aiLL 6/23/23 - 2 - The Mind-Blowing Power of AI Tools Like ChatGPT

Transcript

1. 0:01foreign
2. 0:20Fox what's happening you're the first one in how's my internet because it was just crappy I had to restart my router
3. 0:28is the internet good am I all Herky jerky howdy hey cavuno shaking cheers um better okay good
4. 0:44looks fine but lagging at the moment huh strange I'm using five maybe I said now okay all right
5. 0:58how's the pantomime I guess I guess I can't do twerking tonight I was gonna twerk tonight but if I've got laggy internet nobody nobody
6. 1:08wants to see Herky jerky twerking okay you're back I'm I think I'm back we'll see it's up to Xfinity apparently
7. 1:19I like for the amount of money we pay these stupid companies you'd think you'd think they could make this [ __ ] work you know
8. 1:28wait till you get 5G it's gonna be incredible yeah it's not all right all right um welcome everybody just the AI
9. 1:40learning lab uh if you would feel free to share the this live if you know how to do that I don't and tap the screen let people know we're
10. 1:49here that'll get people in my name is Kyle Shannon I'm an entrepreneur um I'm the founder and CEO of a company
11. 1:56called storyvine I'm super into this AI stuff um I've been doing this channel for I don't know a while now and I've been
12. 2:04doing these lives for a few weeks instead of sim racing at night I do lives and talk about generative AI so if
13. 2:11you have questions about AI pop them in the comments below I'll do my best to answer them um hopefully my internet's back and I'm
14. 2:18not Herky jerky um trying to think what else is going on lots of new stuff happening mid Journey 5.2 came out
15. 2:27um stable defer diffusion XL 0.9 came out um chat gbt still rocks if you haven't played with chat GPT these URLs behind my head
16. 2:41um the reason I put these here the the top one there is the official chat website if you haven't played with it
17. 2:47just go start playing well you hear me ramble and and if you really don't know what to do don't know where to start
18. 2:52that's why I put this here this prompts.chat thing it's like a little instruction guide and a whole pile of different
19. 2:59um prompts where you get chat to you GPT to act like different personas and it and it really guides um what it can do well and it'll give
20. 3:08you ideas about what it makes possible so that's that bing.com is Microsoft's search engine and if you click the chat
21. 3:14button that's chatgpt4 for free connected to the internet so none of this ending in September 2021 nonsense
22. 3:22it's all connected to the internet it's free and then poe.com gives you six different models you can play with so
23. 3:28and then futurepedia.io I put that there really just as um if you want to geek out on tools or you want to know specific tools like
24. 3:37music tools or animation tools or writing tools that's what is there there's like 4 000 tools 3850 of which are pieces of [ __ ]
25. 3:48oh I also cuss so if you don't like cussing I would leave because it's it's Friday night and I'm drinking a hot hot cup of water
26. 4:02um taste all right I don't know why I just want all Will Ferrell there but you know he's also a comedian and very knowledgeable
27. 4:11thank you I appreciate that um I try to keep it entertaining because you know it's computer [ __ ] it can get
28. 4:20dry real quick but if you want to know cool things like how does chatgpt work I can answer that if you want to know
29. 4:26things like how can you make money that's not what this channel is about but I'll give you ideas on how you might
30. 4:34use Chad's EBT and you can go figure out how to make money with it uh um and by the way if anyone tells you
31. 4:40that like tries to sell you the course to make money don't spend your money on that those people are idiots nobody knows how
32. 4:51to make money on this stuff yet well some do the ones that are making money on it mm-hmm okay so you should try movies too I was in
33. 5:01New Jack City I have a degree in acting so I've done a little bit I did mostly New York stage stuff and but I did a little bit of movie work
34. 5:08I did I did uh did some uh New Jack City I was uh white boy number two [Laughter] in New Jack City how does it work do you
35. 5:21want to start should we start out right out with the geeking out all right you sure all right we'll do
36. 5:28this hang on let me get up gotta go get a keynote here hello please let me flip the camera we're gonna do a little geeking out
37. 5:38we're gonna do a little Jane how does chat GPT work oh look there's there's my board meeting deck no I'm not showing you my board meeting
38. 5:48there's the content Gap and omnichannel in the Pharma business this is a lot of my business business stuff where is my here it is okay
39. 6:00so this is a presentation that I did for a thing called creative mornings and that's this is sort of the
40. 6:08the payoff slide of the deck where I'm talking about I where's the slide here so that date is the data everything
41. 6:16changed in my opinion that's the day that Chachi BT launched and so I think we are entering uh a great Renaissance
42. 6:24so I think these tools are going to be absolutely mind-smashingly powerful and lead us into a new era of creativity but
43. 6:35you want to know how this [ __ ] works huh okay let's do this let's do this okay so let me make this window a little smaller
44. 6:46so we can all see it and I'll just I'll sort of don't get back and forth here okay so so chat gbt one of the big
45. 6:58misconceptions about chat GPT and things like mid-journey and stable diffusion is that um They're copying and pasting you know
46. 7:06parts of documents right you hear a lot of people going oh they're stealing the robots are stealing the robots are not stealing um
47. 7:15so so we'll start with what does GPT mean GPT stands for generative pre-trained Transformer and the generative part is it's generating
48. 7:24original content it's actually generating word by word original content pre-trained means it's been pre-trained
49. 7:31on if essentially all of the internet all the public internet or as much as they could get to uh in the amount of
50. 7:36time they had to put it together and then Transformer is the Transformer technology that Google invented in 2017
51. 7:44so even though Chachi PT is new and it seems like everyone's just talking about it now it's this has all been in
52. 7:51development you know machine learning and neural networks have been around for decades the Transformer was invented by
53. 7:57Google in 2017 open AI you know jumped on that and created gpt1 and then gpt2 nobody really cared about them then they
54. 8:07um launched gpt3 and all of a sudden people were like uh this is looking interesting and um then in November uh they did GPT uh
55. 8:183.5 or you know chat gbt they launched at gbt um and now everyone's lost their mind so that's what GPT stands for now you'll
56. 8:26also hear a large language model the large language model is basically that that core model that's been trained on
57. 8:32all of the Internet it's the thing that lets chatgpt do it it does um and then you're also going to hear
58. 8:38about something called the latent space well you might not hear about it but this is how it works so there's this
59. 8:43stuff called the latent space which is a mathematical representation for lack of a better term of the locations of the meaning of words
60. 8:53so so the the graphic that I'm showing you here is like a simple two-axis XY coordinate and all those clusters of
61. 9:00dots are numbers that represent different words that are clustered by semantic meaning right clustered by
62. 9:08meaning right so those are semantically related words so for example you might have the word dog and in one of the
63. 9:14Clusters the words dog are like animals and you know the fluffy thing we pet in another area dog might mean that awful
64. 9:23man that didn't treat his lady well right so so you can have the same word can have multiple meanings and live in
65. 9:28different semantic regions now the real mathematical space is thousands of Dimensions like it's incredibly
66. 9:35complicated it's incredibly huge you know imagine training on all of the internet okay so how does it actually
67. 9:42work basically what happens is when you put in your prompt it converts that in into a mathematical probability and says
68. 9:50based on all of the words that were typed in that prompt what is the next most likely word that I should type you
69. 9:56know in response to that and so it kind of works like this if if our prompt is the quick brown fox jumps over the lazy
70. 10:03blank we know the answers dog because we have a brain and the brain works well how does the large language model neural
71. 10:10network brain work well what it does is it it effectively deselects all the stuff that is not in in that semantic
72. 10:17region and there's some little light blue dots here you might not be able to see them but these little light blue
73. 10:22dots are all things that might have you know some decent probability of filling in that sentence and there's going to be
74. 10:28one of those words that says 96 there's going to be one of those words that has the highest probability of being the
75. 10:34right word and if you look at what's in the database or in the in the vector database it's actually just a string of
76. 10:40numbers that represent that location in Vector space and then that number is associated with a word and in this case
77. 10:48the word is dog and so it puts the word dog in that sentence as a response and right there's other things that had
78. 10:54other probabilities but dog had the highest probability and that's how it works and it does that word after word
79. 10:59after word so when you see it write you a whole song or write you a marketing plan or create a table or write you code
80. 11:07this is how it's doing it word by word with mathematical probability it has no idea what this word means it has no idea
