aiLL 8/5/23 - The Future of Work in an AI World

Transcript

1. 0:01foreign
2. 0:33Silver Fox what's happening all right it is Saturday night I just tried to sign up for a service and
3. 0:48you had to do it through PayPal I haven't logged into PayPal in Forever when it said I wasn't verified but I
4. 0:55have all my crap verified there so [Laughter] I hate it when crap doesn't work uh welcome to AI learning lab everybody hey hey
5. 1:09um happy Saturday night my name is Kyle Shannon we'll get this party started in a little moment I have exciting news I started a YouTube
6. 1:21channel for Ye Olde learning lab so I'm gonna put my uh the recordings of these and I'll probably put over
7. 1:30everything if I can get it all automated I couldn't automate downloading the lives I don't think I might be able to but I couldn't with
8. 1:39whatever tool was recommended last night but I'll show that I think that's a good thing I think that's a good thing I actually thought
9. 1:51it'll be cool to use um maybe we can even do that tonight I wonder if any of the transcripts are there yet I figured I could use um
10. 2:06let's see seven hours ago I figured I could use Ye Olde chat gbt show transcript yeah all right the transcripts there yeah so
11. 2:21we can go play with this tonight maybe we'll go maybe we'll go experiment we'll go spearmint with uh chat gbt because
12. 2:29what I thought would be cool is use chat gbt to go look at the transcript of these long ass ramblings
13. 2:39and come up with a title for it come up with a description and then come up with a uh a set of chapters with time code in them
14. 2:48I think that would be cool I've been screen recording a lot of your stuff and I'm working on an app to organize stuff wow that's awesome
15. 2:58you're gonna sell me back my app that's got my crap organized that would be nice that's awesome oh that's hilarious
16. 3:09um that's cool yes definitely um yes send me what you're working on I I always like seeing seeing that because
17. 3:16it's there's so many ways to skin the cat with this stuff so I love I love seeing different things no charge thank you I appreciate that
18. 3:26I appreciate that um oh wow a bunch of people here all right welcome so here's what this channel is about my name is Kyle Shannon
19. 3:33I'm a long time entrepreneur I've started 12 or 15 companies I'm currently the CEO of a company called storybine
20. 3:40the founder and the CEO of that uh we've been around for 11 years we do automated video storytelling so people answer a
21. 3:48series of questions in an app that goes up to the cloud a few minutes later you got a fully edited video and uh
22. 3:56over the past years I've been digging into this AI stuff I'm completely rethinking my value proposition and the
23. 4:02product roadmap so that's been absolutely insane um I was also an entrepreneur uh a startup guy back in the early days of the world wide
24. 4:12web and what's going on with this generative AI stuff the parallels between what was happening with the early days of the web are absolutely
25. 4:20insane so this channel is all about uh just talking about this generative AI stuff it's it's very new it's very raw it's very profound um
26. 4:31this list is now seven this use this when I first started the channel there was one site here it was it was chat GPT
27. 4:39and then I added some others it was three for like two months three months and now it's seven so the the competition is heating up and
28. 4:48these are all just kind of chat flavors of chat gbt or competitors um so there's a lot going on so if
29. 4:56you're if you're new to generative AI if you haven't played with it that's what this list is all about so let me let me run through these
30. 5:10all right so go ahead and take yield screenshot if you're new to this stuff if you're just getting started you don't want to think about
31. 5:21too much just go to the top one go to chat.openai.com that's the official chatgpt website the 3.5 version is free it's really good
32. 5:32um if you pay the 20 bucks a month there you get gpt4 and you get a couple other things that I think are worth it I'm not
33. 5:38getting paid by them so I don't give a [ __ ] if you do that or not um then you've got Bing and Claude which
34. 5:43are kind of the next two down that that I would I would consider Bing is Microsoft search engine but if you click
35. 5:50on the chat button that's gpt4 uh you used to have to do it within Edge but I think they've opened it up to
36. 5:57Chrome and Safari now it is connected to the internet it can generate images it can read images uh it's it's decent it's not as good as
37. 6:08the in in terms of the the content creation it's not as good as the gpt4 you get here but it's free then
38. 6:16you've got Claude AI over on the left there that is anthropics um Claude 2. and it has a hundred thousand token context window which if
39. 6:27you're new to this stuff just doesn't matter all you need to know is you could copy and paste an entire novel into your
40. 6:35chat prompt and start interacting with it like tell it to summarize the novel um claude's really interesting
41. 6:43um Bard is Google's entry into this space this did not used to be on the list because it was embarrassingly bad
42. 6:49for a while it's it's getting better and I thought that they actually rolled a new large language model called Gemini
43. 6:57into it I don't think they have now that I now that I've said that in about 30 different places but it's much better
44. 7:04than it used to be and when they rolled Gemini out I think it's going to get really good uh labs.perplexity.ai that's actually you
45. 7:13can those are uh the metas Llama Llama 2 models uh are being hosted by perplexity that you can play with for free
46. 7:23um so you don't have to install them and do all that happy horse [ __ ] and then pie in the corner is from a company
47. 7:29called inflection it's very very chatty if you actually want to have a conversation with a large language model
48. 7:34that's a good one and then Poe you can play with a bunch of different large language models and it's got Bots that
49. 7:41you can make it's got a bot maker and really interesting just just in it both of the bottom ones feel a little
50. 7:50more consumery where all these upper ones feel more professional all right so that's going on there okay what else
51. 8:00I got a lot this is like this is like homework this is the homework portion of the evening happy Saturday night let's
52. 8:06go on ahead and get fired up with uh homework okay so here's your homework if you're new to this crap um prompts.chat will teach you about
53. 8:17prompting it will also give you a whole list of personas that you can tell Chachi PD to act like certain things
54. 8:25like act like an accountant act like a screenwriter act like an idiot and it will um and then openart.ai promptbook
55. 8:33that'll teach you how to prompt these image generation tools and I I recommend these three right here if you're just
56. 8:39getting started and then futurepedia.io will give you just a a big old pile of AI tools that you can go geek out on
57. 8:48um I have some news I have news we now have um a YouTube channel that I just started today and AI learning lab was taken so it's
58. 9:03YouTube slash um AI learning lab Dash TT for an old tickety talk and uh we can go over there and take a look at it
59. 9:16look I made I made a little uh a little Banner AI learning lab and I put my little eyeball dude in there and I uploaded
60. 9:27nine and then it it uh rate limited me it wouldn't let me upload more than nines I have nine up there
61. 9:34um so those are the those are the early ones if you want to see hey how fat I was when I started this channel three
62. 9:40months ago I've lost some weight uh I I look a little beefy I'm chubbier than the dog in these in these videos uh you
63. 9:50can go check that out but anyway people have been asking you know if these uh if these lives are being saved they're
64. 9:55going to be I need to find a way to automate it uh right now it's a manual pain in the ass it's a pain in the knee is that tell you opinions
65. 10:08um what I figured we could do tonight also is um the videos that I uploaded have transcripts with them so I figured we
66. 10:18could go play with some plugins that go look at YouTube videos and maybe we could write descriptions of um
67. 10:28descriptions of of the uh of the old lives let's see how good it does at making sense of me just rambling for two
68. 10:36hours maybe it can chapterize them in a decent way and actually make those things usable and discoverable and interesting
69. 10:44that would be nice wouldn't it wouldn't it be nice if this AI technology actually helped us instead of destroyed our lives and
70. 10:52killed all of humanity might be good all right so what do we do here I don't know a ramble I swear I
71. 11:01make jokes I make voices occasionally my jokes and voices get me kicked off this thing for seven days tapping three times
72. 11:08now I'm gonna try to avoid that tonight um and if you've got questions about AI pop them in the comments below I do spend
73. 11:17most of the time on these answering questions or comments or thoughts that you have down below um I think this stuff's really really
74. 11:25important I'm really passionate about it passionate enough that I hop on here for a couple hours every night instead of
75. 11:30doing Sim racing which I love and I'm giving up Sim Racing for you for you to get you up to speed on all this it's crazy all right
76. 11:43let's see let me go back and see what questions we have here what comments we have no comments about the hair although
77. 11:50it's it looks like it's about ready to take off it could go bad pretty quickly here tonight all right all right
78. 11:59let's see you remind me of Tom Arnold I'm sure you've heard that before I have Tom Arnold Rosie O'Donnell uh fat Alec Baldwin
79. 12:12um Philip Seymour Hoffman uh uh um wait who's the other guy John Goodman skinny uh skinny John Goodman uh
80. 12:26someone said Leonardo DiCaprio once and I thought well yeah Leonardo Leonardo DiCaprio if he made much worse life decisions Maybe laughs
81. 12:38if he didn't get that big break when he was you know who's Eating Gilbert Grape uh I've heard some of that stuff before
82. 12:49um yeah that's cool uh add it to Amazon Bedrock what's add what to it I assume uh that the uh llama llama two Labs without perplexity
83. 13:03is that what you're talking about yeah Amazon bedrock's interesting I'm I'm uh I'm building a I'm going to New York in
84. 13:12like 10 days yeah yeah eight days I'm going to New York in eight days to talk to a big Pharma client about
85. 13:21building some some tools for them and uh I think Amazon Bedrock might be a might be a big piece of that I don't
86. 13:29know if it'll be a part of the conversation it might be a big part of how we solve the problem Amazon Bedrock is you can use it's
87. 13:38basically you can hook all your AWS [ __ ] up to large language models and use different models and things like that
88. 13:46um I don't know how far along it is or how good or bad it is I haven't really dug deep into it yet uh um did I see that Skype
89. 13:59now has Bing chat built into it no I haven't seen that that's interesting let me go let me go looky looky at Skype
90. 14:07and see how how much they're hyping it or if they just kind of snuck it in there
91. 14:21Hawaii Five-O someone someone include me and that song just popped in my head last night I tend to have ADD singing
92. 14:29moments when I'm going and searching for things introducing new Bing in Skype interesting do they have a little video of it get social
93. 14:42wait sorry get social with the new Bing in Skype the new Bing in Skype is your AI powered co-pilot can provide helpful answers to your
94. 14:51questions whether you need a quick answer on something or something more in depth learn more they got a cool little video for us to watch
95. 15:02download Skype now just ask Bing hmm interesting all right yeah so it looks like they just popped it in there as part of the
96. 15:15refined things answers they've got it writing poems well we're so [ __ ] early in this in all this all this stuff being integrated
97. 15:29and so a lot of it's just parlor tricks right now look you can make poems yeah okay um interesting well I I don't know
98. 15:41anything about it so I'm seeing it for the first time now um but but that's why I think the there's an important Point here actually um
99. 15:50the the the thing the thing that's significant about that is is you know it's it's further evidence that um these tools are just going to be
100. 16:00ubiquitous they're going to be kind of like uh the World Wide Web where just you know it's it's out there for
101. 16:06everyone to use it's going to get rolled into all the you know all the software we already use every day
102. 16:12um it's not going to be this thing off to the side like it is right now like you know chat gbt is like this you know
103. 16:18we've got you know AI detectors and you know we we've got to figure out if it was written by AI well no we don't it like if
104. 16:30if we're doing things like writing articles that have citations in them as human beings we need to confirm that the
105. 16:36citations are correct because these things lie um but but in the end it doesn't really matter if a human wrote it or if this
106. 16:43wrote it if it gives you value right that was Google's initial response to Chachi PT taking off was you know we're gonna
107. 16:51de-prioritize anything that was written by Chach EBT and then they were like yeah no we're not going to do that if if
108. 16:57it adds value if it's valuable content it's cool if it's shitty content then we'll de-prioritize it and I think
109. 17:04that's the I think that's the the right the right approach but this stuff's going to be everywhere so it behooves
110. 17:10all of us uh to get much more um uh AI literate in the next year we'll we'll pay dividends in the next year year and a half
111. 17:22um if you would since I'm just getting this started tonight if you're willing to share this live with your friends and
112. 17:28let them call you a weirdos for playing with this AI [ __ ] that would be great uh and you know tap the uh
113. 17:35double tap the screen get some hearts going maybe we can get some more folks in here that'd be awesome uh um
114. 17:42what else what other questions do we have what's the best one to use on an iPhone um the the official chat gbt in my opinion is just
115. 17:54it's just the best so I there there is an official iPhone app from openai for chat GPT um it's a very simple website so it
116. 18:04actually I just have it as a uh a bookmark to the website on my home page I actually have it in the bottom four
117. 18:10like it's one of my primary buttons on my iPhone um the other one that's really interesting here is pi dot AI because
118. 18:19you can put it in voice mode and just talk to it and then you pick one of six voices and it talks back to you and it's
119. 18:27connected to the internet um chat EPT is not connected to the internet although with plugins it kinda is and I can I can show that later
120. 18:36um I would say for the iPhone right now chat GPT clawed just because it's got a really big context window and that might be
121. 18:46useful if you want to copy and paste big documents even on your iPhone and then uh Pi dot AI the thing about Pi dot AI
122. 18:55if you want to like like like the way I've been using pi.ai on my iPhone is I get in the car I turn on the Bluetooth
123. 19:01and I turn on pi and just as I'm driving places I just start having conversations with with this large language model and it's
124. 19:09fascinating it's very very chatty so if you actually want to get [ __ ] done don't use Pi it's just too chatty for that
125. 19:16it's like haha isn't that clever Kyle you know it's very funny um but but to have a conversation with
126. 19:25it it's it's actually quite interesting you know after whatever it's been 10 15 20 years of things like
127. 19:32um Alexa and apples talk to me crap have sucked um it doesn't suck it's actually you know it's a little awkward
128. 19:43um I've got I've got it set to some Australian woman voice and she pauses a little too long before she answers but
129. 19:50it's pretty interesting so that's what I play with all right um pie pies tasty Chachi BT available for Android Android uh iPhone and
130. 20:03Android yeah they just announced Android if you're gonna download the chat GPT apps be very very careful I would
131. 20:13strongly recommend you go to openai.com and click on the links for Android or iOS from their site rather than searching for it in the app
132. 20:24store or in Google Play because there are so many copycats that are like just trying to get your money that are
133. 20:31basically charging you money to give you free GPT 3.5 so don't fall for that [ __ ] little [ __ ] scammers are out there ah we can
134. 20:40make money off to clueless people so go there um but yes they're available for both the the challenge right now with uh with
135. 20:52the apps from open AI although I haven't played with them to be fair in the past three or four weeks um they don't have all the features of
