what about elon's hyped up xai what's your view on that um he's gonna do what he wanted Sam Altman
1076. 2:48:28to do he's going to create an open source well he says he's going to create an open source you know AGI seeking company um
1077. 2:48:41we'll see if he actually does it I mean I wouldn't put it past him I mean he's got he's got the resources he's
1078. 2:48:50hired some really [ __ ] smart people he wants to understand the universe like the imperative that he talks about that
1079. 2:48:56this is Elon Musk with his xai company the the imperative that he talks about about understanding the universe is a
1080. 2:49:02really interesting imperative for the AI but uh we'll see we'll see he's got a lot of competition man I mean I mean Chachi PT
1081. 2:49:12doing what it did and sort of unleashing the floodgates of these competitors this quickly plus pi plus meta plus we
1082. 2:49:22haven't heard anything from Apple yet plus xai I mean we're gonna have at least a dozen increasingly powerful multi-modal
1083. 2:49:32tools that just do all this [ __ ] and and and then they're all going to get incorporated into all the tools
1084. 2:49:43we're already using so we're going to have tools outside of what we do day to day and it's it's going to start to
1085. 2:49:47infiltrate all the stuff that we do so it's going to be everywhere um [Music] where can I find this site I just came
1086. 2:49:59up on you live I don't know which site you're talking about oh that site was Pi heypi.com or Pi dot a i
1087. 2:50:09was the one where I was talking to the bot that was Pi that's inflections uh service it's pretty good between my AAA
1088. 2:50:17membership and AI leveled up assistant sounds like the best boyfriend ever that's awesome holy [ __ ] my hand and my
1089. 2:50:26face are hurting from the Kevin McAllister moments every night here all right all right everybody where can
1090. 2:50:34I find pie it's not p-i-e it's p i dot a I or hey pie h-e-y-p-i.com or no dot a wait haypi.com and then Pi dot AI
1091. 2:50:49all right everybody I gotta go it's been fun hanging out with you secret dancer I hope you get a hug I love you too um
1092. 2:51:00being shot of these things if you're if you just got here grab some of you know grab these URLs go start playing with them
1093. 2:51:10that's official chat that's Microsoft it's getting increasingly powerful claude.i AI is a really interesting competitor to chat
1094. 2:51:20gpt4po.com you can make Bots and play with six different models there and then Bard is pulling up the pulling up the rear but
1095. 2:51:30but getting really interesting like they're incorporating uh links out to the internet now in a pretty organic
1096. 2:51:37looking way that's way better than I thought it would be so all of these are good prompts.chat will teach you how to
1097. 2:51:43prompt future pedia we'll let you find all sorts of rabbit holes to run yourself down peace out everybody hope you had a good time tonight
1098. 2:51:51foreign