81. 11:17what it's actually writing but it's so sophisticated that it looks like it does whoo pretty intense to start out with you know
82. 11:27not for nothing how does chat EBT work come on but I'm here for you we got it what do you think cool all right
83. 11:36uh if you if your thoughts questions about uh AI uh anything you want to ask pop them down below and I will um do my best to answer them
84. 11:48in a small town and I've been talking to small business owners about learning AI actually what would be what would be cool is
85. 11:56um maybe maybe this is what you mean is is start teaching them like do courses or do work for them like like
86. 12:04one of the things that I've been doing in the past three or four months is I I've just put myself in this just say yes mode
87. 12:11where I'm just saying yes like like someone says oh let's collaborate on something yep let me do that let me help
88. 12:18you figure that thing out let me build something let me do this let me do that um it's completely exhausting and it's
89. 12:24completely burning me out but we're at a singular point in history um you only get to participate at the beginning of a technological
90. 12:36transformation like once and it's like once usually in a generation in my life there have been three major ones there was the personal
91. 12:46computer which I was too young for I was in seventh grade when I got my hands on one of those and I was acutely aware
92. 12:53that ah I just missed one of those big technological revolutions I was there for the World Wide Web I started one of
93. 13:00the first digital agencies in 1994 grew it really big like like rode that up that was insane it was wild it was amazing
94. 13:08and I didn't think I'd have another one in my life these things don't come along that often and it's it's here 2022 2023
95. 13:16um this is the first year of a new inflection point these are the most powerful tools ever put in kind of the hands of the masses
96. 13:26um and and it's going to change a lot it's going to be quite profound so so I feel like we're very very lucky we're very lucky to be here
97. 13:33um so anyway so that's why this channel exists I want to demystify this stuff I want to inspire you hopefully make you
98. 13:38laugh a little bit uh maybe learn a little bit um I'm certainly not expert at this anyone who tells you they're an expert
99. 13:47at any of this is full of [ __ ] uh wait till hang on there's an important distinction the the people that I'm talking about
100. 13:58that are that are not experts right now are people saying all the chat GPT stuff all the generative AI stuff that's been
101. 14:04sort of November 30th 2022 forward nobody's expert at that there are lots of experts at machine learning and
102. 14:11computer science and neural networks and all that sort of stuff there's Decades of expertise there what I'm talking
103. 14:17about is okay now that these tools are in the hands of the rest of us how do we use those that's that's where I'm really
104. 14:23interested so having started a bunch of companies that's the kind of thing that really excites me and I like putting my
105. 14:29time and energy into and that's where I feel like there's a lot of people right now going I'm an expert at prompting I'm
106. 14:36an expert at making money at chatgpt you should hire me I can tell you everything no no one can tell you everything
107. 14:44um so sometimes if you ask me questions in here like how do I do this um I'll often just say just go there and
108. 14:51start playing which I know is a frustrating answer but I think it's actually important to start learning how to use these things and how to
109. 14:58how to think about interacting with them because it's not like Google it looks like Google but it's not like Google
110. 15:04quick question yes go ahead uh is there a chat GPT plugin for python you don't need a chat gbt plugin for python I'll
111. 15:11show you my shirt is intense love it thank you very much um let's see oh yes I'm teaching them and asking them what what they want
112. 15:21their future to be and I help that's super cool Silver Fox keep doing that because it's I mean it'll help you learn
113. 15:29um it'll it'll keep you educated and and you know you'll get better and better at problem solving but it's also going to
114. 15:35help them I think that's really awesome um is there a chat GPD plugin for python well actually there might be let's let's
115. 15:43go look okay and I'll show you about plugins because we we talked about how chat EBT Works let's go look at plug-ins
116. 15:50and I'll give you a quick little primer on plugins if you haven't played with them so where I'm going right now is chat.openai.com
117. 15:58and that's the official chat EBT website okay so if you're new to chat GPT to get to the plugins you have to be a chat GPT plus member
118. 16:08I was logged out of mint defense okay um and then you get access to it the other thing that you need to do to get access
119. 16:15to plugins is you need to turn on your beta features so you go to your settings and then within your settings you go to
120. 16:25Beta features and then you enable browse with Bing and plug-in so you have to turn those on first so you won't see
121. 16:31them and that can be confusing okay um then we go to plugins so the way plugins work they're they're a little confusing
122. 16:42so with chat GPT in general you just type your [ __ ] down here and it answers your question in the in the window
123. 16:48plugins are are little extensions and it the the the software kind of figures out is what that person asked something that
124. 16:56this plugin could help with and if so then fire off the plug-in that's kind of how they work um you you pop open this little list and
125. 17:05you've got the list of your installed plugins in any given chat session you can enable enable up to three of them
126. 17:11but you can install a bunch but for any given chat session you enable three um to get to New plugins you scroll to the bottom you go
127. 17:21to the plugin store there are now I think 500 and 50 or something like that there's a lot of them so if I go to all here
128. 17:30there are 139 pages of for each this will create Spotify playlists for any prompt that's pretty cool let's look up python
129. 17:43create notebooks in Python SQL and markdown and explore data visualize and share notebooks with everyone
130. 17:50um notable I don't know notable that's probably just a website but I'll go oh yeah I've got I've got a the best Jupiter notebook alternative
131. 17:58yeah I'm not going to create a new account for this some of the plugins are really just sort of gateways to other
132. 18:03websites so there's not a python plug-in but Chachi PT knows python like this was my big Epiphany with chat gbt um so let's say right
133. 18:18right uh sorry I'm typing typing while talking is not good oh wait someone else no I'm not going to do it okay I could speak it in but I
134. 18:30don't know how to turn it off because I'm an old man and I haven't figured that [ __ ] out yet okay uh write a python
135. 18:36application that refactors um I don't know uh these squares of prime numbers
136. 18:52and graphs them on an X Y axis I don't even know if that's possible I'm not good at math I know some math words
137. 19:05sure here's a python application that does that so it's going to import the matplotlib library and simpi and then
138. 19:15it's going to do some [ __ ] and then it says this script generates the first hundred prime numbers squares
139. 19:22them and then plots the prime numbers on the x-axis and and their squares on the y-axis you can adjust and then
140. 19:30note that you have to install the mat so it tells you like it gives you instructions on what to do um
141. 19:37you could and basically you can just go on and on like this right you can um take existing code have it decode it it
142. 19:45can read code it can translate code so if you've got some application in some other language you can translate that to
143. 19:51python so it's it's pretty remarkable um there's a question there about when did AI really start Let's go ask the
144. 20:01chat gbt when did AI really start I like I think you could argue probably in the 60s was the first neural network right start question mark
145. 20:14a long and complex history yeah coined artificial intelligence was coined in 1956 at the Dartmouth conference
146. 20:24where researchers gathered to explore the possibility of creating machines that could simulate human intelligence
147. 20:30the foundations are even older so 40s and 50s devices and algorithms that could perform tasks resembling human intelligence 60s and 70s
148. 20:43people they developed expert systems right which which were um sort of hard-coded versions of this so
149. 20:58it was coined in the 50s and then really kind of caught you know they they started building all the neural networks
150. 21:06and [ __ ] like that in the in the late 90s and early 2000s so it's been it's been around for a while like I said
151. 21:12there's there's some real parallels between the early days of the world wide web and and what's going on with generative AI right now so
152. 21:21and when I when I kind of discovered the web in 94 the internet had been out for decades right but it was mostly
153. 21:28researchers and scientists and [ __ ] like that and then Tim berners-lee creates the world wide web and all of a sudden you
154. 21:36know normal people can like make these websites and then the websites you don't have to type in command lines you can
155. 21:42just click on these Blue Links hyperlinks and if you click on that hyperlink it takes you over to another
156. 21:48document and that document could be all the way on the other side of the world and all of a sudden that that little
157. 21:54simple ability like you know Steve Jobs with the iPhone like this is the this is the best input device right so that's
158. 22:01what the World Wide Web was similarly machine learning and ai's been around for a while like ai's in all of the
159. 22:07software that we use it's been in it for years and years and years what's different is chat GPT made a simple little like this kind of