136. 21:00the website so you just need to check you know go to the website see what features are there what's available and
137. 21:08then see what's available on the uh on the apps what I will say is that the voice recognition in chat GPT the official app uses their
138. 21:21whisper protocol the for to do the trip the the transcription of your voice it's really good so um yeah so I would you know
139. 21:33you could you could use it to do a voice transcription much more accurately than the the iPhone local stuff
140. 21:40good good good little life hack there huh delivering value right off the bat right off the bat at the lab the AI learning
141. 21:48lab what we do here is all value all the time it's just it's a Relentless pursuit of value of Excellence of um wow wow
142. 22:03All Right Moving On they have an app now yes they do just jumped me to the bottom if you're new here it's one of my pet
143. 22:15peeves is Tick Tock jumping me to the bottom I like to answer all the questions I don't like to be at the
144. 22:21bottom I like to be in a place where you don't think I'm actually going to get to your question and chances are good I won't [Laughter]
145. 22:30ah he's so silly he's very silly he's silly I don't you know I come in here and and I just I put it on and it I like
146. 22:41to go to sleep to his voice I like his voice his voice is really nice and uh sometimes he'll make me laugh other
147. 22:48times it's a little cringy a lot of dad jokes a lot of dad jokes he's not really qualified to talk about this stuff but
148. 22:56he seems to like doing it so uh yes I think it's good all right I've never used that but that voice before ever in my life kind of hurt
149. 23:12[Laughter] Tobias here with his plus one awesome Tobias is your plus one a puppy or a human or a kitty cat or a goldfish
150. 23:26welcome welcome welcome what'd you do on date night uh I want to talk about it no you didn't tell me that's probably
151. 23:38best if you don't know now tell me and we watched this guy on Tick Tock for two hours what [Laughter]
152. 23:52uh down I downloaded a chat gbt and it was a fake I couldn't close it uninstall it oh that sucks bolognes I know that's that's
153. 24:07again like why I put these here this is not for me to make money I'm not getting paid by any of these [ __ ] um they should pay me
154. 24:16you know all we do all they just sit here and talk about them someone should be writing a check but um
155. 24:23yeah that sucks it's one of the reasons I have them here because there's just so many scammers out there trying to
156. 24:29trying to take advantage of the uh excitement about the future of communication um what game do you use for Sim racing
157. 24:38assetto Corsa competition GT3 stuff and I raced the McLaren I'm thinking of switching I might switch
158. 24:52to the Honda and I might switch to a Porsche but I don't I don't think the Porsches are that good they're easy to lose lose control of
159. 25:03what's good what's good uh oh here's something that's good this is new um people have been asking but Kyle do
160. 25:12you have a YouTube channel and I have not until today we do now self was already taken by someone else the bastard
161. 25:22um so it's YouTube come on Focus you jerk ah come on you can do it
162. 25:37anybody know Tim Cook um YouTube is at AI learning lab slash TT as in Tick Tock huh clever so I've got nine uh I I have the first
163. 25:53nine lives up there at some point I'll get up all the tick tocks up there and I'll get i'll get all the lives up there
164. 25:59and then hopefully I can automate this [ __ ] because it's a lot um but but we'll go I I think we're gonna go play one of the things I'm
165. 26:06gonna do tonight is I'm gonna go see if we can make a a prompt that will give us a YouTube description with chapters from these
166. 26:16tick tocks from The Tick Tock live transcripts so that should be fun so we can do that together we can do a little
167. 26:21we can do a little prop crafting together and we can learn we can learn things together that's fantastic
168. 26:31yeah it seems he's he's like a teacher but but you know a little he's he's attached he's a little touched he's touched
169. 26:44all right you're behind I am behind and I will be behind I will get increasingly behind as more people come in here and
170. 26:51ask me questions thank you Robert I appreciate the donut it's obvious that I appreciate Donuts I'm trying man I'm losing weight slowly
171. 27:06Donuts is tasty um encourage more viewers to share yeah viewers I'm encouraging you um Can AI build similar HTML Pages based
172. 27:19on the design of another page yes it's really good at that [ __ ] if you grab you know your CSS style sheets and your HTML
173. 27:27of a thing and then you say you know based on this give me a page that says that and you can have it write the content
174. 27:36generate the HTML update to CSS if you need to it's quite good at that I've even done it where where um
175. 27:44for like bulk emails um you know you can do you can send plain emails which look like [ __ ] or you can send HTML emails
176. 27:52well HTML emails are very very particular because email programs still can't handle much HTML so it understands
177. 28:02that so you can just say take this shitty looking email and format it using HTML that's appropriate for this
178. 28:09particular service and it it does it good and then and then rather than going into the HTML and editing [ __ ] yourself you
179. 28:18can just go yeah make the Line Orange now make it a darker orange or you know here's a hex code or recommend a hex
180. 28:25code and go change it you know throughout the throughout the HTML it'll just go do it it's really good at that
181. 28:31um and of of all of these I would say um chat GPT gpt4 is probably best but I would experiment with with Bard I think
182. 28:40Bard is actually decent at coding which makes sense considering it's from Google was literally wait will literally invent
183. 28:49citations for sources that don't exist it will so there's a pretty famous lawyer with a with a Scarlet eye on his forehead for idiot
184. 29:01that had Chachi Piti write a legal opinion for him and he handed it to a [ __ ] judge and the judge actually clicked on the
185. 29:11citations and said uh sir it appears that your citations are non-existent which you care to explain [Laughter] and I will argue
186. 29:24up one side and down the other that's not Chachi PT's fault that's that lawyer's fault if if you're doing things like oh say
187. 29:33trying to prove a case do your own work you know verify your own work um hang on but it is important to know all of these
188. 29:43things do what's called hallucinating it's the the developers you know they came up with that term because it's
189. 29:49better than you know it will [ __ ] you I like to describe these things as mansplaining as a service um I didn't come up with that but I I
190. 29:58love it um they will confidently answer you even if they're wrong like every good man in your life um the the way they work
191. 30:13what it looks like they're doing it looks like they're a sentient being on the other side of the computer just
192. 30:19typing all this remarkable crap they don't actually understand what they wrote what they're doing is using a
193. 30:25probability engine figuring out one word at a time what's the next most probable word so after you write your prompt it
194. 30:33sort of creates a mathematical waiting it goes into this insanely complicated Vector space and it picks out a vector
195. 30:41number that's associated with a word and it plops it in there and then it does that over and over and over again so but
196. 30:47it but what it doesn't do is it doesn't go back and look at what it wrote and understand if it's right or wrong so it
197. 30:52has no [ __ ] clue if what it gave you is real or not and you can tell it oh you got it wrong could you please get it
198. 30:59right now it'll be like Oh I'm terribly sorry let me try to that again and it'll get it wrong again sometimes so so you
199. 31:06just need to be conscious of that um doesn't mean don't use them these are the least capable they will ever be
200. 31:15right now the the speed with which these things are improving is absolutely insane so don't dismiss them just
201. 31:22because they're not perfect right now I know nothing about AI or chat GPT where do I begin and where do I go all right Mama LA
202. 31:29let's get you going if you're just getting started I would just go to the top one don't worry about all these
203. 31:34other ones just go to chat.openai.com and go to prompts.chat that very top URL that's a document that will explain to
204. 31:46you what these large language models are a little bit about how they work there's little primers in there for free and
205. 31:52then there's a whole series of prompts that that put chat gbt into different modes like you can have it act like a writing
206. 32:04partner or a gardening expert or a historian or a therapist and you can start interacting with it in that way
207. 32:12and those prompts believe it or not when you tell it to act like a certain thing it it actually gets much better at that
208. 32:19thing it's it's kind of counter-intuitive from how we're used to using Google which is give it as as concise a prompt as possible and Google
209. 32:28will sort of figure out what you need these things need a lot more context and so that prompts.chat um document will teach you about that
210. 32:39it's really good start there here's how I would approach it start out doing what I call parlor tricks just
211. 32:47play with it have it write you a country song or have it write you a poem or have it explain photosynthesis to you or give
212. 32:55you advice on your gardening or whatever you might enjoy just start playing with it and then rather than try to figure out well what
213. 33:04do I use AI for flip it around and say what's something I'm trying to do in my life like what's a problem I'm trying to
214. 33:11solve either with my family it could be like we want to eat better and I want to do some meal planning
215. 33:18um or it might be with a community group where you you want to plan a bake sale or it might be something at work where you know
216. 33:27I don't know you're struggling with marketing and they're looking to you to come up with a marketing plan take some
217. 33:32problem in your life that you need to solve and see if you can solve it with these things that's the best way to start to get your
218. 33:40head around what they make possible because anywhere where you know how you would have done it without using these
219. 33:46things when you use these things you will likely have what I call your Kevin McAllister moment which is this foreign alone um
220. 33:58and that's that's the moment you want to get to because when you get to that then you realize these things are profoundly powerful and
221. 34:07they're amplifiers they're Humanity amplifiers their ability amplifiers you want to make a marketing plan okay I
222. 34:15gotta think about okay what are the different kinds of marketing and who's our target audience and this and that or
223. 34:20you just go to this thing and say Here's my business I need to create a marketing plan is that something you can help with sure here you go
224. 34:33and you'll be like really and you'll read it and like 80 of it will be like better than you thought and then 20 will
225. 34:41suck and then tell it hey this 20 sucks oh sorry about that let's try that again and we'll get better and you say that's good but it seems a
226. 34:50little expected do you have can you give me some unexpected ideas sure I can do that and within 10 or 15 minutes
227. 35:01you'll have a fully fleshed out marketing plan and your job you should have little scabs on your chin from your jaw
228. 35:07dragging across the floor so that's how I would start with playing then solve a problem then get hooked and then your life will never
229. 35:19be the same you will step over the precipice and enter a world largely unseen by anyone before
230. 35:34you'll look back and it'll look quaint and then you'll be hit with raging fear oh my God is my job going to be the same no it's not
231. 35:50but you'll be way more prepared than people who aren't exploring with this stuff so yeah give us your YouTube again okay
232. 36:04AI learning lab Dash TT someone else some other dude got AI learning lab bastard I avoided it for three months so it's my
233. 36:18own stupid fault but you gotta add the dash TT at the end he might be a good dude I don't know but he got my AI
234. 36:25learning lab and it makes me cranky uh only lonely people use Ai No that's not true I think lonely people may start
235. 36:38having conversations with AI and that's not necessarily A Bad Thing I can tell you that since using AI
236. 36:49I am more engaged I started a community called the AI salon so if you go to the salon.ai it's free you seem like one of those that would be
237. 36:59a little bitter oh I was just trying to get us to join us his thing for money no it's free it's just a community we
238. 37:07meet every other week we have a online Discord server where you can hang out make mid-journey images ask questions um
239. 37:16I've been more connected more creatively engaged um more productive than I've been probably in my life since using AI
240. 37:32so I do not resonate with only use lonely people use AI at all we're too see here's the deal the the cliche if you haven't used AI
241. 37:43which I assume you haven't played with it much um based on the comment the cliche is that the machines are
242. 37:50doing all the work right and that that humans aren't necessary the reality is is that humans are collaborating with these machines to
243. 38:00do work and to figure this [ __ ] out there's no books on it there's no manuals there's no university courses on
244. 38:10generative AI like the post chat GPT [ __ ] like how do you use this stuff in your marketing department
245. 38:17there's no [ __ ] manual for that right now so you know what you have to do talk to people so what ends up happening is the more
246. 38:24you use this [ __ ] the more you're like oh my God how do how do you use this stuff oh here's what I do oh my God
247. 38:29that's amazing here's what I do oh my God that's amazing that's the stage of the business we're in right now that's the stage of the of
248. 38:37the technology it's it's this is all kind of spit and duct tape right now it's built on Decades of research
249. 38:45um and and it's incredibly powerful but the tools themselves like Chachi PT is very very thin from a from a a product standpoint
250. 38:55Pi is like barely functional but they just raised 1.3 billion dollars and and they're setting out to make a personal assistant
251. 39:09and that personal assistant might do things like help you socialize more help you get more dates help you save money on your next trip
252. 39:18teach you about personal finance right so yeah I do not resonate at all with that statement um I mean I mean quite honestly it's
253. 39:32it's like saying you know people that use computers are anti-social yeah they are but that's not the point that is not
254. 39:40the point little mister see what I'm saying you see how I do this you know what that means neither do I
255. 39:51all right hello Rachel what's happening amen what is a good AI State site to animate a still picture for example making rain animated let's
256. 40:01go now if I can get this stupid thing to focus on the stupid screen come on there we go all right we're gonna go
257. 40:12we're gonna go first to Mid Journey and we're gonna make what did you want you wanted rain that was a that was a bear making a beef
258. 40:20sandwich in a Chicago restaurant we did that the other night actually okay so if you've not made a mid-journey
259. 40:29image this is Discord here's here's your Discord servers those are your different communities I'm in the AI Salon Discord
260. 40:37server here's your different conversations and so mid Journey play and mid Journey lab where you can make stuff
261. 40:45and so the way you make this is you go slash imagine oops not slash m imagine
262. 41:04and then we're gonna go what did you say you said making rain animated so let's say um down let's see photograph
263. 41:19of down for in back yard with laundry hanging um and then I'm gonna go dash dash AR
264. 41:36space 16 colon 9. and if you're asking but Kyle why are you having to write code to make a picture because that's the way that's
265. 41:48where we are right now it's just the dumb thing about it uh dash dash AR stands for aspect ratio and 16x9 is like a you know like a movie
266. 41:59um aspect ratio so that's what we're gonna do and then now it's gonna go it's gonna go make us hopefully a decent downpour
267. 42:10and then we're gonna go to another tool and we're going to turn this into a video and hopefully you'll have a Kevin McAllister moment like
268. 42:18I had no idea did that but Kyle that seems like magic or it'll suck and that's been it's been kind of the theme this week is like I