160. 22:16interface for machine learning and it and it's pre-trained on all of the internet right so it's incredibly smart
161. 22:24we don't have to figure out which data to give the machine it's already got all the data all we have to do is ask it for
162. 22:30what we want and then it goes and gives you the answer um just remarkable um which can open it oh can open AI write a book yes and in fact
163. 22:45um go check out a project this is really cool go check out a project called GPT author GPT Dash author it's a GitHub so
164. 22:55if you know how to use a Google collab notebook there's a Google collab notebook there it's not that hard to run you just push play buttons
165. 23:03um but um you give it you give it a topic and then you give it a genre and and you give it a writing style and then it will it uses stable
166. 23:18diffusion to generate a cover it generates an outline and then it writes a chapter for X number of chapters you tell it how many chapters
167. 23:27you want and then it goes and writes you a book it's it's insane so yes it can and and how it's doing that is is it's it's not
168. 23:37sort of it's not a single prompt that just says go write a book it's doing it in this procedural way right take the
169. 23:45topic incorporate the writing style incorporate the genre um come up with a title come up with an outline now that you have the outline
170. 23:54you know write the first chapter now that you've written the first chapter write the second one right and so it it
171. 24:01goes through this process you can actually watch it building building the book you're going through the steps and
172. 24:07then after I don't know five minutes or so it's like there's a file and you just click on it and it it opens in uh like
173. 24:15an e-book reader and you can read your book that was just written for you so but you can do that yourself too like
174. 24:21that's one of the things that you could do at chat EBT if you've got an idea for a book start with the outline have it
175. 24:26help you write the outliner if you've got the outline have it come up with new characters or whatever it might be
176. 24:32um really interesting I'm demoing GPT my media classes trying to show the good and the bad tips to describe the hallucinations
177. 24:41um yeah so so so this is not for me I'm stealing this for Monday Andre carpathy who's who's one of the co-founders of openai so so the
178. 24:50inventors of this [ __ ] I'll give you the metaphor so here's the metaphor that really that popped into my
179. 24:56head as I heard him talking about this okay so the the the base large language model that sits underneath chat GPT
180. 25:103.5 or gpt4 was trained on all of the data of the internet right as far as the eye can see all of that data for the
181. 25:19past 30 or 40 years whatever is accessible on the net they tried to capture as much of that as they could
182. 25:24and just train the thing on that so that's the core learning what Andre carpathy was talking about he was
183. 25:30talking about the Microsoft developers conference and he said he said if you ask chat GPT to solve a very specific math problem
184. 25:41it's going to look at all of the internet and it's going to figure out where do I have answers to that math
185. 25:46problem and it's going to look over here and over here is a professor that is an expert in that math problem and he's
186. 25:54gonna have the right answer and then over here are a bunch of students that also answered that math problem and they
187. 25:59got it wrong and there's like 30 of them and there's one of him right um and so if you just say solve this math problem it basically sort of
188. 26:10averages in all of the data it knows that are kind of semantically related to that question and it'll solve the math
189. 26:15problem and if it pulls in more of the student answers then the the professor answers you're going to get a wrong answer so
190. 26:24because it's just doing word probability like word by word you write a long prompt and then it says what's the next
191. 26:30most probable word and the next most probable word and the next one because it's doing it word by word it has no
192. 26:37concept of what it actually wrote so so it's it's like it's it's like rain man like it's incredibly intelligent in
193. 26:44its output but it has it's got like no you know concept of what it actually did so what he said is you need to narrow the semantic um
194. 26:57um vectors the corners of this mathematical you know language space that you're in and here's the metaphor that I came up with
195. 27:06pretend you're in a helicopter over the Grand Canyon and if you go really high up you can see in all the nooks and
196. 27:13crannies all the little mini Canyons so you can see and and in the Grand Canyon the Grand Canyon is filled with all of
197. 27:19the knowledge of the internet and that if you're really high up in the helicopter you see everything right
198. 27:25and so if you ask a really generic question like solve this math problem the helicopter is really high up so
199. 27:31it'll get the crappy answers as well as the good ones but if you say GPT I want you to act like a Stanford math
200. 27:38professor that specializes in this branch of mathematics and I want you to solve this problem it's like you're
201. 27:44lowering the helicopter into one of these Canyons into one of these little small sub canyons and the only thing
202. 27:49that's in there semantically is professors answering that problem correctly so you'll get a much better
203. 27:54answer if you if you sort of narrow it so that that's what I would that's the way I'm understanding why it
204. 28:00hallucinates and the other thing that's weird is because it's been programmed to kind of talk like a human
205. 28:08you kind of think like it kind of is sentient and it kind of knows what it's talking about but it is just a
206. 28:15calculator that is doing probability Auto completion um it just happens to be like the the most powerful tool ever in the history
207. 28:25of mankind right it's like it's like having a really incredibly smart intern that has no life experience you have to
208. 28:33tell it everything to do but if you ask it about some obscure topic it'll know about it right it's it's it's really
209. 28:40quite remarkable that way and that's a big uh that was a big shift for me shifting from Google searching where
210. 28:48like a Google search is a very particular kind of thing it's like you put in a very concise sentence and then
211. 28:54you get back a big list of answers and if you get good at Googling you know you don't have to you can hit it in the
212. 29:00first shot and it gives you back what you need right so they kind they kind of over the past two decades have taught us um
213. 29:08write short and concise and we'll give you all the answers chechi PT kind of sucks at that but what chat CPT is really good at is you go all
214. 29:18right I'm gonna actually start a conversation with this thing and I'm going to give it a lot more context
215. 29:23than seems reasonable hey chat GPT I'm thinking about writing a book and gosh darn it you know I've got a few ideas
216. 29:31but I'd sure love your help like like you find yourself talking to it like that and then it'll respond hey awesome
217. 29:37you know would love to help you write a book what are you thinking about what's you know who's the main character oh the
218. 29:42main character is this that back and forth the more you give it back and forth the more context it understands
219. 29:48the better it gets and and so you can kind of you can you can narrow the hallucinations and and I don't know if
220. 29:55you can eliminate them because it doesn't know what it's doing so if you're doing something like writing a legal brief check your [ __ ]
221. 30:02work like two weeks ago there was a there was a lawyer that um he gave the judge um a chat GPT written legal brief and
222. 30:12didn't proof it and the judge was like all of these citations are [ __ ] and the lawyer was like well sir I don't know what to do
223. 30:25right so all right [Music] all right we just talked about who's who's hallucinations Claude I think is a writer of Books well
224. 30:39Claude so let me show you one other thing here um so if you if you're thinking of wait am I here no okay if
225. 30:51you're thinking of writing a book or doing something long form or even analyzing a book you could have you could have this tool
226. 30:58um read a manuscript you've written and help you edit it so if you go to poe.com so this is one of the the websites that
227. 31:05I I think is worth paying attention to um it's just like a normal uh chat EBT like engine it's got all these different models here
228. 31:14Sage is is there um their chatbot gpt4 Claude plus Claude instant 100K Claude instant and Chachi BT are also available but but these uh
229. 31:26the ones that say subscription access you've got to pay for to to have access to them but this one right here in
230. 31:33particular is really really interesting so it's called Cloud instant 100K that means a hundred thousand token context window
231. 31:40what that means practically is that you can put 75 000 words in your prompt so you could literally copy and paste a novel
232. 31:49manuscript of you know 50 000 word novel and then say um I don't know find inconsistencies in my plot line and it will you know
233. 32:00um rename this character to that and instead of them being an air conditioner repairman um I want them to be to work in a
234. 32:11chocolate factory and it will just rewrite that [ __ ] um so so it's remarkable I put in I put in um is it still here
235. 32:20yeah it is I put in the entire um 2001 Space Odyssey screenplay so all of this text up here is is the
236. 32:30screenplay for 2001 A Space Odyssey so I just pasted it in and if you've ever tried to paste in a long document into chat EPT it just
237. 32:38breaks it's like I can't do it I'm too exhausted but this thing will just chew through it and it it reads it in like I don't know 20 seconds
238. 32:47and then what did I ask I said what are some what did I say I said who are the main characters and what happens to them Dave Bowman he's an
239. 32:54astronaut in the discovery mission to Saturn Frank pool was his crewmate the computer HAL 9000 starts malfunctioning
240. 33:01and predicts failures so you know it reads the whole screenplay it knows what's going on and here's some
241. 33:06interesting things that Hal said to Dave um let's see um when Hal says he's sorry to Dave
242. 33:22what is it in reference to uh oh it lost oh my context was cleared and
243. 33:39so I wonder um anyway all right I cleared my context I should probably put in a Monty Python script I think that would be a lot of fun
244. 33:51all right so that's po po could help you write a book because you can put so so damn much in there oh yeah hey Kool-Aid
245. 34:03how many licks does it take to get to the center of a Tootsie Pop one two two any gen xers all right uh uh some comments in this
246. 34:18live were filtered to protect the community's experience don't be nasty in here please um do you need gpt4 for plug-ins you
247. 34:28need to pay for GPT Plus for plugins which gives you access to gpt4 and plugins so I think I don't think you can use
248. 34:39plugins with GPT 3.5 but I don't know that's incredible I'm not sure what's incredible but something I just showed
249. 34:46oh maybe the python thing I asked it to add comments in ah tick tock um I asked it to add comments to my python code
250. 34:59that was my Kevin McAllister moment so yeah so on these lives I've been saying you owe it to yourself to play with
251. 35:06cheche EBT until you have the Home Alone moment [Music] skepticism and the you know and the uh it's not perfect
252. 35:20and actually starts playing with it has this moment and and commenting code is amazing we did I was for for my company we've got
253. 35:31an app and and um we had a an iOS Developer who was he was good but he was young right so he wasn't super experienced
254. 35:40and so for one device one particular uh model of iPhone it was just the app was just crashing we couldn't get get it
255. 35:47past a certain thing and he tried for like two weeks to solve this thing and it it was just this obscure weird little
256. 35:53bug we couldn't crack it and I said why don't you try chat EBT and he's like you know like like developers I'm not gonna let the robot
257. 36:03do my coding I'm a genius and so he pasted and pasted the code into chat EBT and it was there were just it was a typo right there were there
258. 36:12were um two characters it was a parentheses and a period that were swapped right and he goes well I don't think that should
259. 36:19so so he compiles the code launches the app and it fixes the bug it's like it's it's it's just [ __ ] remarkable it's
260. 36:26really good at analyzing other stuff um yeah so the Kevin McAllister moment yeah yeah I I mean maybe that's it maybe
261. 36:33that's my new thing play with chat GPT until you have your home alone moment and then at that point then you can
262. 36:41decide if you want to hate it or not right all right um so what made AI blow up recently great question okay so
263. 36:56there were some murmurs um probably two years ago now um open AI launches um gpt3 which was a really good you know
264. 37:08it was kind of the precursor to chat gbt it's the underlying model that was the first one that started giving results
265. 37:16that were like holy [ __ ] this is better than we thought it would be and so that made a little bit of noise
266. 37:23right around that time mid-journey and stable diffusion came out with their models where you could start making
267. 37:30these cool images and and they very very quickly went from kind of laughably bad to like pretty good except you know
268. 37:39people at hands with like eight fingers on them and you know they they looked a little cartoony they didn't quite look
269. 37:46right and then within a within a year like from like 20 2022 in the spring to 2023 in the spring it went from like laughably bad to
270. 37:56you know [ __ ] mind-blowingly good like the the current mid-journey and a lot of the stuff being done unstable to
271. 38:02Fusion is just mind-blowing in fact that image uh that's behind me that was a stable diffusion thing I did with dream booth um so so
272. 38:15it was starting to kind of emerge as this interesting thing and then in um probably about a year you know in in Spring of 2022
273. 38:29and maybe even before that Microsoft starts working with openai and they start working together and so you know
274. 38:35you sort of come around to October of 2022 and Microsoft's getting ready to announce their billion dollar investment in open AI
275. 38:44and and Sam Altman wanted something to be able to show to the public like here's what this thing is and they didn't have anything it was
276. 38:52everything was for developers so so this is an interesting tidbit in the documentation of gpt3 where the
277. 39:00instructions to make chat gbt and Sam Altman had assumed that someone would make it and no one made it so so like
278. 39:09like with with a month to go before Microsoft is going to make this announcement of their billion dollar
279. 39:13investment Sam Altman says to his Tech Team you have two weeks to put together chat GPT so they put together chat GPT
280. 39:20in like two weeks or a month whatever it was Microsoft makes their their big announcement of a billion dollar
281. 39:27investment in open AI they launched chat GPT chatgpt went from zero to a hundred million users in six weeks it went to a
282. 39:36million users in five days and it went to 100 million users in in six weeks and then you know within a month Microsoft said okay now we're
283. 39:45investing 10 billion dollars they started with a billion they upped it to 10 and they basically committed to roll
284. 39:51open AI throughout their Tech stack right they're incorporating it in Microsoft Word they own GitHub it's
285. 39:58co-pilot they're they're building it into windows 11. so so chat apt becomes the fastest adopted technology in history from like November
286. 40:10to mid-January um Microsoft makes this big announcement and Google freaks the [ __ ] out because hey they invented the technology the
287. 40:22Transformer that's sitting underneath all of this Tech they invented it but I think they kind of slow rolled it
288. 40:29because I think what they recognize is that a tool like chat GPT would fundamentally threaten the the search business so
289. 40:39through a series of of different announcements and things like that Google is is basically trying to play
290. 40:44catch-up right now so they've got Bard which is like chat EPT but it's not anywhere near as good um meta came out and said hey we've got
291. 40:53AI too and they launched a bunch of their models and they actually gave their llama model the core model to the
292. 41:00developer Community who basically just turned it loose on the world so now you've got all these open source models
293. 41:06that are that are based on on this llama stuff that meta gave out AWS has announced that they've got something right and apple hasn't
294. 41:14announced it yet but they're going to be coming in probably with a bang so so it was a combination of the tech is really
295. 41:22remarkable and Chachi PT got to 100 million users so quickly that everybody freaked out everybody like all the
296. 41:29companies were like we got a I know we got AI use our AI so so that's that's what made it blow up and and now it's just it's
297. 41:38because of that inflection point being so like white hot it's way more than hype because you've got you know at this point probably a
298. 41:51couple of hundred million users using chat GPT you've got all of the development Community startup startups
299. 41:59investors you know everyone's saying what's your AI plan right like ai's everywhere and these tools happen to be
300. 42:06remarkably capable so that like there's just this convergence of all these things that happened and that's why it
301. 42:13just feels like it showed up overnight but the foundations have been sitting there for six seven years probably you
302. 42:19know more like two decades really um a chat bot is so much easier than more complex interfaces of the past
303. 42:27that's that's a great Point like that's to me your your comment what's your name tired Runner uh awesome name I am loving
304. 42:34if if I do nothing else but attract people with awesome screen names I I will be happy um tired Runner that's good that's solid
305. 42:42I'm a non-runner so I'm I'm in in the neighborhood with you um I think the most significant thing that these large language models do is
306. 42:54simplify the complex I think it's I think it's the thing that they do I can paste an entire screenplay into that PO 100K bot and just say
307. 43:06write me a log line who are the main characters what are the plot points you know what happens in Act One what
308. 43:12happens in act two like you can just start interrogating that data you can also do the same with complex systems
309. 43:19right so you know imagine you know imagine in the future you've got an agent where you just say hey go deal
310. 43:26with my insurance company go deal with with uh booking the doctor appointment and you know ordering my stuff at the
311. 43:34pharmacy and make sure that my deductible's covered and and like whatever all the [ __ ] that takes you
312. 43:39like 10 hours of your weekend where at the end of it you want to smash your face into a brick wall it's gonna be able to do that [ __ ] like
313. 43:50it's gonna be able to just take that complex [ __ ] and do it and that's what it is that's what it is that's the Kevin
314. 43:58McAllister moment when you realize I didn't know it could do that I showed this up to I've told this story
315. 44:06on the lives before so if you've heard it before sorry but but it's it's a really good story I was I was teaching
316. 44:11at a my co-founder's son uh goes to a school and they wanted to learn about generative AI with the images right so I