269. 42:32went to I went to yesterday I went to play on open ai's playground and it literally the site just didn't work it
270. 42:38gave me back like two asterisks instead of like writing what I wanted it to write and I looked like a dum-dum it made me
271. 42:45look like a dum dum dum dum dum dum dum dum all right that's not really working photograph of a downpour let's see
272. 42:54all right we're gonna go slash those are cool but they're not what we asked for slash imagine um a photograph of heavy rain
273. 43:11um drops um falling past the camera
274. 43:26in a Farm back yard with laundry hanging all right let's try
275. 43:41that we'll see if we can get something that actually shows us the rain falling and by the way uh if you're still here
276. 43:57it uh J trulius um the tool we're going to use is Runway ml's gen 2. which is you can also do text to video so you could just go
277. 44:12describe rain falling and it'll make you video of rain falling but what I like about doing it in mid Journey first is mid
278. 44:20Journeys are dedicated it's it's the best of the image generation tools it just gives you the highest quality stuff
279. 44:26and you can start it there and then you can use this image as your prompt for Gen 2 to make a video so these do not look like rain but we'll
280. 44:38see they're not done yet um does this have this is this has got a little bit of rain all right so let's do we'll do
281. 44:52variations of two so if you've never used mid Journey you've got these these little things at the bottom you one two
282. 44:58three four and V one two three four the user upscale so you can take any of these that you like and upscale them the
283. 45:04Visa variations so this one has a little bit of rain in it so I'm going to do variations on that one and we'll see if it gives us one that's
284. 45:12got a little more rain oh
285. 45:33we haven't flown on a Saturday night everybody having a good time is it good is it like a rock concert is this as
286. 45:39good as a Taylor Swift concert oh yeah you know it is you know what it is no I don't think so but did you hear she just gave two hundred
287. 45:52thousand dollar bonuses to her 50 truck drivers that drove her crap around for that tour she's amazing she are amazing amazing
288. 46:01amazing amazing um I don't see any rain in these although this one has water water on the ground so let's grab this
289. 46:11one so we'll scale this one up that's three we're gonna upscale three and then I am gonna save as save image
290. 46:25I'll just call it oh that's good it's got my prompt in there save that to downloads okay so now we're gonna hop over to Gen 2 runwayml.com
291. 46:39[Laughter] [Music] here we come to save the day all right so I went to Runway ml I have an account there obviously
292. 46:55um I said text to Text slash image to video so I can type in a prompt and I'm gonna upload my my uh image right here
293. 47:21all right so there's my little image and then let's say um uh hard rain I just typed in hard rain let's see I've
294. 47:32noticed that if it's sometimes better if you don't type in a prompt in fact let me kill the prompt for now let's just do it with the
295. 47:41image and let's see if it if it figures anything interesting out with the image so I'm just going to use the image as
296. 47:46the as the prompt and so I say generate and so now it's going to go off and it's going to do its thing look you can see me in the forest I'm
297. 47:56like a little I'm like a little ghostly figure in the forest oh papa don't you see the face in the woods
298. 48:05no Sally what are you talking about I don't know I see him I'm terrified papa papa don't you um what oh it's an ad sorry
299. 48:19I was giving a performance there I was just trying to win the win at me it would complete me I tell you
300. 48:39no it wouldn't do you ever see that Jim Carrey thing where he's he's like I'm not just Jim Carrey I'm Golden Globe winner two-time
301. 48:54Golden Globe winner Jim Carrey and I don't go to bed at night I go to bed at night as two-time Golden Globe winner Jim Carrey and you know
302. 49:08what I dream about being three-time Golden Globe winner Jim Carrey because then I'd be enough that's
303. 49:16pretty good all right let's see so we're gonna hit play here you ready let's see if we get some rain no we don't get rain but we do get sort
304. 49:25of driving through that's actually really good that's beautiful so there's Gen 2 right okay so that's cool so now
305. 49:37let me stop that because we'll get seasick so now I'm going to put um heavy rain falling and we'll try another one generate
306. 49:53and let's see what it does the thing that I've noticed about when you actually add words to The Prompt is that it doesn't
307. 50:01um use this exact image as the first frame it kind of takes the basic idea of this image it takes some of the themes
308. 50:09of it and then it adds its other [ __ ] so it's going to be interesting to see what it does here but I gotta tell you that
309. 50:15that animation that it just did of you know sort of blasting down the the driveway or whatever that is that's pretty cool
310. 50:27like I could absolutely see that you know so I mean this is the kind of thing that if you get if you get a style
311. 50:32dialed in in mid-journey and and you sort of figure out your storyboard of the kinds of shots that you want and you
312. 50:39just established the first um the first frame of each of those shots um you know this this is something you
313. 50:47could do now this is this is not super cheap like I have 193 seconds left and I think I'm paying I forget what it is whatever the bottom
314. 50:57one is um there's another trick that you can do that someone told me about that I thought was pretty brilliant so if this
315. 51:07is your first frame you can go to your last frame you can do you can screenshot this or grab the last frame of this video
316. 51:15and then use that as the first frame of of the next video and then you could string a bunch of those together and
317. 51:22make like a 30 second clip Gen 2 only does four second clips yeah so this is a completely different scene but it does look like rain's
318. 51:30falling so here we go yeah so you put an umbrella there it is kind of in the woods but it it completely recomposed
319. 51:45recompose the image but that's rain falling so how can we improve Gen 2 fill out a short survey all right your email I'll I'll go I'll
320. 51:58do it how How likely are you to recommend Runway to a friend or colleague well I just did it to 84 people so pretty pretty likely How would
321. 52:09how can we improve runway for you when adding a prompt and an image at the same time comma
322. 52:25retain as much of the image as possible period I either get image coherence with no prompt or
323. 52:40something completely different with a prompt period it would be nice to be able to add something like Heavy Rain comma and still have the
324. 52:52scene look the same with the addition of rain period what's the main benefit you get from Runway um experimenting
325. 53:07what type of people do you think will benefit um I don't know creatives I don't really care maybe contact you
326. 53:19yes okay all right so we gave them some feedback we've we're we're good citizens but anyway so yeah so that's so Runway
327. 53:28would be the tool I would use to do that you're gonna have to experiment with with how you go from like what what
328. 53:34would probably make more sense for you to do is if you can get mid-journey to give you a better you know something
329. 53:41that's more clearly raining that gives it the clue that it's raining um or just use text to video at Runway
330. 53:49at Gen 2 and just describe the scene that you want um it to be raining in and it'll just make that video for you
331. 54:03oh just jump me to the bottom tick tock just I can't I can't take it anymore I want to be where it was
332. 54:16why did you do this to me all I want to do is help the people but I can't help the people if I can't see their
333. 54:23questions Tick Tock please stop it you're making me hurt inside uh for those of you that are new here I
334. 54:38don't like it when Tick Tock jumps me to the bottom foreign [Music]
335. 55:15all right I use Chachi BT too much but I have upgraded is there a cut off I don't I don't think so I think they're upping
336. 55:27I think they're upping their rate limits so I've never hit I tend to use chai gbt in little bursts and I think they reset it every three
337. 55:39hours anyway so I've never really hit a rate limit with chat gbt but if you're doing some intense stuff and you get
338. 55:45rate limited the the other the other way that you can do it Rachel there's there's kind of two ways to use chat gbt
339. 55:51you can use it at chat.openai.com but you can also use the openai playground which is platform dot openai.com
340. 56:01playground it's a different interface but you can still use chat gpt4 as the model and and use it that way so that if
341. 56:11you're getting rate rate Limited that might be good also if you haven't turned on plugins and code interpreter
342. 56:20and custom instructions do that on chat GPT if you're paying the 20 bucks go get the extra features because they're really good
343. 56:29they're really really good except the plugins are a little bit poo poo the plugins plugins are not great there's 800 of them but they're not
344. 56:38great but there's some good ones so we're gonna I'm gonna play with some tonight I'm gonna go do something with
345. 56:42my with my uh YouTube our new YouTube channel um and we're gonna do some chat gptn with the transcript
346. 56:52haha I fell asleep to you once let's see it's it's that's my favorite compliment slash troll I like to watch the AI learning lab to
347. 57:05get some well-earned shut eye Kyle's voice just puts me into Slumber like nothing else [Laughter] and I used to drink whiskey but now
348. 57:20it's the AI learning lab for me oh my wife she's going to be in Denver next month oh nice has treats for chance that's awesome
349. 57:31Tobias that's great champ will be happy he loves treats he loves small stuffed animals too he'll carry him around for three weeks and
350. 57:40that is one day that they're just murdered there's just small animal we call him we call him his babies he's carrying his carrying around
351. 57:49his babies and then we're like champ why did you kill your baby and there's just like stuffed animal stuffing all over
352. 57:56the floor it's hilarious it's very cute all right can't find your YouTube can you add it to the AI Salon screen yeah that's a good idea
353. 58:10um the the YouTube is AI learning lab Dash TT I just did it tonight there's another AI learning lab that bastard got
354. 58:20there before me because I waited for three months before I did it um AI learning lab Dash TT as in tick tock because
355. 58:34I didn't use chat GPT to have any imagination for the title I had AI Dash learning lab and then I was like no that
356. 58:43like No One's Gonna put the dash in there so AI learning lab Dash TT but yeah I'll add it to my other I've
357. 58:52gotta I've got to go redo my link trees anyway ah there it is oh cool great I do know Tim Cook that's good could you tell Tim Cook to
358. 59:01to fix the fix the the auto focus on the [ __ ] back cameras on these things this could also be a uh a tick tock issue but Jesus it's just a
359. 59:14it's just horrible um it just it just don't it just don't Focus it's just like it's not looking do you see there's text there maybe you
360. 59:25could use a little of your fancy AI [ __ ] and have it Focus your Apple figure it out they don't have Steve Jobs there going
361. 59:34uh just make it work anyway um I'm I'm less pissed off at Apple about the focus than I am at Tick Tock for
362. 59:46jumping me down to the bottom of the comments when I'm when I'm scrolling when someone puts something new in all right
363. 59:53um are you gonna get your retina scan when you go to New York I am yep and and listen I they claim that they're
364. 1:00:00not keeping the the biometric data I don't I don't really care if they do or not like I feel like my data's already
365. 1:00:05out there like I I I made I consciously made a deal with the devil I don't know 30 years ago when I was
366. 1:00:13doing agency.com back in the olden timey days where it's just like I'm just gonna put my data out there I don't I don't give a [ __ ] um
367. 1:00:24I so so what what uh what's your name cuppy cuppy206 what cuppy's talking about is Sam Altman last week launched this
368. 1:00:34thing called World coin and it's got three components to it it's got a transaction service thank you very much
369. 1:00:41for the rose appreciate that uh there's a transaction kind of component like PayPal there's a cryptocurrency called
370. 1:00:48World coin and then there's an identification protocol called World ID and in New York Miami and San Francisco
371. 1:00:57you can go to these physical locations and they have this round orb they call it the orb and if they wanted to make it if they wanted
372. 1:01:06to make a scarier sci-fi name for the thing they couldn't they couldn't have so it's a it's it's a stupid it's a
373. 1:01:13stupid name that is that is gonna perpetuate the paranoia that the robots are gonna kill us but whatever but you
374. 1:01:20go to one of these major cities and you look into this orbit scans your iris it all it's basically doing is going yep
375. 1:01:26that's that's a human Iris or it's an iris whatever we're taking the biometric fingerprint we're verifying you as human
376. 1:01:31and then you get a blockchain verified ID that says you're human what's that why is that important well I I think we are about to learn
377. 1:01:44the importance of human verification in the 2024 U.S presidential election uh because there is going to be so much
378. 1:01:53misinformation and and probably personalized misinformation um and you know things just things that look
379. 1:02:04real that aren't and things that you know came from a trusted source that didn't actually come from The Trusted source
380. 1:02:11so there's a verification component to it that's really important I I personally just want to be part of it
381. 1:02:17early enough just because I think the applications for it are kind of insanely powerful imagine something like a Twitter or X as as it
382. 1:02:27likes to be called now where you knew that everyone that you were interacting with on that service was a human was a verified human and had
383. 1:02:37a very a blockchain verifiable history that you could go follow all right they signed up for Twitter on this date and this on this date
384. 1:02:45and they're verified human with this world ID cool and then you can layer on top of that you know here's a video that Kyle made
385. 1:02:53and I can verify you know on the blockchain that that hasn't been modified okay so I verified that he's
386. 1:02:59human I've verified that this is video that he created and that it's not been modified okay I can trust that video
387. 1:03:05become really really important so there's a layer this verification layer of of content and people is sorely
388. 1:03:15missing right now so I think it's interesting on that front the other thing about world coin that's that's I I don't know like
389. 1:03:26the the sort of depth that they're thinking they've been thinking about this stuff for a lot of years and so and it shows right because
390. 1:03:35they've thought about the implications so one of the implications of these tools is that they are going to cost
391. 1:03:41people jobs there is going to be disruption people are going to be displaced jobs are going to be reinvented or
392. 1:03:48completely automated out with with this technology and and not just this the image stuff the video stuff all of it
393. 1:03:54right the tools that you look at now um are the worst they're ever going to be these are going to get incredibly
394. 1:04:02sophisticated incredibly quickly so jobs are going to be eliminated so there's a couple of things going on
395. 1:04:09Sam Altman when he when he flipped open AI from a non-profit company to a for-profit company he's doing it as a capped for-profit
396. 1:04:19so what that means is if if an investor put in a billion dollars they can um the maximum they can get returned
397. 1:04:28from that investment is a hundred X so if they put in a billion they can get a hundred billion out Sam Altman feels that this is going to
398. 1:04:37be trillions McKinsey agrees this is going to be trillions of dollars of prosperity generation so anything above that cap above above
399. 1:04:47that 100x cap is going to basically go into the fund so the other thing about world coin is that he's setting it up as
400. 1:04:53a universal basic income for members and um and he's basically going to turn it into a
401. 1:05:07wealth redistribution platform so that as these large language models start to displace jobs and start to increase
402. 1:05:16Prosperity that there's basically a tax on that that's going back to the people that were displaced so I think it's a
403. 1:05:23fascinating experiment I want to get in on it um thanks to the stupid [ __ ] sec everyone else in the world when they