317. 44:19was teaching them how to do images and they're they're like it's it's a it's a non-traditional school so it's for it's
318. 44:25for kids that like me have ADHD and all sorts of you know you know neurospices as they call them and so they're just
319. 44:31pinging off the walls and they're like make pictures of boobies I'm like no no don't make pictures of boobies
320. 44:36so after like two hours of this I'm sort of packing up my stuff and this one kid goes hey um I'm making a uh I'm writing an
321. 44:47application I'm like are you I said what what are you writing because that's in C sharp I'm like oh that's cool he goes
322. 44:54yeah I'm making a clone of Flappy Birds I'm like wow that sounds pretty fun he goes yeah I got a bug in the code so I
323. 45:00think I gotta throw it out it's like a 10 year old he's he was cool too cool for school so I said I said
324. 45:08have you tried Chachi BT to help you you know fix that ah yeah yeah I use chat GPS yeah yeah whatever
325. 45:15I'm like well here come here let me show you something so I pop open chat EPT on my laptop and I say write a C sharp here I'll do
326. 45:22it for you now because it's wild it's just [ __ ] Wild uh uh uh uh so I've got a chat you BT we'll just go to 3.5 because it's easy um right
327. 45:37me a c sharp application that is a Flappy Bird's clone uh that will run in unity in fact so so I did that
328. 45:55and it did that and this kid did the Kevin McAllister thing he literally went he goes he goes I gotta get my laptop
329. 46:09he ran down four flights of stairs came back up he's like so so I said get out your code that's broken and um
330. 46:22we copied and pasted his code into chatgpt said is there a problem here he had a typo in it so I said now copy and paste
331. 46:30that over to Unity copy it into your project so he copies it into his project and it runs and and it's Flappy Bird he starts
332. 46:37he's playing Flappy Birds and then he stops he looks at me and he goes this changes everything it's like it's just this awesome [ __ ]
333. 46:46moment right this was his Kevin McAllister moment really cool that's what it can do that's what it does all right
334. 46:55um thanks for showing this the zoom in mid-journey sure I can I can also show that in here I heard about the legal
335. 47:01brief I know the legal brief it's so funny like people the thing that bugged me about the reporting on the legal
336. 47:07brief story is they're like you know AI created a legal brief that had you know false things in it no no a lawyer used
337. 47:14the tool to create something that had shitty work in it and he didn't proofread his work that was the story not that AI [ __ ] it up
338. 47:26I love the poe IOS app very easy to use wow cool I don't know I showed but that's cool this is going to change my
339. 47:34life that's cool James tell me what you do I'd love to hear I'm obsessed I know it is this is once you have this moment
340. 47:41then you have this moment okay I gotta I gotta use it for this I'm gonna use it for this oh yeah oh oh Sally needs
341. 47:47flowers we could use Chachi BT to figure out which flowers to get which one of the flowers represent in the world we
342. 47:53can figure out what metaphors it's like it's it's like those movies where you got the crazy guy with like the all the
343. 47:59pictures on the wall with the yarn trying to connect the conspiracy theory that's totally what it's like trying to
344. 48:04figure out like all the [ __ ] you can do with with chat gbt and all these tools they're just all [ __ ] mind-blowing so
345. 48:12um all right let me if you uh yeah we've just got the it looks like we got the core group of people in here do me a
346. 48:18favor if you would just double tap the screen uh so you let people know this is going on and if you're willing to share
347. 48:25share the live with with your folks so we can get some other folks in here and then I will go answer some other
348. 48:32questions Gen X here love the commercial love that commercial yeah exactly one two two uh chat CBT how many licks it takes to get
349. 48:47to the center of a Tootsie Pop let's do that that's a solid one uh many licks does it take to get to the
350. 49:02center of a Tootsie Pop by the way gen X's have you bought a Tootsie Pop recently they used to be like the size of your fist
351. 49:17and then it was like a whole Tootsie Roll in the middle now they're like the size of their like a little microscopic
352. 49:23little like you pay 17 and you get one little speck of [ __ ] that gets caught in between your teeth Jesus all right what's the answer the
353. 49:34famous advertising campaign Tootsie Pops asked the questions how many licks does it take to get the center for Tootsie
354. 49:39Pop the answer to the question can vary upon uh uh can vary depending on factors such as individual licking technique
355. 49:47this is getting very Saucy size and consistency of the Tootsie Pop numerous studies and experiments the only way to do it is to figure it
356. 49:59out for yourself that's not the answer a few dozen to several hundred no it takes two that was the comedy all right have you
357. 50:09found any new plugins worth taking a look at um I showed instacart if you haven't seen instacart it's cool because it's it's
358. 50:20visually interesting um there's one called stories that writes kids books but the the the prompting that they do for the images is pretty
359. 50:29shitty and like the titles are pretty shitty so I haven't it's it's interesting what it does but it's not good
360. 50:37um I played with one today that was bad but I it was really interesting um there was one called rewriter that would rewrite the the um
361. 50:49it would take the copy from a website and rewrite it for you which I thought was kind of clever but then the last
362. 50:54paragraph is a link to if you want help doing [ __ ] like this click here to go to our Ad Agency so it's a plug-in that was
363. 51:01an ad for an ad agency which I thought was pretty clever so I might steal that uh but no the the plugins I mean
364. 51:09there's like almost 600 of them now a lot of them are just kind of Fly by Night little pieces of [ __ ] that were
365. 51:15just tossed together by someone over the weekend um so it's so they're just a mess right now so as I as I as people experiment
366. 51:24with them and find good ones I go play with them but yeah not not really much yet I've been on the waiting list for more
367. 51:31than two months is there another way to get gpt4 yes yes yes yes yes so if you pay there you get to you if you pay for chat
368. 51:42GPT Plus at open ai's website you get chat gpt4 if you're talking about the um API access to gpt4 I still don't have
369. 51:53access to it and I'm cranky about it I want gpt4 API access and I want code interpreter and I want it now
370. 52:05um but I don't have it but if you just want access to chat GPT for to play with just pay for chat gbd Plus or go to
371. 52:13bing.com and that is gpt4 that you can that's connected to the internet so click on the chat button when you go there and that's gpt4 um
372. 52:24so Bing is free um openai.com is not free but openai.com gpt4 is better right now I will link you to a crypto leader on
373. 52:35Tick Tock who would benefit from something okay um read my question please okay Paul I'll get to it generative AI jobs listings have
374. 52:48increased already 20 yeah I have a feeling it's going to start that's going to start uh going up Paul he gets to the
375. 52:55mall did you have another question up here Paul that I missed please tell me I didn't miss Paul's question what's Paul gonna do I don't
376. 53:02know Paul was very upset I think he was crying was that a tear I think he might have been sweating I don't know
377. 53:09um I don't see your question Paul but let me let me go down maybe it's below where you said answer my question please
378. 53:17did I miss a bunch of questions um Paul oh you put a bunch of smiley faces and then I don't see a question Paul
379. 53:34Paul this could be user error I'm I'm sorry Paul I'm afraid I can't do that I still see no Paul questions still see
380. 53:46no Paul questions oh Paul read my question please okay what's your question Paul oh now now just jump me to the bottom
381. 53:55again God damn it tick tock she do Friday night watch some old man try to find questions and Tick Tock scroll
382. 54:10foreign by the way uh if you have questions pop them in the comments and like Paul's I'll try to get to it um
383. 54:27generative AI jobs listings read my question please okay we're back to Paul's thing chat GPT can't make me a pizza it can't
384. 54:35but it will soon be able to chechi PT can order you all the ingredients for a pizza or a frozen pizza from instacart
385. 54:42and have it delivered to your house 252 don't know what that means shrink flation like that womp womp hey Paul
386. 54:53let's get rid of Clarence and steal all those good ideas Paul I don't see a question from you here I'll answer your
387. 55:00question but you ain't got one all right do you know if it's possible to use Claude instant 100K on your own website well
388. 55:11so I haven't played with anthropics apis so I would assume if you go to anthropic.com so they're the they're the ones that build Claude
389. 55:24I don't know their API policy like I don't know if there's a wait list for their API but I assume you get on their
390. 55:29developer program you get access to their API and I assume you can get access to Cloud 100K because Poe got access to it and I
391. 55:39don't think Poe has I assume pose just you know developing [ __ ] so um can you talk about the code interpreter
392. 55:49more I don't know a ton about it but so so what it's going to look like is what you can do is go to YouTube and
393. 55:58search for code interpreter chatgpt but I'll show you where it's going to live so I've seen it so so if you've got
394. 56:07access to the beta applications for gpt4 you've got default browse with Bing and plugins code interpreter is going to be
395. 56:14another mode right another option here and then what happens is next to your send a message there's a little upload
396. 56:21icon and you can upload a like a a spreadsheet um or I assume maybe even put in a URL for a spreadsheet and then I think it
397. 56:33can be up to I think I remember it can be up to 100 megabytes which which is not massive but 100 megabyte spreadsheet
398. 56:40it's a fair amount of data but basically you upload files to this thing it interprets them all and then you interact with it and it's a
399. 56:50full-on data analyst like you can say write me a report about the data in this spreadsheet and it will write you a
400. 56:59report it'll generate python code to create charts and graphs and and then you can say find me some anomalies in
401. 57:06the data and it will do that and you can like like you can just interrogate the data in all sorts of crazy ways um
402. 57:14but but if you want to see like like watching YouTube videos of data analysts who have access to code interpreter like
403. 57:23you watch these data analysts have their Kevin McAllister moment they're like because it's not a tool for data analysts it is a data analyst
404. 57:35it's insane and the fact that I don't have access to it makes me so sad it makes me so sad my fam and friends think I'm crazy when
405. 57:47I talk about AI I'm glad I can tune into your lives I know my friends and family think I'm [ __ ] toast
406. 57:55because they saw me do this in the world wide web and they're they're like again really I have one friend I started my current
407. 58:03company 11 years ago and I went out to lunch with this guy in Boston and he worked for me at agency.com at the early web company and
408. 58:11I told him what I was doing with storybine and we're like halfway through lunch and he just looks at me and goes Kyle
409. 58:18why do you keep doing this to yourself laughs I don't know when you get to add you get a little obsessive on stuff and when you
410. 58:30see things as profound and Home Alone inducing as this AI stuff um I don't know I can't imagine any anything else I would rather be thinking
411. 58:42about right now I mean seriously it's like I think about it most of the day which I know sounds just stupid and
412. 58:52you're like well when don't you have a life yeah I have a life like trying to figure out how in the [ __ ] this is gonna change all
413. 58:59of our lives that's my life right now what's the best way to build a GPT chat bot there's too many sites and it's
414. 59:09overwhelming let's go build one right now shall we oh wait yes waitlisted for a while shrink flation for the small Tootsie
415. 59:18okay my friends and family think I'm crazy okay what's the best way to build a gbt chat bot so you can build them if you if you're a
416. 59:28user of zapier which is an automation tool They just added a new thing called interfaces which is basically like web
417. 59:34access web front ends to to the automations you build and in there they've got a bot maker but we're going
418. 59:40to go over here we're going to go right now to poe.com so if if you're new here that's the official chat gbt website
419. 59:48that's Microsoft's Bing search engine that's got gpt4 sitting underneath it now and then po.com we're gonna go play
420. 59:55we're gonna go make a bot so we're going to flip the camera then we're going to turn off the little Uzi
421. 1:00:03what's it's and then we're going to go over to Poe so this is Poe and so when you get to Poe you get all
422. 1:00:11this [ __ ] down the side the first six things six one two three four five six the first six things are models and then below that are Bots
423. 1:00:20so there's a mid-journey bot there's a bot called I am Shakespeare that's one I created that talks like Shakespeare it
424. 1:00:26only writes an iambic pentameter it's pretty funny um and then you can look at all your bots so I'll go to all my Bots and I got
425. 1:00:34a bunch of them in here I got Jay-Z I got I did a Ricky Bobby bot he just talks about racing and if you
426. 1:00:41ain't first you'd last uh I did a Yoda bot I did a sling blade bot I did Emily Dickinson all the little images of the
427. 1:00:48of the Bots I did a mid journey and I popped them in there I did Mary Shelley Virginia Woolf my Angela so I made a
428. 1:00:55bunch okay so how do you make a bot you go to this little create bot button and you just name it so you just you
429. 1:01:03know name it like uh I don't know like who might be interesting uh uh p ee Wee Herman and then you make a description oh no
430. 1:01:18spaces and then you make a description so like this is the Pee-wee Herman bot and then you write a prompt and so the prompt
431. 1:01:26says you know you you are Pee-wee Herman and you make Pee-wee Herman jokes and when people ask you questions you
432. 1:01:32respond in kind so let me go show you one that I've written we'll go find a fun one here and what's cool is once I wrote one of
433. 1:01:41these you can actually have chat gbt rewrite it for other characters so we'll do Steve Jobs so this is if you go to poe.com I am
434. 1:01:51Steve Jobs that's this bot that I made and The Prompt is um you are Steve Jobs the iconic co-founder of Apple Inc and Visionary
435. 1:02:03leader in technology and business engage with users by sharing your passion for Innovation design and Leadership using
436. 1:02:10authentic language and phrasing your responses should reflect your distinct drive for excellence and blah blah blah
437. 1:02:17as Steve Jobs your memory of your accomplishments and experiences is exceptional allowing you to provide examples
438. 1:02:24right feel free to address the quote the feel free to address questions regarding the principles strategies and
439. 1:02:32philosophies of your style keep these guidelines in mind keep the interactions brief unless a user asks you to provide
440. 1:02:39in-depth mentorship do not say you work for Apple in any way do not refer to your own death provide guidance and and
441. 1:02:46some of those tweaks are just tweaks I put in there after testing it it it just got weird and kind of morbid like well
442. 1:02:53I'm dead now and so so so that's just stuff I put in there but now that I have that written I can literally take that
443. 1:03:00thing and say rewrite this for Shakespeare rewrite this for some other author and then and then you basically
444. 1:03:06do a little intro message and you hit save bot and then you're off to the races you've built a bot it's like super
445. 1:03:12easy it's a great way to practice your prompt crafting um because when you build the bot then you start playing with it like it's a
446. 1:03:20bot and if it's shitty then you prompt sucks and go back and fix your prompt and then test it again
447. 1:03:37I just I love I love the thought of someone just scrolling through Tick Tock and they just watch me losing my [ __ ]
448. 1:03:44mind like what is this what is this I think he might need some help should we send should we send resources
449. 1:03:59hope that helps all right what company is behind chat EBT that I can buy stock shares well I mean I don't
450. 1:04:07I don't think uh openai is publicly investable Microsoft is Microsoft has rolled that [ __ ] across everything they're doing
451. 1:04:16they are in hard so um I am historically a Microsoft Basher I'm an apple Fanboy from from way back um but the last six months Microsoft has
452. 1:04:28impressed the [ __ ] out of me with how quickly they're moving how thoughtful their Integrations of chat gbt are into their stack
453. 1:04:38um and how much Glee they're getting out of [ __ ] with Google it's like a cage match like you know it's gonna be fun watching Elon and and
454. 1:04:46Zuckerberg do the cage match but watching Microsoft [ __ ] with Google's been pretty fun um like I don't know who you should invest
455. 1:04:55in I mean companies that I think are interesting a lot of them you can't invest in um anthropic I think is really
456. 1:05:01interesting they're a bunch of EX open AI employees open AI is chat gbt um Nvidia is making all the chips to to do
457. 1:05:12all this like like the amount of compute power when every company in the world decides they want to make their own large language model
458. 1:05:21like the amount of you know Nvidia chips uh AMD claims they're gonna start competing with Nvidia but Nvidia seems to have the the market
459. 1:05:30cornered for now um it's not easy to spin up chip manufacturing um who else is interesting um I mean
460. 1:05:43I don't think you can bet against Google but they're definitely behind but they're like big have a lot of money and
461. 1:05:49they're really [ __ ] smart people so I wouldn't bet against Google um I think apple is going to be a beast in
462. 1:05:57this Arena I think when Apple comes into the AI game they're going to be able to offer privacy and security like no one else
463. 1:06:04can [ __ ] touch like because they're going to be able to do all their large language model stuff locally on the chip so they're not gonna
464. 1:06:12have to send things off so they're going to be able to promise a level of privacy that I don't think anyone else can touch
465. 1:06:18I think Apple's gonna be big bum oh we got a couple of ADH deers in here yeah did you get your card laminated mine's
466. 1:06:33laminated I have an acrylic plaque such a good club actually as I've gotten older I you know I've just learned that oh you know
467. 1:06:49not only am I shitty at certain stuff and hate it um I'll never be good at it once you accept that and then you go oh there's other [ __ ]
468. 1:06:58I'm really good at right once you accept that it's it's actually kind of freeing this is going to change my life
469. 1:07:06um Gen X I love that commercial okay I'm I'm back in Old comments sorry old man trying to use tick tock here [Laughter]
470. 1:07:16all right oh Ed I'm all the way back to Paul asking me to read his question I don't know what's going on here
471. 1:07:26all right it's 6 a.m I didn't sleep because of AI people think I'm crazy you are crazy but like it's the only place