404. 1:05:33scan their eye they get free cryptocurrency but in the United States we can't get free cryptocurrency because
405. 1:05:39it's a Ponzi scheme and it's evil and we're not allowed but anyway yeah so yes I am already subscribed cool he is touched
406. 1:05:49thank you I appreciate that do you think AI comedians will be funnier than real comedians not with their current timing
407. 1:05:58go go listen to pie with when you talk to pie you say something funny she goes that was a funny one Kyle
408. 1:06:08no I don't think they're gonna be funny other comedians for some time um we did an experiment a couple of
409. 1:06:17weeks ago on the live where we went to all of these different tools I wrote a prompt um to write a joke about avocados
410. 1:06:25um you know not being ripe for a long time and then they're ripe and then they're rotten um and chat gpt4 was the only one that
411. 1:06:34actually wrote a joke that could have could be turned into something a comedian might actually say all of the others were embarrassingly bad
412. 1:06:42um but they're gonna get better they're gonna you know someone's gonna write a joke writing um actually you could if you wanted to
413. 1:06:50play with it go to Poe and create a joke writing bot I bet I bet you could do one pretty decent all right
414. 1:07:06is there anything you would not use AI what are you doing there Cowboy liaison oh thank you the bear welcome the Bears
415. 1:07:16good to see you thank you for the Hattie hat thank you for the rose Adam that's nice uh we did a we did a little run the other
416. 1:07:25night where a bunch of people were putting stupid [ __ ] on screen and I got all NPC like a rose a hat I mean
417. 1:07:34oh those sights are funny did you have you seen the there's a dude that um so there's the the people doing the NPC
418. 1:07:41thing where they like you know clickety-clack and whatever as people pay them [ __ ] they they talk stupid and then um this guy
419. 1:07:49put an iPhone on his robotic lawnmower and he's just putting it on Tick Tock lives and and for like four hours at a
420. 1:07:59shot he puts it on Tick Tock lives as this lawnmower just drives around his yard mowing very very slowly and he's
421. 1:08:06made like 2 500 bucks in two weeks so you can make money with Chachi you can make money with tick tock oh it's [ __ ] hilarious
422. 1:08:18ah what is there anything I would not use AI for well hmm
423. 1:08:35I don't think that there's anything that I would use AI exclusively for right now although I I you know that's probably
424. 1:08:42not true I probably have done that like my so so I've got it you know my my art project on Instagram is called Kyle
425. 1:08:51Shannon dreams and it's I've also got a tick tock Channel called Kyle Shannon dreams where I took these you know AI
426. 1:08:57generated self portraits and and put them up there um you know that that was 100 AI but most of most of the work that I do when
427. 1:09:06I'm doing when i'm doing things to do things it tends to be a collaboration between me and the tools I think that right now
428. 1:09:19I think it's I think it's in all of our interest to and in fact Ethan Malik said this he was at the uh the marketing AI Institute
429. 1:09:32just had their annual conference and Ethan Malik spoke and he said he said the best thing you can do right now
430. 1:09:39is for everything that you do at work or at school or whatever it is try to use one of these tools when you do that thing
431. 1:09:50try to figure out if if it can do it if it can do it better if it can accelerate something because the way these things do what they do
432. 1:10:00we've never had anything remotely like this so none of us know how it's going to change how we write emails how we
433. 1:10:10read emails how we read annual reports you know how we create PowerPoint presentations how we modify them
434. 1:10:18um how we create computer stuff like everything's going to change so I kind of feel right now like everything needs to be an experiment
435. 1:10:29look everything is an experiment because these tools are so new and they're so imperfect right so you can't
436. 1:10:37you just can't trust these things are gonna do the [ __ ] you think they're gonna do exactly how you think they're
437. 1:10:43gonna do it like like the thing that I just experimented with where I created the image of the backyard and it was
438. 1:10:49sort of raining but not and we put it in the video thing and it made a cool video but it wasn't raining and then when I
439. 1:10:55said Make It Rain it added an umbrella which we didn't want a [ __ ] umbrella it's a backyard we don't need an umbrella right
440. 1:11:03um and it didn't get that so but it was a good experiment right like like I feel like I learned something in just kind of
441. 1:11:11getting the Nuance of how it did what it did but it's not going to act like that a year from now right a year from now it's
442. 1:11:20probably going to have a level of coherence that's just mind-blowing to what we're doing today so I might not use that for anything
443. 1:11:27today but six months from now I might use it for 50 of the things like that that I need and a year from now
444. 1:11:34that might be the primary way I do it and and two years from now that might be the primary way that all videographers do what they do
445. 1:11:44we don't know so so I don't think so I think right now I I think right now like like where my head is is every time I have a problem
446. 1:12:01and not a problem every time I have something that I need to solve at work I'm now I'm now at the place and this
447. 1:12:09took me five or six months I'm now at the place where I first think is there anything I could do with AI with that
448. 1:12:18and sometimes the answer is no but a lot of times the answer is like yeah let me go let me hop into chat gbt here let me
449. 1:12:24run over here let me pre-visualize something over there thanks Adam appreciate the rose um that's a really weird switch it's a
450. 1:12:34really weird ah thank you very much it's a really weird switch to go from if if you're in well we're all problem
451. 1:12:41solvers but but like I pride myself in general sense oh shut up Siri seriously come on back to Tim Cook
452. 1:12:57[Laughter] um what the hell was I talking about uh uh I don't know what I was talking about Syrian or after me she broke my brain oh
453. 1:13:13anyway all right I've been wanting to look into it for a while but didn't know where to start oh absolutely mama
454. 1:13:22um yeah start with start with chat.openingi.com have it make you a meal plan like the meal plan thing is
455. 1:13:28really cool or an exercise plan or like you know some sort of thing like that just give it sort of a general you know kind of
456. 1:13:37you know diet that you want to play with and then it'll make you a big meal plan oh you know what we'll do something cool
457. 1:13:43I'll show you how to I'll show you how to use plugins let's let's go do a little bit of uh let's go do a little bit of chat GPT in together
458. 1:13:57all right I'm going to show you a couple of things so here's chat gbt they've added these fancy they're they're kind
459. 1:14:04of useless they're sort of thought starters but you know whatever it's cute it's nice they're they're they're
460. 1:14:10they're they're doing they're doing what they can to improve it they need to do things like return access to the
461. 1:14:15internet to to improve it but you know some thought starters are nice okay so if you're if you're at chat GPT
462. 1:14:25um if you pay your 20 bucks a month you get this little yellow flag and then you get access to code interpreter and
463. 1:14:31plugins as well as custom instructions and I'll talk about those in a sec um if you don't see these if you're
464. 1:14:38paying for the if you're paying your 20 bucks a month but you don't see code interpreter and plugins you actually
465. 1:14:43need to explicitly enable them so you go to your settings and then in settings go to Beta features and turn that [ __ ] on
466. 1:14:53see the little green pills you want green pills not gray pills gray pills make you sleepy green pills
467. 1:14:59that make you all excited you'll be like I could play with code interpreter now oh yeah all right so we're gonna play with plugins first
468. 1:15:09so I was talking to Mama LA about um about making a meal plan so I'm gonna turn on a plug-in where are it
469. 1:15:24um did they get rid of it or do they just miss it there it is instacart okay so I'm gonna say let's see
470. 1:15:40make me a meal plan for the next seven days for a family of four with a chubby dog with a heavy emphasis on Meatloaf all right
471. 1:15:55[Laughter] why why not sure I can help with that here's a simple meal plan all right so let's see what it gives us
472. 1:16:05breakfast scrambled eggs with toast chicken salad sandwiches for lunch meatloaf with mashed potatoes for dinner let's see
473. 1:16:14day two lunch leftover meatloaf sandwiches that's actually a brilliant day too um spaghetti bolognese is kind of
474. 1:16:24shredded meatloaf on pasta which is good meatloaf with roasted vegetables on day four baked salmon baked salmon no all right fine barbecue
475. 1:16:37ribs steak and potatoes as for your chubby dog it's important to consult with a vet all right so there's all of our
476. 1:16:51ingredients there's our instructions okay now what it what it says after giving me that shopping list is would
477. 1:16:57you like to create a shopping list for these meals and I can say yes and what this is going to do now it's
478. 1:17:03actually going to fire off the instacart plug-in it's going to understand what are the ingredients I need
479. 1:17:12to execute that meal plan and turn that into an instacart shopping list and then I can just have that [ __ ] delivered
480. 1:17:20so plug-ins you kind of you activate them for for a given chat session you can activate up to three I just have one in this case um for demo
481. 1:17:31you can activate up to three you can install as many as you want but in any given chat session um you just do you just do one so
482. 1:17:45today nine stores are delivering that's not actually my ZIP code but that's fine but like there's our there's the stuff
483. 1:17:52right ketchup we're having meatloaf we're getting we're getting the three pack of ketchup damn right we are
484. 1:17:59we're getting a roasted chicken we got some meatballs in there some berries some tomato sauce this looks expensive we're going back to
485. 1:18:08macaroni and cheese um so so that's kind of how plugins work um so now let's let's try something new I'm gonna go to plugins I'm gonna turn
486. 1:18:22off instacart and I'm gonna turn on video insights I think there might be another one Vox script so both of these will read YouTube
487. 1:18:37transcripts so what we're going to do is we're going to hop over to the newly minted look at this wait let me go let
488. 1:18:44me go back to the home page AI learning lab YouTube channel huh I did it you did it yay um so I'm gonna go to videos and I'm
489. 1:18:57gonna go grab let's see what's the first one April 30th April 29th was when it all started all right I'm gonna go to April 29th everybody
490. 1:19:12look there's two of me I'm the same size too look look how look look how much fatter I was good Lord it's disturbing okay shut up
491. 1:19:28okay so now we're gonna come back over to chat gbt and what we're gonna do is we're gonna say um first thing I'm going to do I'm going to
492. 1:19:38say do you understand how to create video titles descriptions and
493. 1:19:52chapters for YouTube yes I can certainly help you understand video titles should be engaging
494. 1:20:08video descriptions there's a reason I'm doing this so so one of the things that really helps a lot with Chach EBT is if you give it
495. 1:20:17context and one of the ways you can give it context is you can have it educate you right and so they're there it's telling me
496. 1:20:25um here's how to make chapters you know with within YouTube um here's all this stuff so now when I start asking it to do this for me it
497. 1:20:36will follow its own rules so so it will make it better um just in doing that okay so what we're going to do is we're going to say um
498. 1:20:47this video and then I'm going to paste in the URL and I'm going to hit shift return let's see this video um
499. 1:21:01is a recording of a fairly long live session where the person talks about AI period
500. 1:21:16can you come up with a title that encapsulates what was talked about comma a description that highlights some of the more
501. 1:21:32interesting things and creates a chapter list of the relevant sections that might be interested to an audience trying to
502. 1:21:42learn about generative AI period so now in theory in theory this is going to run off okay so it's grabbing video insights
503. 1:21:55it grabbed that plug-in So based on the transcript what is generative AI
504. 1:22:18ethical implications of AI let me I don't think this is the right um I'm gonna put I don't think that's the right format
505. 1:22:28for the chapters for YouTube can you please create the description and the chapters so that they are appropriately formatted
506. 1:22:41to put in a YouTube description and have them function as hyperlinks let's see if it fixes that ice I apologize
507. 1:23:09I don't think I don't think it did it based on the transcription because it's got introduction what is generative AI
508. 1:23:20applications of generative AI Deep dive into generative models ethical implications Future q a session
509. 1:23:27conclusion that's not how my lives work this is [ __ ] this is [ __ ] so this is this is what we call in the
510. 1:23:34business an experiment we experimenting and the experiment was a miserable failure it did not look at the transcription at all um
511. 1:23:45it appears you did not look at the transcription of that video at all based on the chapters you created period
512. 1:23:53can you see the transcription of the video I gave you earlier ah if now it's using a different plugin that's interesting
513. 1:24:09so this is box script and so what we might find here is that Vox script did better and it's actually looking promising the
514. 1:24:16fact that it's doing multiple calls we can open these up and see what it is because it was a long transcript I bet
515. 1:24:21it's having to do the summarization um separately or yeah let's see
516. 1:24:37next transcription next transcript token yeah so Vox script is automatically breaking up a longer the longer transcript into into
517. 1:24:47like 3 500 token chunks which is this this is actually impressive to me um so so if you don't know Chachi PT
518. 1:24:57kind of has this limit of 4 000 tokens it's moving to eight it's not I I don't quite quite know what it is on gbt4s but
519. 1:25:06it's basically you can't put in a really long transcript or it just chokes it out so what Vox script just did was it
520. 1:25:12automatically broke that up for me all right this is looking better introduction AI in the music industry
521. 1:25:21intellectual property challenges Ai and e-commerce large language model optimization Ai and project management Ai and content
522. 1:25:29Ai and perfume Ai and energy please note that the time stamps are placeholders um let's in this comprehensive discussion we
523. 1:25:44explore it's a fascinating world of generative AI um let's see can you make the description
524. 1:25:59less flowery and more factual absolutely here's a straightforward
525. 1:26:13version this video provides an in-depth discussion on generative AI it's applications and the potential implant
526. 1:26:20impact on various Industries the speaker discusses how AI is being uh I don't know if that's true but anyway let's
527. 1:26:31I'll deal with the with the description some other time but let's go to YouTube and then how do I edit this I go to
528. 1:26:45my YouTube studio and this is 429 so we're gonna go content and then we're gonna find 429 where's 429 there it is and we're gonna
529. 1:27:01edit it how do I edit it what where's the um there we go I'm gonna pop in the description I'm
530. 1:27:16gonna put in all those chapters and I'm going to save it and then we're gonna go back out to the channel
531. 1:27:32and now I should where's the description why is it not there
532. 1:27:47foreign no views nobody's views my videos but they've been up for an hour a show transcript where's where's the description
533. 1:28:02hmm maybe and what why does it not work instantly and I'll address anything
534. 1:28:17um one of the things that I've talked about on the channel I've actually got this thing what I'm trying
535. 1:28:24to do I'm trying to see if the if the chapters are um functional like if if they're actually what they say they are
536. 1:28:35I have a sneaking suspicion they're not but edit oh it didn't save it save
537. 1:28:58foreign there we go okay so now show more okay so now we've got
538. 1:29:14our little chapters although it didn't hyperlink the bottom ones that's probably some weird copy paste problem
539. 1:29:22but large language model optimization at 35 minutes in I only need 30 seconds of anyone's voice probably less than that to really