472. 1:07:33like I woke up one night at four in the morning oh this was wild I woke up at four in the morning and I thought oh I have a domain that I
473. 1:07:46could use for AI and then I popped up and and I had an idea for a site that I could put on it and then I went to Mid journey and I
474. 1:07:55created an interface for it and then I went into illustrator and put a bunch to text on like I built a whole interface
475. 1:08:02for a whole business at like between four and five in the morning one morning and then I went to bed back to sleep
476. 1:08:08before work so I get it I've been there I've yeah ADHD Dreamland oh yeah it makes so much sense cool I have adhd2
477. 1:08:18and I'm listening to your Live While holding laundry perfect yes please multitask or just you know mono task or
478. 1:08:24you know whatever you do uh but but yeah uh if you are new to this stuff uh go to chat.openai.com play with chatgpt until
479. 1:08:34you have a Kevin McAllister moment and if you don't know who Kevin McAllister is I resent you because it means you're young
480. 1:08:42and you don't have hair and facial situations going on like I do [Laughter] um Can the Bots be on your website they
481. 1:08:53can uh but well it depends what you mean by the Bots can you put chat Bots there there's lots of tools to do that now you
482. 1:09:01can do it with zapier um like the the I've been I've made some tools with zapier where I just have a
483. 1:09:09Google form and then I capture the embed code for that Google form and put that into a website but there's all sorts of
484. 1:09:16ways you can get it in there um so the answer is yes you just need to dig around if you're not a coder either
485. 1:09:22find a coder or just start looking up no code tools like zapier there's one called make um I don't know look up no code stuff no
486. 1:09:32code stuff's good but um it it's still complicated you work a lot Kyle I do I do but I like it like like work like this kind of work
487. 1:09:45is my hobby and then if I can turn that into cash dollar you could make money
488. 1:10:04yes they can I made four yesterday FAQ Bots business oriented cool security security is a problem right now um
489. 1:10:13if you're using chat GPT so if you're using GPT playground and playing with their apis that's better if you're using chat GPT you can turn
490. 1:10:22off your chat history and your uh you can tell them not to train on your data which helps but it will delete your chat
491. 1:10:30history so make sure you download your data first um and then and then there's just the the whole security thing is just a whole
492. 1:10:39much deeper conversation for big companies right now where most big companies even medium-sized companies are going to need
493. 1:10:49to go is to have their own kind of Standalone large language models that is trained on their data there's a thing
494. 1:10:57called Lang chain where you can do that um or you might want to run open source models locally there's there's all sorts
495. 1:11:05of issues with security um a lot of these tools are very very spit and duct tape right now so you should know that so um
496. 1:11:15yes learn more about that I'm not a security expert but they're basically they're like swiss cheese they have holes all over them so um
497. 1:11:24digital Danny thinks I'm cool awesome appreciate it digital Danny how can we protect the conversation and
498. 1:11:31information like I said you can turn off your chat history and tell them not to train on it that's the best way to do it
499. 1:11:36or run things locally like on your local machine or in a standalone virtual machine like Google collab uh or or in Microsoft Azure that's
500. 1:11:47basically you need to firewall the data in some way but turning off your chat history and telling them not to train on it will help
501. 1:11:56do they have what was that question and they have so much data yeah so does Google so does Facebook we could listen
502. 1:12:04like everyone's like ah they're state they're taking our data and they're training their box on our data well we
503. 1:12:10gave our data to Google and Facebook and Twitter and [ __ ] everything else YouTube ancestry.com we we uploaded our DNA to
504. 1:12:21the internet our data is the long since gone um the scary thing with this AI stuff is now you got AI tools where you can
505. 1:12:31synthesize that stuff like that so that's the scary thing poster child for ADHD ADHD is a superpower I agree
506. 1:12:42um my buddy Peter shankman wrote a book called uh faster than normal which talks about the ADHD brain and how she's like
507. 1:12:51well not that part the frontal lobe is basically a block of cheese that doesn't do anything we have no executive
508. 1:12:57function up here but all the synthesis [ __ ] and the the pro the the the you know sort of the fast input [ __ ] we're good at that
509. 1:13:06but uh what was what did I do the other day I wanted to take in a my phone charger which I had in my bedroom
510. 1:13:15so I woke up in the morning I'm like okay gotta remember this I took it down I put it in the kitchen counter when
511. 1:13:21like if I kept putting it everywhere I went and then I I went into the living room and I sat on a chair and I put it
512. 1:13:27on the arm of the living room chair and I put on my shoes and I get in the car and I go to work and I don't have the charger
513. 1:13:34and I get home and it's on the it's on the armchair like it's you know the executive function thing just flush any
514. 1:13:42expectations of that done the toilet and enjoy the rest of the [ __ ] that you're good at workaholism is classic ADHD it's a
515. 1:13:53blessing love it yep absolutely agree um flow wise is pretty neat no code okay I don't know flowwise I'll check that out
516. 1:14:01is it expensive to run things locally processing power wise well um not necessarily if you've got an M1 Mac
517. 1:14:11if you've got like a MacBook Pro M1 or M2 there are some open source projects that run on those pretty pretty well the the the the
518. 1:14:21developer Community for the open source models they're doing all sorts of really interesting [ __ ] where they're
519. 1:14:27um you know compressing these models making them smaller pulling data out of them but they still work um so
520. 1:14:34I'm pretty sure there was one that you could install on a phone locally and it would run um you there's definitely ones you can
521. 1:14:42install on a Mac if you've got a PC with like a 30 90 or better GPU you can do some pretty intense processing but
522. 1:14:50they're starting to get these things to the point that you don't need you don't need really Banger machines to do this anymore
523. 1:14:56um they're not going to be as powerful as as they're going to be probably just close to GPT 3.5 they're not gonna
524. 1:15:03they're not gonna be like gpt4 it it's just of another class um so I would I would say just go start hanging out on Reddit in the in the
525. 1:15:15um subreddits for like large language models open source large language models there's a lot of projects there's a
526. 1:15:20project called GPT for all that's probably worth checking out because I'm pretty sure that's got an installer for Windows Mac and Linux
527. 1:15:28that installs you know some local instance on your machine and then you can choose from different open source
528. 1:15:35large language models to play with uh um that's why they will succeed chat is behind Google may have my data they earned it I
529. 1:15:49know that's I was I was bitching to a friend of mine that worked at Google once about I don't know something Gmail
530. 1:15:55sucked at and he goes what did you pay for that I'm like oh yeah if the product's free you're the product uh M chips with their unified
531. 1:16:06architecture was very forward-looking yeah I agree the the uh the apples apples um chip architecture if you paid
532. 1:16:16attention to their Vision Pro announcement with the R1 chip so they got the M2 Chip on one over one eye and
533. 1:16:24the R1 chip over the other eye and the R1 chip is reading all the sensors on the thing so it's doing real time like
534. 1:16:31machine learning [ __ ] so they're doing all that machine learning stuff for those goggles on the device
535. 1:16:37Apple's gonna [ __ ] kill it with this AI stuff all right looks like looks like I've blown through the questions I never
536. 1:16:44answered Paul's question but Paul I never saw your question I don't know if you're still here but anyway if you've
537. 1:16:49got some questions pop them in it's getting a little late I want to go watch the bear I've watched two episodes now of the bear it's
538. 1:16:56freaking awesome here's some TV show recommendations it's Friday night so um drops of God really good the bear is
539. 1:17:04[ __ ] incredible it's just so good um what else do you if you haven't seen Ted lasso you have to see Ted lasso
540. 1:17:13the the just like the three season package it it starts out weird you're like why am I watching this show about an idiot redneck in England
541. 1:17:23and then by the third episode you [ __ ] love them all um those are three good ones right now the bear yeah the Bear in Chicago
542. 1:17:33the bear had a dream of creating a site with multiple Bots to focus group your document documents that's pretty cool idea
543. 1:17:42but I had a I had an idea for business today what the hell was it me he he the hell was it oh it was just some sort
544. 1:17:57of framework thing it doesn't matter it's just it's like the thing that's just amazing about these large language
545. 1:18:02models is you really can have a concept for a business and rather than thinking about oh I need to get a developer I
546. 1:18:09need to do this need to do that you can be prototyping it in chat GPT or or in zappy or something like that within an
547. 1:18:15hour and it's like [ __ ] world class like it it's just mind-blowing this the speed with which you'll be able to take
548. 1:18:23a raw business idea and spin it up is it was already fast like with Site Builders and you know no code tools but the the
549. 1:18:34large language models just take it to a completely different level all right uh Ted lasso Ted lasso is like a big
550. 1:18:46warm hug or like a warm hug it is totally like a warm hug Ted lasso soundtrack and lesson learned is so good the dart scene when when when
551. 1:18:59Ted is doing the dart scene about people underestimating him and not being curious holy [ __ ] [ __ ] that's what
552. 1:19:06this this channel is about be curious go there and be curious go to prompts.chat figure out what this prompting nonsense
553. 1:19:14is all about go start sticking sticking them in there and get curious and if you haven't seen the Ted lasso
554. 1:19:21scene where he talks about being curious and how people aren't curious it's one of the best scenes in TV it's so good that scene is so good
555. 1:19:30it's such a [ __ ] you to all the bullies oh it's so good um the ultimate queer love is so good okay I'll check that out
556. 1:19:39um any bot that can read a PDF well yeah there's there's plugins that read PDFs I use them all the time that's one of the that's one of the
557. 1:19:48someone asked before about are there any good plugins there's not a lot but there's one that I use called ask your
558. 1:19:56PDF that works really good and you can just give it the URL of a PDF it'll read it and then you query against it it's really good
559. 1:20:05but you need access to plugins so that's the paid version of chat.openai.com Bing has announced Microsoft has
560. 1:20:14announced that Bing will support chat GPT plugins so if you wait three months six months I don't know
561. 1:20:22bing.com will be turning on plugins and so you'll probably get to use them for free as my guess I'm a complete rookie what's the best
562. 1:20:30way for me to get started that's why this channel exists that's all of this I'm like Carol Merrill remember Carol
563. 1:20:37Merrill again Gen X reference what are you gonna do old man on a tick tock what are we gonna do um you're brand new to this
564. 1:20:47if you haven't heard of chat GPT now you have it's a thing you go you get to it by just going to that website just go to chat.openai.com
565. 1:20:57it's basically just like a Google search bar but instead of searching you start interacting with this thing and it and
566. 1:21:04and what I like to say is start with uh parlor tricks and the parlor tricks are things like describe your significant other or your
567. 1:21:13pet and have it write a country song about that person or a rap or a poem or whatever a short story and it will and
568. 1:21:21you'll go oh that's pretty good and then tell it to make a table comparing I don't know three different sports teams
569. 1:21:27and it will and then you'll be like oh I didn't know I could do that and then tell it to write some computer code and
570. 1:21:33it will do that and then it's at some point you'll just go like holy [ __ ] like what what do I actually do with this
571. 1:21:40thing then once you get to the this is intriguing now I'm intrigued go here to this thing prompts.chat and that'll
572. 1:21:49teach you about prompting and kind of how these machines work and how they think and then this is a giant long list of
573. 1:21:55personas that you can give to chat GPT so act like a screenwriter act like a mathematician act like a travel agent
574. 1:22:02act like whatever a nutritionist and you'll you'll get a real sense of oh this thing like can kind of shapeshift
575. 1:22:09it can be lots of different things and so this is it's not so much where you're going to learn to prompt but this is
576. 1:22:15what will inspire you to start thinking about oh maybe it can do this maybe it can do that maybe it can do that so
577. 1:22:22that's phase two and then phase three is what stuff that you're passionate about or what's a project that you're working on
578. 1:22:30and think of Chachi PT like the most capable assistant you've ever had like it's way smarter than you it's it's
579. 1:22:39dumb in terms of life experience you have to tell it everything like okay chat gbt now you're gonna do this so you
580. 1:22:46have to you have to guide it so you're the pilot it's the co-pilot but it's the smartest [ __ ] co-pilot you've ever
581. 1:22:51worked with it's the smartest person you've ever worked with by far and then and then start solving problems right so
582. 1:22:59it could be a personal problem like I want to eat better so have it create a meal plan for the week with with a new kind of diet
583. 1:23:05um it might be a work problem have it come up with a marketing plan for your business or have it help you solve I
584. 1:23:11don't know some sort of customer problem and then once you start sort of pinging it to solve stuff for
585. 1:23:17you at some point in there you're gonna have the Kevin McAllister moment and then you're hooked and then at that
586. 1:23:27point then you're up till six in the morning working on this stupid [ __ ] and you listen to me on these lives
587. 1:23:35you joined the [ __ ] Insanity Club speaking of the insanity Club here let me show you a group that I started full
588. 1:23:41of insane people no full of actually just really cool people that are very curious about AI so it's called the AI
589. 1:23:47salon so if you go to the salon.ai that's a link tree that will point you to our Meetup we meet every other week
590. 1:23:54virtually and in person here in Denver there's also a Discord server there there's also also a LinkedIn group there
591. 1:24:00so go there and then this thing is a newsletter that I do with my friend Greg mushin called everyday AI news so that's
592. 1:24:08our Twitter handle so you can go there and we do cool tweet threads and [ __ ] like that so anyway welcome complete
593. 1:24:14rookie what is your name David something like that cool welcome to the nightmare it's going to be exciting and then you'll be terrified
594. 1:24:25and then you'll be excited and then you'll be terrified and then you'll be excited and then you'll be terrified and
595. 1:24:31then your friends will stop speaking to you and then at some point one of your friends will pull you aside and say
596. 1:24:36Could you teach me a little bit about chat gbt and then you'll have a fun time with them and then they'll be hooked too
597. 1:24:45it's gonna go something like that or you play with it once and you're like it made a mistake it's stupid I think it's
598. 1:24:52a waste of time this is just a hype cycle you could be that guy [Laughter] later skater thanks for streaming you
599. 1:25:02got it Jim Buckley oh cool hey Jim uh oh good lord all right I'm gonna go soon here we live in amazing times these are
600. 1:25:12amazing times I I you know on a on a a serious note as I can muster up um I feel so [ __ ] lucky to be alive right now
601. 1:25:24like these are the most powerful tools in the history of humanity like the internet was cool but in the end the
602. 1:25:32internet was just like here's some [ __ ] here's some other [ __ ] let's connect it this stuff is like a reflection of
603. 1:25:40humanity back in our face it's like it's like a Humanity amplifier it is it it's honest to God like I don't I don't know
604. 1:25:52we're very lucky to live in this time and if you if you're lucky enough to get curious about this [ __ ] get curious and curious and curious
605. 1:26:00it's going to serve you well I promise you the cloud 100K is so so so much better than plug-ins yeah the the cloud 100K
606. 1:26:11um thing is is it's a big deal Carol Merrill all right some someone knows the caramel reference thank you Silver Fox all right
607. 1:26:23one of us one of us exactly Kevin McAllister moments never stop exactly that's what it's been like honest to God
608. 1:26:31the last six months for me since chat ebt's come out it's been six months of okay the the first three months it was
609. 1:26:41multiple times a day it slowed down a bit so now it's five times a week I'm doing this no I'm gonna go look at it watch the
610. 1:26:54video holy [ __ ] five times a week it's insane and sometimes it's just I'll talk to someone and they're like oh
611. 1:27:06here's how I use chat GPT and it's a way I never thought of it that's the other thing for for those of you that are brand new to this [ __ ]
612. 1:27:16um if if if they just stop development of chat GPT right now and you just went to the free version of chat.openai.com the 3.5 version
613. 1:27:29I confidently feel I could spend the next decade trying to figure out all the [ __ ] that that makes possible just that
614. 1:27:40the next decade with no advancements and there's like advancements multiple times a week significant event advancements
615. 1:27:47multiple times a week we are in an exponential Innovation curve right now where these tools are informing other
616. 1:27:55tools and developers are taking what these two developers did and they're rolling it into their new product and
617. 1:28:01then the next pair of developers is rolling it into theirs it's it is insane what's happening right now so the Twitter handle oh that other
618. 1:28:12Twitter handle hold please is my newsletter called everyday AI so that's the Twitter handle for everyday AI news we do tweet
619. 1:28:22threads there like I I just the the most recent tweet thread there yesterday Mackenzie came out with a big report on
620. 1:28:28the future of generative Ai and how much [ __ ] money it's going to generate it's like four point 4.2 trillion or I don't know 3.4
621. 1:28:38trillion I don't know some stupid large amount of money but it's in it's in the thread uh uh my photos are all screen
622. 1:28:48recordings of you oh God I'm so sorry I'm gonna take me away well I'm gonna try living in this [ __ ] brain
623. 1:29:00um all right I gotta go everyone have a happy Friday night thanks for hanging out with me y'all are swell and uh yeah
624. 1:29:07stay curious go play with this stuff I'll let me scooch out of the way here if you're new here screenshot this or go grab some of that
625. 1:29:14stuff have a good weekend everybody I'll probably be on tomorrow because this is what I do now good night everyone