540. 1:29:32synthesize it so so anyone who's out there um you know whether it's actors or voice over actors this isn't project management
541. 1:29:44um relatively replicable things like blog posts or not replicable things that are more formulaic where it's the same kind of thing over
542. 1:29:53and over again I would say here's air in the music industry what its implications are how to incorporate it into your
543. 1:30:00business or into your life or if those two are very blurry um how to deal with that Synthesia uh s-y-n-t-h-e-s-i-a is one of them and
544. 1:30:14dash id.com in fact I can go there hang on good this is AI and economy great question I wouldn't have had an answer
545. 1:30:22to this but I I talked to a friend of mine who said he'd been thinking about AI in a 30 to 50 year time frame it's wrong
546. 1:30:30all right that was a failed experiment I I I'll experiment with that but but hopefully you can learn you know at
547. 1:30:37least what I did there right you know I had a I had a plug-in that goes and looks at a at a transcript the the fact
548. 1:30:43that it broke that transcript up into all these components is actually promising there's just something about how I wrote the the prompt
549. 1:30:52um for it to create these chapters it didn't quite it didn't quite do it right so um oh you know what it didn't do this is interesting
550. 1:31:03if I go look at what data it gave back it gave back parts of the transcript but did it it did not include it did not include time code
551. 1:31:18so let's let's do an experiment here um it looks like when you broke the transcription up into components you did not include the time code
552. 1:31:33in the summarized chunks and therefore the chapters that you created are not accurate to what the person is actually saying period
553. 1:31:46can you re-analyze the transcript and generate up to three chapters per chunk with accurate time code and descriptions
554. 1:32:05yeah there's more than one way to skin a cat and actually what I'm what I'm doing right now like like this kind of okay
555. 1:32:15using large language models right now is kind of like being in an escape room and you know it'll sort of do what you
556. 1:32:27ask it to do but then it doesn't quite and like the lock doesn't unlock you like damn it we were so close
557. 1:32:33and then and then you're like I know we could ask it to do it this way and it'll do it that way and so yeah now it it
558. 1:32:43what did it say I apologize for the confusion earlier let's revisit the transcript one more time with accurate
559. 1:32:49um okay that's that's [ __ ] so I think I've still got the URL there so let's try again um you still didn't get it please start over
560. 1:33:04with the transcript of this video uh wait uh this video
561. 1:33:25um with a transcript of this video um and when you and and let's see let's just focus on creating accurate
562. 1:33:41chapter names and time code so make sure you include time code when you analyze the transcript all right we'll give it one more shot
563. 1:33:56we'll see if it fire it should fire off Vox script nope it used video insights I'm going to stop generating um please use the Vox script
564. 1:34:15plugin plugin did and we'll see if it actually puts any um time code in the summaries
565. 1:34:32it did not wait did it
566. 1:34:48maybe it did no no no this is a fail it's a fail fail whale
567. 1:35:04we come we saw we conquered then we realized we did not conquered and it was an abject failure we are leaving in humiliation nay laughs
568. 1:35:22[Music] so so I uh that's that's the way a lot of this goes and
569. 1:35:37you know if you're sitting there going that seems miserable well that's where it is right now um two years from now
570. 1:35:45this stuff's gonna be magical right because all all of this mess is gonna get solved and these tools are just going to get increasingly
571. 1:35:58sophisticated like you can see that what it's what it's trying to do is in the right neighborhood right the fact that
572. 1:36:04it's do it's breaking up a transcript into sub into um chunks small enough that chat CPT can do is good the fact that Vox script as a
573. 1:36:14plug-in isn't smart enough to include the time code is [ __ ] stupid right okay they'll fix that at some point or
574. 1:36:23someone will come along with something better or YouTube will come along with something better and you'll just be able
575. 1:36:28to pop the pop the write my description and chapters for all my videos button on YouTube right it's Google so they could
576. 1:36:40do that probably tomorrow um but the process that I just went through of thinking like oh I don't want to have to
577. 1:36:50manually watch all of my [ __ ] lives again and figure out what I was talking about but they're essentially useless
578. 1:36:58um if you don't kind of chapterize them and understand what's in them so I've got to find a way to solve this problem
579. 1:37:05um it's definitely going to be some version of chat CBT or I might go to Claude so so here's an actually you know what let's go ah
580. 1:37:14oh oh it's so exciting look stimming stimming stimming I just had an idea we're gonna go to Claude Claude is a beast okay so here's
581. 1:37:27what we're gonna do we're gonna go to um so I'm on YouTube I clicked on the little three dots right here I'm gonna say show transcript
582. 1:37:38so here's my transcript with time code can I copy this with time code what's it what's it let me do here toggle timestamps select all no
583. 1:37:52okay I'm gonna just go select all of this okay now so so this is a good example of when you might use different tools okay so we just tried
584. 1:38:08oh we're gonna do we're gonna do two tools this is this is like a Hands-On little workshop here tonight kids it's
585. 1:38:14gonna be fantastic what'd you do Saturday night I watched a guy fail miserably over and over and over again sounds awesome um
586. 1:38:26so we tried plugins that was a fail Claude has a hundred thousand token context window gee Kyle what does that mean
587. 1:38:34that means I can paste in up to 75 000 words into this chat and it will under it will read it and be able to understand what's there so
588. 1:38:50let's start a chat here with Claude and and the the two hour or whatever it was hour and a half transcript um
589. 1:38:59probably is like twenty thousand words so I think we're well within so I'm gonna say do you understand how to write YouTube
590. 1:39:08titles descriptions and chapters so that the chapters are linkable in the YouTube description yes
591. 1:39:23yes can help generate optimized YouTube titles blah blah blah blah blah then in the description we're going to
592. 1:39:29blah blah blah blah blah blah just put the time code dash okay great perfect make sure the chapter titles in the
593. 1:39:38descriptions the same keywords are relevant for the content let me know if you need any clarification I'm going to say um no this is perfect
594. 1:39:47period now I would like you to consider this transcript colon shift return and I'm going to paste in
595. 1:39:58my entire transcript and it it knocks it into a little 50 kilobyte thing you can put in up to 10 megabyte things so that transcript is
596. 1:40:07well within I think it's going to be able to understand it now consider this transcript okay so since it's got it in there
597. 1:40:16consider this transcript and write me a description for the inaugural
598. 1:40:30live session of the AI learning lab period the speaker's name is Kyle Shannon period make the description interesting but not overly
599. 1:40:47flowery period provide a title that encapsulates what the session covered
600. 1:41:04and create chapters of the most interesting topics that were discussed all right let's see if this does anything interesting
601. 1:41:22and by the way if this does this then this is going to be a yay Ai and the music industry intellectual
602. 1:41:37property challenges so huh you know what's interesting is that's what the other one said so that that the other one might have been right
603. 1:41:46alternative and Indus wait alternative energy can AI help solve our problems opportunities for displaced workers this seems close
604. 1:41:56um it did something wrong here [Music] you didn't write the chapters in The Format you described above period I need
605. 1:42:06the chapters to be click clickable when they are pasted into YouTube description area period
606. 1:42:21also can you add five additional chapters based on the next most on the next most interesting topics
607. 1:42:38because it was a little light but let's let's see what this does looking promising Kyle Shannon Tech entrepreneur and AI
608. 1:42:55enthusiast hosted the first ever session on the AI learning lab to discuss the current state blah blah
609. 1:43:03all right so now it did them correctly so let's copy all this copy let's go over to Ye Old YouTube go back to our little
610. 1:43:18YouTube Studio welcome everybody looks like we got some new folks here welcome I'm I'm just geeking out uh I'll explain what I'm
611. 1:43:27doing in a second here let me see if I can get this thing working I'll pop back out and breathe here in a second
612. 1:43:34so we're gonna get rid of all that crap we're gonna put in that crap let's go grab our title too because it's
613. 1:43:42not wait the AI Revolution is here implications and opportunities that sounds like something I would say that
614. 1:43:50sounds like some pompous ass talking about the beginning of uh the future what should I do with dates here I don't care it doesn't really
615. 1:43:59matter right now we're just spearminting okay we're gonna say save we good we saved so let's go back over to our little Channel
616. 1:44:10we'll go to our little videos we'll go down to the 29th if you have any questions about a
617. 1:44:25show more all right we have a description we have a title and we have chapters and they all work look at Claude coming in for the win
618. 1:44:34starting a business leveraging AI 41-12 um I can also show you some things here on oh you know what we can do let's
619. 1:44:47make Kyle talk a little faster to do anything talk faster boy um project management consulting any advice um
620. 1:44:57um the project management SAS company is really interesting one because I think a lot of the project management is going
621. 1:45:01to be handled yeah um all right that's that's definitely talking about starting a business I think this just worked I think this just
622. 1:45:10worked double tap the screen for Claude coming through and actually solving um a pretty significant challenge let's see um
623. 1:45:19Ai and project management rethinking if you understand something um so any of these courses you're going
624. 1:45:32to turn it behind the vehicles things are things are moving so fast so um if you feel really lost and like you don't know where
625. 1:45:38was the language process independently better so I think my um sort of fake assumption assumption based on the knowledge is that you can
626. 1:45:45understand programming logic it understands logic better um so so um they had a dramatic breakthrough reasons
627. 1:45:50their gbp3 was so good I think it was a combination of programming understanding code and understanding this massive data
628. 1:45:54set all the internet is different um and a GPT 4 is just a much more sophisticated version of the China thing
629. 1:45:58um how open AI did what they did is probably relatively straightforward from a technical perspective to replicate all
630. 1:46:03right then I'm talking about China and intellectual property I think it did it um I'll have to go back in and like
631. 1:46:10really really dig into that but but that that did it that's pretty insane that's pretty insane okay so for those of you that are new here
632. 1:46:22people on the old Ticky talk have been going hey Kyle you got you got any of these here tick tocks on the YouTube If
633. 1:46:29I want to go back and watch them I was like no I don't even know how to do that because I'm old uh and so I figured it out today and so
634. 1:46:36I uploaded a bunch of things and then I thought well I don't want to write descriptions for these and these things
635. 1:46:41are kind of useless without chapters and descriptions um so what I just did was I went and I tried chat gbt and it failed using
636. 1:46:53plugins thanks silver fox and then it hit me oh right I've got Claude the reason I have seven of these things up here is because they're
637. 1:47:03they're good at different things so Claude with its 75 000 word context window I can paste a two hour transcript into that
638. 1:47:14and have it write me a good compelling title a good compelling description and then chapters that link to different parts of the
639. 1:47:25video and it'll make these videos actually kind of usable um I can probably also use it to do things like make YouTube shorts and
640. 1:47:33whatever happy other horse [ __ ] I want to do so so now that I know it can do that now I'll now I'll go work on that prompt
641. 1:47:41right like that's the kind of thing that that I'll probably spend an hour or two working on that prompt because when you
642. 1:47:48get that prompt dialed in the workflow then becomes you know grab the transcript paste in the prompt
643. 1:47:56you know paste in the transcript hit the button copy and paste it right you could you could you know do a pile of videos really quickly and
644. 1:48:04ideally you want to automate that so so I will probably automate that at some point but that was kind of cool
645. 1:48:14um we could probably also give that a shot with code interpreter where I could save that as a file put it into code
646. 1:48:19interpreter and do the same thing that I just did with uh with Claude with code interpreter should I do that you want to see that
647. 1:48:27I'm happy to do that but I'm also happy to move on thanks so much I've been wanting to look into this that's what
648. 1:48:33started that whole thing I was being doxxed online once I now own the a-hole's name online forever that's good
649. 1:48:40I hate [ __ ] trolls and thieves and [ __ ] like that um all right yeah it's like SEO that's why I'm learning it all right can you talk
650. 1:48:54about more about machine learning um I don't know what you want to know about it it's not my expertise I can
651. 1:49:00tell you how chat GPT works I can tell you how the image models work I can tell you a little bit about embedding and what that is
652. 1:49:08um if you if you add if you ask me this is from GT 357 um if you tell me a little bit more about what you're looking for I can do that
653. 1:49:19Apocrypha hello Kyle hello Apocrypha how are you I need a Lamborghini up close front quarter View doors closed St Louis
654. 1:49:27background all right let's go let's go make an image for Tim Diner Tim Diner is it Denair yeah Denair all right Tim wants it for our
655. 1:49:36Lamborghini what color oh Tim fine slash imagine
656. 1:49:57photograph of a Lamborghini up close front quarter View comma doors closed
657. 1:50:11St Louis background at sunset dash dash a r sixteen nine
658. 1:50:25sixteen nine um I hate yellow Lamborghinis let's do a green one because it's unexpected
659. 1:50:41all right let's see what we get and then if it if it's interesting we'll pop it over to uh to gen one and see if
660. 1:50:49we can turn it into a video if the videos at all interesting and by the way if it if it does St Louis without the
661. 1:50:57arch I'm gonna be disappointed It's gotta have the arch in it right it's St Louis I don't see St Louis there maybe it'll
662. 1:51:04be way in the distance it's definitely green Lamborghini looking oh sorry uh the focus thing dear Tim Cook my name's Kyle Shannon
663. 1:51:17I have as many as 27 friends who watch me on the internet okay that's not St Louis but whatever uh we'll take which one of those
664. 1:51:32I guess number one I'm just gonna upscale it did it it's not really St Louis like that's one that I would muck with right
665. 1:51:42but let's save this and let's go over to Gen 2 and turn this into a movie we're gonna make this a movie using that artificial intelligence
666. 1:51:55yeah yeah I was watching this guy he he well I'll tell you what he was doing he was he was using that artificial
667. 1:52:01intelligence and he uh he he went in and he made a picture of a of a Lamborghini you know I'm not a fan of Lamborghinis
668. 1:52:10I'm more uh I'm more a Detroit man if you know what I mean but but uh he made he being one of them there Lamborghinis
669. 1:52:18and uh and he made it a color green it was it was a bit of a it was a little bit of a a little sissified green if you
670. 1:52:25know what I mean uh uh he made it green in any uh and he put it in St Louis but it didn't look like St Louis to me but
671. 1:52:32anyway they went over this other thing I don't know what it was called it was called a Runway ml that's what it was
672. 1:52:38called and uh he stuck it in there and then he made a movie out of it it was pretty scary I think the robots are going to kill us
673. 1:52:49all right let's see there it is let's bring it in all right so I'm in Runway ml right now and I've now got
674. 1:52:59the Lamborghini is now my prompt so I'm gonna that's the movie I just made before so I'm gonna generate
675. 1:53:06and so it's now going to go off and turn that image into a video it's interesting because the car clearly
676. 1:53:15looks like it's parked so I doubt it's gonna make the car move but maybe what it'll do is a camera move around it but we'll see
677. 1:53:24it'll be interesting interesting interesting interesting now there's a lot of people here welcome everybody my
678. 1:53:31name is Kyle Shannon this the AI learning lab uh if you have questions pop them in the comments below here's
679. 1:53:36the deal with the questions I try to answer all of them I go through them sequentially so I am more than 99
680. 1:53:42messages behind so if it seems like I'm not answering anyone's messages that's because I'm not but I am
681. 1:53:49but it's I'm a time traveler you're in the future if you think of these questions like the scrolling credits of the fourth Star
682. 1:53:59Wars movie which was the first for me [ __ ] number four was number one for me okay so you're down here on the comments I'm way back there
683. 1:54:12remember those com the the title in Star Wars I'm answering these questions yours is up here hold your [ __ ] horses okay let's see
684. 1:54:24what Gen 2 did with this little image uh how did it roll without the wheels rolling well
685. 1:54:42but it's actually not bad like if the wheels actually rolled like that would be something to you know to dick around with
686. 1:54:48it's not bad is it other than the fact that the wheels are are still it's like it's being towed towed by a cable hahaha
687. 1:54:58stealing my Lamborghini [Laughter] oh that's funny foreign all right and and it looks like we got some new folks here so the the URLs that
688. 1:55:14I have here the top one is that's the official chat GPT website all of the others are either some different flavor
689. 1:55:21of Chachi PT or some competitor to Chad gbt they all do something really interesting we just did an experiment
690. 1:55:28where we tried to get chat gbt with plugins to um write descriptions and chapters for a really long YouTube video with with
691. 1:55:38plugins failed miserably um then we went to Claude AI which can take really long um transcripts and it can understand
692. 1:55:47them you can just paste it in there and it reads it and it succeeded perfectly it was great um so so that's that's what these are these
693. 1:55:57are these are for things for you if you're new to this stuff if you're new to generative Ai and you haven't played
694. 1:56:02with chat EBT like go to the top one or really go to any if you have played with the top one go experiment with the other ones the
695. 1:56:10other thing that is worth me putting up here if you're new here is um prompts.chat will teach you how to prompt and it will teach you um
696. 1:56:23different ways you can tell chat GPT to behave where it will kind of shapeshift and do very very different things for
697. 1:56:31you like you can have it act like a social media manager or your therapist or your nutritionist or your
698. 1:56:38motivational speaker or a screenwriter or whatever and it does it really surprisingly well and that document will
699. 1:56:47teach you how to do that and it you'll you'll be able to start to get your head around how these tools well sorry these tools
700. 1:56:55are way more than just a fancy Google search they are profoundly powerful and that's what this channel is all about
701. 1:57:03there's it's not just these but it's it's the technology that sits underneath these and how do we apply them to our
702. 1:57:09work and our lives and all that sort of happy horse [ __ ] all right there are these tools trying to charge
703. 1:57:16to split videos up into the best parts is there a lower price one I don't know there's yeah there's that one video
704. 1:57:24v-i-d-y-o I I'm subscribed to that one but it's like so expensive that like I got a lot of videos that are long that I
705. 1:57:32want to chop up into Parts I don't I don't know if there's a better one if you find one chance 150 please let me know
706. 1:57:41all right love your show bro thank you very much purpose Master good name I like it I'm gonna go nine out of 10 on
707. 1:57:47the name I like it outside Chicago girl comparing yourself to Swift to Taylor Swift I know I don't quite have the
708. 1:57:54dance moves of Taylor Swift but I mean listen this audience it this is almost like a hundred thousand people in a stadium
709. 1:58:02there's a lot of cheering there's a lot a lot of foot stamping what was the thing was it in in the Seattle show the the crowd was dancing
710. 1:58:12so hard and so in unison that it it measured on the Richter scale there's they cause seismic activity
711. 1:58:21it's like holy [ __ ] [ __ ] that tour is insane she's amazing like as a just as a business Force she's like you're not gonna give me the
712. 1:58:35rights to my music back [ __ ] you I'll go re-record it and tell my fans to re-buy all of my music and they did
713. 1:58:44and then she didn't tour for six albums and someone said to her well what's this gonna be like a three and a half hour
714. 1:58:52concert and she's like yeah foreign she's grossing a billion dollars on this tour and and like and like clearing like 550
715. 1:59:06million and she just gave the 50 truck drivers that drove the tour around the country two hundred thousand dollar bonuses each [ __ ] insane
716. 1:59:19it's great [ __ ] awesome awesome awesome awesome awesome awesome awesome website or program um I don't know what you're talking
717. 1:59:30about VK Rook um chat gbt is a website they also have an Android and iOS app but be careful go to open AI to find the
718. 1:59:45links to their official apps because there's a lot of fakers on iOS and probably more on Android so just be
719. 1:59:52really careful get the official ones what's that website again I don't know what song it was that I was looking at
720. 2:00:02very cool cool beans High peeps the pair the bear micro very good am I late hey Emilio's wife I don't think so you've
721. 2:00:13been here for a while you've seen some of the some of the cool stuff I just did with the with the prompting
722. 2:00:20the prompting oh and Amelia's wife I I now have a YouTube channel I think you saw me working on the thing it's um AI learning lab Dash TT
723. 2:00:32somebody else had AI learning lab because I was too slow to register it because I'm a dum-dum I'm a dum dum dum
724. 2:00:40dum dum dum dum dum dum dum dum dum dum dum dum dum so I had to get Dash TT is my uh YouTube channel all right
725. 2:00:55best way to create my own AI without code well a couple of ways and it depends when you say my own AI it
726. 2:01:10depends on how how big my is if Maya is I've had a company for a decade and I have six terabytes of data that I want to be my
727. 2:01:26chat bot that's a very different conversation than I kind of want it to talk like me if you kind of wanted to talk like you
728. 2:01:34you can go to poe.com and they have a bot maker and it's really cool like you just name it give it a description
729. 2:01:43give it a like a greeting like what does it say when it when you launch it and then give it a prompt and the prompt is
730. 2:01:49basically here's what I want you to act like and so if you have got a particular personality you can just put it in there and then
731. 2:01:57you can name it yourself and then you can give people poe.com well your name's Jay by the way 1.5 a single letter
732. 2:02:04although that's interesting I'm going to revise my score because because Jay seems very lazy but getting a single letter seems good that
733. 2:02:14seems that seems like a power move it's one in 26 so I'm gonna go 8.5 on the name now Jay um anyway I go to poe.com and do that
734. 2:02:26the other thing that you could do is if you've played around with zapier zapier now has a thing called interfaces and
735. 2:02:35one of the interfaces that that they've got a whatever a template for is a bot so you can do kind of decently
736. 2:02:43sophisticated prompting and create a bot with zapier and if you pay whatever they're 30 Buck a month subscription extortion price then you
737. 2:02:54can put your own your url there I don't think you can do that with Poe so that could be interesting and then if you've got you know a
738. 2:03:03decades worth of data and you want to be able to kind of make your company.gpt then you've got to go dick around with a
739. 2:03:10project called Lang chain Lang chain as in language chain um and that's going to require you to be a bit more of a geeky mcgeekington
740. 2:03:23you gotta be geeky you got to learn that they're Python and you got to learn about permissions and and uh code and
741. 2:03:32you know interfaces and apis and you know you gotta you gotta be like Unix happy you got to go all command line
742. 2:03:41it's like commando but you can wear underwear it's command line so it depends all right moving on bookshelf oh tick
743. 2:03:55tock huh they just jumped me to the bottom again hang on let me go find where I was in the stupid place in line
744. 2:04:15what'd you do Saturday night I don't know I watched Tick Tock scrolled on him and seemed to pop a vein in his forehead I
745. 2:04:26don't know what happened it didn't look good how long did you watched him for two hours what are you doing with your life I don't know
746. 2:04:35it's not good tap that screen if you like stupid [ __ ] you give me money [Laughter]
747. 2:04:52oh good lord all right let's not even let me scroll up to where I was well I missed some questions sorry about that
748. 2:04:58people lose questions that I missed oh no here's love your show bro and someone talking about Taylor Swift all
749. 2:05:03right am I late no the Bears here Amelia's wife is here okay that's okay my comment is at the bottom yeah if your comments at the bottom
750. 2:05:16you're kind of screwed with me tips for prompts for chat EBT um go to prompts.chat to start go to that thing to start it's got some
751. 2:05:26good primers there um probably the most important tip for prompting any of these large language models any of these is
752. 2:05:38to think of it not as a um not as a single prompt but more like a conversation so here's a good way to
753. 2:05:53think about these things they're like knowing they're idiot savants is what they are so imagine that you just were given an
754. 2:06:04intern and or or an assistant and the assistance a little slow on the uptake like they gotta you gotta tell
755. 2:06:12them everything they're like I don't know I don't know what to do you're like well could you write a poem okay here's
756. 2:06:19a poem and you're like well that's a crappy poem like okay I'll make another one here oh that's a crappy poem too so
757. 2:06:24you gotta you gotta give them a lot of instruction but they're an idiot savant they know everything so what these things have been trained
758. 2:06:34on is like all of the public internet so they know all of the public internet up to September 2021 so think about whatever
759. 2:06:44has been put on the net for the past 35 or 40 years are essentially sitting inside these things it's it's astounding but they're really dumb
760. 2:06:55so you got to say like okay I want you to act like a screenwriter can you tell me about how I might structure a
761. 2:07:01screenplay if I wanted to write one oh absolutely here you go there's four different methods there's Sid fields of
762. 2:07:07his little book there's Joseph Campbell in the hero's journey there's that there's Aristotle there's save the cat
763. 2:07:15here which one would you like me to do oh okay uh that one okay so so the thing about prompting is that's the conversation you're gonna
764. 2:07:29have so so think about they call it prompt stacking now they came up with a name for it it's prompt stacking so if
765. 2:07:36you want to impress your friends what were you doing last night oh I don't know I was just do a little prompt
766. 2:07:41stacking if you know what I mean foreign no I don't know what you mean what are you talking about that's not important
767. 2:07:47right now it's an AI thing you probably wouldn't be interested um but prom stacking is basically just
768. 2:07:53having a conversation so start out with like um do you understand uh social media marketing sure I understand it here's some theories great
769. 2:08:04um I've got a company that does this that and the other thing and I'm trying to figure out how I can get more people
770. 2:08:10to come to my pizza shop great here's some ideas oh those are shitty can you give me better ones sure here's you know
771. 2:08:16and just go back and forth just like you would be working with an intern Just Pretend This is a remote employee that you have
772. 2:08:22that will do anything you tell it to do but it will also lie it'll also get stuff wrong and then there's some times when it'll just
773. 2:08:33um frustrate the crap out of you because you think it should be able to do the thing and then it can't
774. 2:08:40and you're like all right it's an intern [ __ ] interns but then occasionally get an intern that's [ __ ] brilliant you're like oh
775. 2:08:47that was great I never would have thought of that that intern brought A New Perspective it does that too and it knows everything
776. 2:08:57and the ones that are connected to the internet so Bing and Bard and Pi are connected to the internet so those you can ask about the Taylor
777. 2:09:07Swift tour Chachi BT is not connected to the internet anymore because they made a boo-boo claude's not connected to the internet um
778. 2:09:19lab stop perplexity that's the Llama 2 models they're not connected to the internet so that would be the tip think of it
779. 2:09:27like an intern like the smartest intern you've ever worked with that needs a lot of Direction and will lie to you like they have no remorse
780. 2:09:41but they don't have an evil bone in their body like in fact most of these things with the exception of the Llama 2
781. 2:09:47models which don't have safety guard rails they are very much designed to give you what you want that's that's how
782. 2:09:53they work they're like I want to give this person what they want I'll do everything I can to give them what they
783. 2:09:58want and that's why they'll confidently give you answers and they say no that's wrong when they're like ah so sorry I got that wrong
784. 2:10:05let me try again nope that's still wrong ah sorry about that let me try again that's like three times in a row you got
785. 2:10:13it wrong all right let me try it again it's like a golden retriever and sometimes it has the intelligence of
786. 2:10:20a golden retriever from a language perspective and then other times it's like [ __ ] Einstein it's amazing
787. 2:10:30but it's all an experiment right now just just sort of here's here's another good way to think about prompting right now choose
788. 2:10:40to go on an adventure and think of it like that playing with these things right now this early in the game is an adventure
789. 2:10:52and rather than saying like I have to learn AI that sucks yeah having to learn [ __ ] does suck but going on an adventure is fun
790. 2:11:01and it's exciting and it's scary and I might get killed by the dragon but I might find some gold that's this right now
791. 2:11:12and I and I promise you that getting literate with these tools right now for the next 12 to 18 months is going to be the best investment any
792. 2:11:23of us can make because the requirement to know how to use these tools well is going to go it's gonna it's right now we're kind of like
793. 2:11:34this and it's starting to do this and at some point it's gonna go like this where you got to know your [ __ ] with this stuff
794. 2:11:45and if you don't they're going to be 17 people behind you that do an entire job categories are going to be decimated
795. 2:11:58but they're going to be filled with some amount less people that know how to use this stuff so one of the scariest things about
796. 2:12:08these is is when you have your Kevin McAllister moment when you have this moment remember home alone when you have that
797. 2:12:16moment and you realize holy [ __ ] these things are powerful and then and then you have the moment of existential dread
798. 2:12:22where you're like oh [ __ ] oh God is this gonna take my job it is very likely going to transform every one of our jobs transform them
799. 2:12:36entirely and it's going to take some jobs but if you're competent with these things and if you're curious and if you
800. 2:12:43think about it like being on an adventure your chance of participating in whatever the future of work becomes goes up dramatically
801. 2:12:55and the next 12 to 18 months like the amount of opportunity for all of us in the next 12 to 18 months is insane
802. 2:13:04and if you ignore it and say I'll get to it next year it's it's I don't think it's going to be pretty I think it's going to be painful
803. 2:13:15even if you're in the adventure even if you're playing I think it's going to be painful but for me personally I would much
804. 2:13:22rather be educated about this stuff and and you know on the surfboard rather than in the Cozy beach hut as this tsunami is coming in
805. 2:13:34you know surf the tsunami don't just wait for it to come in because it's going to come in quick and devastating I think
806. 2:13:44or used during the coronation of the British Monarchy Sam the king interesting here growth geek um yeah they are
807. 2:13:54I you know I feel like they could have they could have not used the term orb for for this thing that scans your
808. 2:14:07retina I feel like they it might have it might have behooved them to do a little bit of focus grouping on what the
809. 2:14:16[ __ ] we call this thing did not make it sound so terrifying but whatever all right how's it doing time wise all right we're
810. 2:14:27gonna go to bed at some point here Ubi through blockchain wow that's genius we'll get most people to use open ai's
811. 2:14:36resource versus others until it's not yep I agree everyone here right now so what everyone's commenting on right now
812. 2:14:42in my comments is something I talked about like 45 minutes ago which is Sam Altman's new company World coin they
813. 2:14:49have this thing called the orb which is these Chrome balls that are about this size that have a hole in in the front
814. 2:14:57and you you stick your eye on it it scans your retina and verifies you're a human and then if you're not in America you
815. 2:15:06get cryptocurrency if you are in America that Gary Gensler can tell you to go [ __ ] yourself um you don't get cryptocurrency because
816. 2:15:15it's not safe because we can't manage our own crypto um not your keys not your crypto dude
817. 2:15:31um but anyway you scan your retina and they give you a thing called A World ID which is a blockchain verified human identity
818. 2:15:42and he's gonna use it as a platform for Universal basic income and for wealth redistribution of the capped profit
819. 2:15:53model of open AI which I think it's actually really interesting I think someone needs to solve the human verification problem and the job
820. 2:16:02disruption problem and that's the purpose of it is it the right model will it be the one that turns into something I don't know
821. 2:16:11probably not maybe possibly you know they're well funded they're smarter than [ __ ] they're thinking about this stuff in
822. 2:16:22in a measured way like if you compare if if you compare interviews with the CEO of mid-journey and the interviews with Sam Altman
823. 2:16:32they're very different interviews like mid-journey the chances of them getting sued into Oblivion is is like 99 percent
824. 2:16:42the chance of open AI getting sued into Oblivion is like one percent right Sam's a smart dude very very thoughtful you know anyway
825. 2:16:53what's the best way to create my own AI without coding we talked about that PJ just gave shade emotion after I did not
826. 2:17:01answer correctly and said haha be nice to each other don't be [ __ ] in here please thanks maybe pie is
827. 2:17:09appropriately responding to your humor Kyle probably [Laughter] we were discussing Plato Plato oh plateau
828. 2:17:21oh I love discussing the Ancients in the philosophical underpinnings of language in The Human Condition I saw the lawnmower guy
829. 2:17:34very very cool get fired quicker it's good I don't know what that was in reference to it's funny I think if AI
830. 2:17:42first always now yeah it's awesome Silver Fox it that that's a really weird transition that transition from someone presents you with with a
831. 2:17:51you know here's a problem to solve like where my brain for 58 years went was okay here's what I need to do I need to
832. 2:17:59think about this I need to go Google something I need to right like you just go into problem solving mode and now my
833. 2:18:05problem solving mode is someone presents me the other problem I'm like I need to go see if I can do this on chat GPT
834. 2:18:13and then you know 99 of the times it at least gets me over the hump of how to approach the problem right it's amazing
835. 2:18:25but it's that that discipline is that's taken me months to get my head out of I need to solve this no maybe I don't
836. 2:18:35and as these tools get better and more sophisticated and more automated autonomous it's gonna get better and
837. 2:18:42better and better this is the worst this technology is ever going to be it's the least capable it will ever be right now
838. 2:18:52like we're going to look back on the way these things work right now and laugh at it and probably not that far in the
839. 2:18:59future like a year and a half from now we're gonna be like remember when you could only put 3 000 words in chat GPT oh my God
840. 2:19:09how did we ever get away without a billion for a billion token context window who knows which by the way that paper came out a month ago
841. 2:19:23some Microsoft researchers figured out how to create a billion token context window so I say 750 million words you could put in your prompt
842. 2:19:35so you could you could take all of the words you've ever read in your lifetime and put them in the prompt and start interacting with it
843. 2:19:46like that's coming that ain't that far away they have they have the mechanism for how to do it they the paper's out and
844. 2:19:55it's fascinating with how they're doing it like I I read I groked it just enough to get like how they're kind of doing it like
845. 2:20:03they're thinking of language as fractals basically it's insane have you ever heard do you know what a
846. 2:20:10gigapixel image is you know those those images where like you see a landscape photo but then you can just sort of
847. 2:20:17infinitely zoom in on it and you can zoom in all the way to see like the guy drinking a cup of coffee on the patio in
848. 2:20:24the building you didn't even know there was a town there and you can zoom in and you realize there's a town and then you
849. 2:20:29realize there's a building and then you realize there's a dude on a patio that's what they're doing with language to get
850. 2:20:35that billion parameter model they look at massive amounts of language and they sort of abstract it out and then if you
851. 2:20:41start zooming in on an area of the prompt it basically I don't know what it does these these [ __ ] people are geniuses
852. 2:20:50um anyway building a chatbot of me train on all my content oh that's cool um what are you and what are you using are
853. 2:21:00you using Lang chain that sounds really really cool um I tried modifying G-Code for my 3D printer and it couldn't do it so Bradley
854. 2:21:09um it can likely do it but I I would try different tools so if you if you weren't using gpt4 use gpt4 it's
855. 2:21:19better at coding than 3.5 also try Bard Bard is apparently better at coding um and and Claude is probably not too
856. 2:21:28shabby just because those are X openai guys um I think the Llama 13B model is better at coding if I remember right I I'm guessing llama
857. 2:21:43might not be the trick right now I would say bard gpt4 and then potentially uh co-pilot github's co-pilot it might it might be better to
858. 2:21:57do it there the other thing about the G-Code is something that you could do at Claude is that you could take the documentation
859. 2:22:11for G-Code paste that in and take the code for your 3D printer and paste that in and then have it modify your code
860. 2:22:22having all of the documentation of G-Code in its memory you can do that at Claude or you could do it with code interpreter
861. 2:22:29at at a chatgpt that's interesting though Bradley all right Siri played creepy song last week creepy
862. 2:22:41oh interesting yeah I know like a series of this they gotta just I hope they don't re I hope they don't keep that
863. 2:22:49branding I hope they flush that branding down the toilet because it's just never worked right some people love it but I don't like chat Bots
864. 2:23:03um you were talking about how you've been using AI to solve problems uh sorry I don't know what I was talking
865. 2:23:09about Siri never responds when I want to talk to her and then random interruptions exactly Emilia's wife like
866. 2:23:18when I asked for her she's never there and then when I'm just rambling on she jumps in how can I help you no stop
867. 2:23:27that appropriate I tried to get chat CPT to write write as tragedy where the main character on the lives by the police
868. 2:23:38yeah it might not do that like that it couldn't do it yeah Bradley so so for stuff like that if you if you want to
869. 2:23:46write content that's got sensitive subjects that are going to get caught by the safety guard rails of all these
870. 2:23:51things then you're going to want to go to labs.perplexity.ai that's the those are the Llama two models from meta that
871. 2:23:58are open source they don't have safety guard rails on them um and it should be able to do it do you still opt out of chat history and
872. 2:24:07training I don't because I'm lazy um I haven't created a prompt Library System yet so I just rely on chat GPT to
873. 2:24:18do it and I try not to put anything in there that's too sensitive but it's a good thing to do is um opt out of
874. 2:24:29saving chat history until openai not to save you [ __ ] if you're doing anything um that you don't want it trained on
875. 2:24:39ski which crypto will Spike next you can do that it it likely won't know which crypto will Spike less next but you can
876. 2:24:48pop in a whole pile of historical data and you know potentially you know real-time data like you know up to the
877. 2:24:55up to the hour data and have it do the analysis and give you picks yes it can probably do that quite well
878. 2:25:04will the picks actually be good or not I don't know it depends on the analysis you tell it to do if you haven't played
879. 2:25:10with code interpreter Richard this is Richard Watkins um go play with code interpreter because you could take crypto historical data
880. 2:25:19get that into a PDF or a CSV pop it into code interpreter and start doing analysis on it um I don't know [ __ ] about that world but
881. 2:25:30if you do you will be blown away it'd be amazing that's so cool maybe I can cancel hellofresh that was when I was using instacart's plug-in
882. 2:25:42I got you a gift well thank you Adam I appreciate that I don't know what that means but thank you what do I do
883. 2:25:52I I'll probably find out after this I'm you know don't forget I am an old guy on Tick Tock and I haven't learned to twerk yet
884. 2:26:01did you give me a song for for twerking at an end of the channel [Laughter] all right um yeah if you're interested in this stuff
885. 2:26:16give me a follow I've got you know I do Tick Tock videos but I've been doing I've been doing more of these live lately than static videos
886. 2:26:25um because I because I find them a bit more interesting but um but I got a lot of content on the channel so go follow there
887. 2:26:31um if you're digging this feel free to share this live I'll go for maybe another 25 minutes or so and uh yeah and
888. 2:26:41if you're new here the the reason I have a list of URLs here I'm not associated with any of these companies these are the seven chat GPT
889. 2:26:52variants or competitors to chat GPT that I think are worth paying attention to right now if you haven't played with
890. 2:26:59chat gbt at all then just start with the top one that's the official chat gbt website just start playing with it and also go to prompts.chat
891. 2:27:09the top thing right there and that'll teach you about prompting and how to use these tools all right what website is that I don't
892. 2:27:18know what it was Kyle I find it fascinating that AI gives me something to to get started yeah that's that's the thing about this is
893. 2:27:29I don't know if you've ever experienced this I assume we all have but but you know how sometimes you'll be thinking about something
894. 2:27:37you're like I kind of have an idea but it's kind of a weak idea and it's just kind of a whisper of an idea and you're
895. 2:27:42like yeah but that would be too much work to deal with and screw it and then the idea just you know vaporizes and goes away
896. 2:27:50with these things you can take that sort of hint of an idea and go oh you know what [ __ ] it let me just throw it in
897. 2:27:57chat EBT and before you know you you're like inspired and you're like oh yeah that actually is a good idea and uh let
898. 2:28:03me expand on that let me intend to it oh you know what I should do I should go to Mid Journey if you got 80d it's this is like crack
899. 2:28:14I do and it is um but but even if you don't even if it's just simple problem solving stuff where where you know historically you sort of
900. 2:28:23have to think through stuff and and and you kind of do mental stack ranking really like you know this idea is better than that one
901. 2:28:31and that's better than that one with these tools you can kind of go explore them all um and it really does sort of get you
902. 2:28:38started in in a in a very powerful way I got the same answer for another of your videos um as you did so hallucinations
903. 2:28:47interesting yeah that that first thing that I did was pretty bad Silver Fox but the quad thing was pretty good [Music] um
904. 2:28:57what are you teaching here well Richie Rich I don't know that I'm teaching anything here I'm rambling incoherently about
905. 2:29:05some things that you might learn something from or not uh it's mostly generative AI it's how we can use it what its implications are
906. 2:29:21why I think it's interesting um I was a a founder of one of the early digital agencies in the mid 90s so I was
907. 2:29:32there at the beginning of the world wide web and that whole run up an explosion and you know bit of a bust but you know it actually
908. 2:29:40just kept going up um so I've seen this before and this is the parallels to the early days of the web and this are very very similar
909. 2:29:50um but this is way more profound and happening way faster so to put it in context the World Wide Web to get to 100 million
910. 2:30:01users it took about six years chat GPT it took six weeks six weeks to 100 million users fastest adoption of technology in history um
911. 2:30:14and what this stuff does compared to the early days of the web is it is so profoundly powerful that it it's kind of eye-wateringly
912. 2:30:27challenging to get your head around what what it even makes possible it makes so much possible and it and it's just like it it's it's
913. 2:30:40it looks just like a Google search box but this ain't fancy Google search this is way way deeper than that so that's what we're
914. 2:30:50talking about that's what I'm talking about and and I basically try to answer all the questions like like who the hell are you what are
915. 2:30:58you teaching here what are your qualifications sir do you have papers do you have papers no I don't well I do but it's an acting degree
916. 2:31:10[Laughter] I do have those papers plus one wants to know what you think about Chrome GPT I don't know what
917. 2:31:19Chrome GPT is but let's go look at it did oh no yeah welcome to save the day
918. 2:31:36GPT oh it's an extension
919. 2:32:07um it looks fine so so it's a it's an extension for Chrome that basically puts um chat EPT in the context of it I you
920. 2:32:17know I switched my default browser to Bing which which if if you had bet me serious money a year ago that I would
921. 2:32:26have Bing as my default or Edge as my default browser and Bing is my default search engine um I I would I would have taken the bet
922. 2:32:35I would have left you you know out of the room um Bing has gpt4 incorporated into it and it it or Edge does um so so does Bing but Edge has
923. 2:32:53um gpt4 integrated into it in a pretty slick way I assume this extension does something similar the ability to search
924. 2:33:03on the internet you know um do searches that you know that are context sensitive to the page you're on things like that
925. 2:33:11um so I haven't been using any of the GPT extensions here here's my main reason why the the um when when Google
926. 2:33:25starts to incorporate Bard into Chrome and and when Microsoft starts to incorporate GPT four and five into all
927. 2:33:34of their product Suite all of those extensions become kind of irrelevant um and so I don't I don't want to get
928. 2:33:41married to a certain way of working or relying on something that is going to be in competition with what the core
929. 2:33:48operating system of the browser or or you know of the of the apps that I already use are so I'm just kind of
930. 2:33:56ignoring them so I don't I don't really know it but I understand why you'd do it but I would say download Edge and play
931. 2:34:03around with Edge and see if you like it better or worse I'd be curious to hear what you think um I sent you an email about a stack I
932. 2:34:10started made with Claude to do something like this oh cool that's great thanks over Fox that's awesome
933. 2:34:16um can't you just export the transcript and then give it to chat GPT and have it analyze it again you you can't no you
934. 2:34:24can't because chat GPT has a 4 000 token context window so that the transcript's way too big this is this is one of the problems
935. 2:34:34um the the one plug-in that we used broke the transcript into sections but it didn't maintain the time code so that
936. 2:34:42was stupid on its part and then um clawed because it's got a bigger context window I could just paste the whole
937. 2:34:48transcript in there so that works um and then chat GPT I could put the transcript into code interpreter and that would probably work
938. 2:34:58maybe the problem is with limitations with the plugins yeah it was that plug-in sucked actually it was llama oh interesting
939. 2:35:07um I was using Claude in to to Aid in writing it was learning my writing style but now is saying that it does not have that capability
940. 2:35:16did that just change um I don't know Andy Lee rocks um so
941. 2:35:32so the memory of a given chat session does not carry over to another chat session so only within that chat will it
942. 2:35:42have the memory of whatever writing it was learning so it might just be that that maybe you're in another chat and it
943. 2:35:49doesn't have the memory of the other thing that you talked about so that like that 75 000 context window is only within a single chat um
944. 2:36:01so there's there's there's a thing called uh what's it called Model creep um so model creep is model creep is
945. 2:36:16um all of these guys wait not that all of these guys are tweaking and
946. 2:36:31um refining these models as we're working with them so very often what will happen is you'll have a prompt that works and
947. 2:36:40then three weeks later it just doesn't seem to work anymore that's model creep and it sucks and and it's just again it's part of the nature
948. 2:36:49so it's possible that you hit that it's also possible that that um you're assuming that it's carrying over
949. 2:36:58its memory from previous conversations and it doesn't do that it will in the future or you know I don't know if
950. 2:37:04Claude will but you know some of somewhere in these someone's going to start adding memory to these where it
951. 2:37:10will actually remember the context of all of our different conversations and all of our documents and all that sort
952. 2:37:17of [ __ ] it's going to get really good right now they're just kind of it just doesn't do that um for anyone new Kyle goes back through
953. 2:37:26the comments and answers all the questions thank you Joker holy Cali figured it out oh that's awesome
954. 2:37:33I just figured it out what'd you figure out Josh wait Kyle it left out the part about how you're completely unqualified exactly [Laughter]
955. 2:37:42Claude exactly Claude for the win that was a pretty cool one um can you have it summarize the steps and share it so I can screenshot please
956. 2:37:51the chapter project can you have it summarize the steps summarize what steps over Fox um I can go find where's my clog
957. 2:38:05where can I paste these I can't I can't paste these anywhere can I uh
958. 2:38:22I don't quite know what you're asking but I've got that prompt saved so um so I can get I can get it to you somehow
959. 2:38:31like the The Prompt right now is is not all that refined what did I put um it said you didn't write the chapters in
960. 2:38:41The Format you described above I need the chapters to be clickable when they're pasted into a YouTube description area also
961. 2:38:50can you add five additional chapters um yeah I don't know what I did but uh I'll I'll get you what I can Silver Fox just
962. 2:39:03DM me and I can with what your specific request is and I can get you that um okay let's see I never realized you said
963. 2:39:14um so much until you sped it up I know me too it's like I'm I'm I'm I'm I'm I'm I'm I'm I'm I'm I'm I'm I'm I'm I'm I'm um
964. 2:39:22foreign what are you gonna do I'm just making this [ __ ] up on the fly so that's why you get the ums well I know what I'm
965. 2:39:29talking about I don't um but I I'm a lot what is the best CD DVD for a Dell 7940 laptop that's
966. 2:39:45pretty funny I don't have a [ __ ] clue I assume it's a joke um and what other whatever other happy horse
967. 2:39:55thank you quoting my my stupid sayings uh but about I did SEO back in the day I think seo seo is going to be a really
968. 2:40:05interest it's got a really interesting future ahead of it I don't know that it's a good one I don't know I I think there might be
969. 2:40:13llmo like how do you get your [ __ ] in large language models and I think how you get it in large language models is you have good content
970. 2:40:23all this SEO and injection [ __ ] might go the way of the dodo bird it's going to be interesting to see how that
971. 2:40:31all plays out what do you think of Google bard I I think it's um it's at least not embarrassingly bad now what's interesting about Bard what
972. 2:40:43they're doing really well they're incorporating hyperlinks in the answers at appropriate times with appropriate context and and in a pretty
973. 2:40:53sophisticated way so I like how it incorporates the internet into its answers um I hear it's better at coding than it is
974. 2:41:02at language and it's it's okay at language right now it's not great but it's okay um and and apparently Gemini which is their next model
975. 2:41:12is going to be a gpt4 competitor um considering where bardi is right now I think Gemini might be actually really good so
976. 2:41:23I it's it's on the list like it wasn't on this list for three months two months three months like it was laughably bad it's not anymore
977. 2:41:34I wrote AI code in the military in the Mi military a 2023 Shelby GT500 KR arrest me red oh
978. 2:41:50Todd Hendricks he wants more pictures I don't know that I don't know that it would be good at being that specific
979. 2:41:57although let's let's let's try that let's do this I'm gonna be slash imagine paste that bad boy in there of a red
980. 2:42:162023 Shelby GT 500 KR
981. 2:42:32close up front Corner view doors closed St Louis arch in the background all right I'll make that one let's see
982. 2:42:44if that turns into anything interesting all right I wrote AI code in the mid oh in the in the in the mid 1990s yeah
983. 2:42:52listen I I mean ai's ai's been around for decades so so the the reason that that I think this is such a similar time
984. 2:43:00to to the early days of the World Wide Web the internet had been around for decades and it was primarily used by scientists
985. 2:43:08and researchers and Geeks and then the World Wide Web came along and it made it simple enough that just regular human beings could use it
986. 2:43:18same thing with machine learning and you know neural networks and and you know expert systems and all of the stuff that
987. 2:43:25has been the foundation that has gotten us here and Transformers in 2017 and what changed was Chet GPT comes along
988. 2:43:37Geeks look at chat gbt and they're like well you could just do that in the playground yeah but the difference
989. 2:43:42between chat GPT and the playground is you don't have any of the settings in chat GPT it's simple like simple enough that anyone who could
990. 2:43:52use a Google search bar could use it and that's why it got a hundred million users using it in six weeks because it
991. 2:44:02was like a Google search bar but then when you type [ __ ] into it what came back was [ __ ] miraculous
992. 2:44:08it wasn't a list of SEO injected links eighty percent of which are irrelevant to you right and so that's the that's the
993. 2:44:18difference so so you have the foundation of all of that stuff this gave really good images but I explicitly asked for the St Louis Arch
994. 2:44:27and we got nothing we got no St Louis Arch how does it miss that but these Mustangs look pretty you got
995. 2:44:37the lights right it got the hood shape right it got the cobra they're not bad but where's the St Louis Arch come on people what are we doing
996. 2:44:51what are we doing uh are we down is that Lexica no what was I using I don't know what I was using this is this is mid-journey this is uh
997. 2:45:05mid-journey I don't know what I was using before that might have looked like lexico I see it now mid-journing okay um
998. 2:45:16which website is this that was mid-journey like a Shelby nice I saw a tiny finger pushing that Matchbox car exactly that was pretty funny
999. 2:45:29um long form to short form repurpose.io or get much not perfect but Time Savers yeah by the way uh hair growth geek
1000. 2:45:39um I signed up for repurposed dot IO today for an account it doesn't do lives it'll it'll it'll automatically
1001. 2:45:49download and re repurpose your channel but it won't do lives so I gotta find a tool that'll do the lives I'm just doing
1002. 2:45:58them manually now because I think I was starting to lose some of the early ones I think I lost three or four of them
1003. 2:46:05uh we've paid far more than we did for Taylor Swift tickets yeah in your sanity um I love her for it glitching uh she
1004. 2:46:15also gave a lot of money to the food banks um in the cities where she's gone on tour oh that's cool yeah she's I mean listen she she
1005. 2:46:31I mean she was clearly already a powerful woman but like in the last two or three years she stepped into her power
1006. 2:46:40like she's like a [ __ ] superhero right now she's just [ __ ] in this [ __ ] Zone it's it's very impressive she's very impressive
1007. 2:46:52ugh I thought I might I missed the prompts info oh I missed the prompts info I don't know what prompts you were
1008. 2:46:58talking about the prompting I did with I don't know what it was but I now have a a YouTube channel so
1009. 2:47:06you can see I now have a YouTube channel um it's got some some of the early lives on it the AI learning lab uh AI learning lab Dash TT
1010. 2:47:18on YouTube so that's the new YouTube channel so this stuff will all be up there so and and we figured out we can use Claude
1011. 2:47:28to actually generate chapters for for all the videos I gotta figure out a way to automate that [ __ ] I wonder if Claude
1012. 2:47:35does anyone know if claude's in zapier am I late was it was an hour ago Mr Shannon oh man yeah well I know
1013. 2:47:47quality of experience should be the basis for AI not Comfort or happiness that's interesting the quality of
1014. 2:47:54experience should be the basis for AI not Comfort or happiness I think that's right I mean I I listen I think ostensibly what these tools do
1015. 2:48:07is they do stuff at least faster and and in a lot of cases better than we do and and if if you think about how
1016. 2:48:23they're going to improve in the future they're going to start to get way better at us at almost everything and so I
1017. 2:48:30think you're right I think I think how you use these things the quality of what they produce and how we engage with them
1018. 2:48:39is going to be the thing um and it's going to cause a lot of pain and disruption it's gonna suck but on the other side of it
1019. 2:48:51is going to be some some pretty amazing stuff I think I hope oh it just jumped me to the bottom again okay I'm gonna do I'm gonna answer a
1020. 2:49:00couple more questions I'm just gonna slide up from the bottom a little bit here and answer a few questions and then
1021. 2:49:05I'm gonna go because I'm tired I'm cranky my ass is numb and I'm almost out of water mm-hmm can a complex company Implement code
1022. 2:49:19interpreter in an enclosed space does openai need to be involved I don't know so I don't think code interpreters available via the API yet
1023. 2:49:31um I don't know enough about Azure so Microsoft Azure has the you can buy I think the cheapest ones like 25 Grand so
1024. 2:49:40I think you can create these sort of stand-alone azure instances where I think you're running chat GPT within that instance
1025. 2:49:51where you're not actually making a call out to open ai's instance their servers um I think that's how that works but
1026. 2:50:02you would obviously need to do that you know you would need to be working in a company that's got budget to do that so I would look up
1027. 2:50:11um what what Microsoft is doing with Azure with with the Standalone um server offerings with open Ai and my my
1028. 2:50:21strong suspicion is that at a minimum you could run a gpt4 instance where you're not making a call to open AI
1029. 2:50:31I don't know about code interpreter within that um now the interesting thing about code interpreter as I understand it
1030. 2:50:45is when you fire up a code interpreter instance it's spinning up a virtual machine essentially within your chat right or
1031. 2:50:54within whatever the you know whatever the [ __ ] going on in the back end but it's spinning up a virtual machine and
1032. 2:51:00all of that python code that it's writing and the things that it's executing are are really only active
1033. 2:51:05when that virtual machine is active if you leave a code interpreter session for you know like a half an hour and come
1034. 2:51:12back to it you'll often lose all of the charts and the calculations that it did because the machine is shut down so a
1035. 2:51:21power move is if you're using Code interpreter copy and paste your results out to other other documents because you might lose them
1036. 2:51:28but but my instinct is is that anything that you do with encode interpreter is not being you know stored and sent to to open AI
1037. 2:51:38but like probably in a cached kind of way but I doubt it's being stored there but that's total speculation so if
1038. 2:51:49if if you if you've got private information you can't have get out then you need you need answers like from Azure not from someone like me
1039. 2:51:58I'm a crumb oh Chrome GPT love I thought you said crumb GPT that's funny there's a symbol I was talking about uh it's so
1040. 2:52:07revolutionary it made Microsoft browsers and search engines relevant again I know totally I avoid Edge like to play now I
1041. 2:52:14must reconsider well here's the deal Rebel so Edge is not bad here's the problem with Edge it it leaks memory like a a [ __ ]
1042. 2:52:23like minnow net it's it's horrible like like I I I am a I'm a tab hoarder so I usually have between 75 and 150 tabs
1043. 2:52:34open at any given time and I even have tabs shut down [ __ ] running and just every day and a half um Edge just shits the bed and I I have
1044. 2:52:44to come it'll like completely lock up my machine so it's it's just a dog that way but how they have chat EBT incorporated
1045. 2:52:52into it's pretty slick so it's actually worth it for for me um but if I need [ __ ] stable I go back to Chrome and just do do stuff over
1046. 2:53:01there but about 90 of what I'm doing I do an edge um just just because of convenience I I also I also wanna
1047. 2:53:10I've been trying to pay attention to how the implementation in Edge evolves like I'll notice that some days I've got
1048. 2:53:19a 4 000 token context window and some days it's two thousand and some days I can upload images and some days I can't
1049. 2:53:25and some days the little sidebar is aware of the web page to the left of it and sometimes it's not like it's it it's
1050. 2:53:33weird it's like they they're clearly doing hot fixes on all this stuff right as is open AI as are all these folks
1051. 2:53:42right they're all making hot fixes on the fly so what worked today won't work tomorrow and then all of a sudden just
1052. 2:53:48some [ __ ] miraculous feature will appear and you'll be like this is awesome and then it'll disappear that's
1053. 2:53:54just the nature where it is right now all right I'm gonna answer one more question and then I'm Gonna Leave You
1054. 2:54:00All for a lovely Saturday probably morning for some of you sorry about that I love Edge usability is great and low
1055. 2:54:07on resources too until it leaks memory and then it blows my [ __ ] up I've been on it for a year I've been on it for nine months now
1056. 2:54:15um I've grown accustomed to Edge can someone Define a token yes okay so that's a good question Lewis Louise Scott Louis Scott
1057. 2:54:25so so when they say it's got a 4 000 token context window or a 8 000 token context window um it counts so when you have a
1058. 2:54:37paragraph you've got punctuation and spaces and words so tokens are basically the little elements that make it up so so
1059. 2:54:46punctuation is a token I think spaces are tokens but they might not be it might just be punctuation and words and
1060. 2:54:52I think long words might even get broken up into into sub words so so they're basically the the elements of data that
1061. 2:55:02companies like open AI can count right so that's how they bill so for every thousand tokens you get billed
1062. 2:55:13three cents or two cents or six cents depending on the model you're using which is basically just to process this
1063. 2:55:20amount of combination of letters and punctuation um it costs them a penny and they're charging us three pennies or two pennies
1064. 2:55:29or whatever it is um what it translates into though is about three quarters of the total amount of tokens is words on
1065. 2:55:42average so if you have 4 000 tokens that means three thousand words or in claude's case it can do a hundred thousand tokens
1066. 2:55:52which means about 75 000 words it might be more or less than that depending on the kind of data you put in there but
1067. 2:55:59that's it's kind of a rule of thumb um so it's and then the context window is basically like in a given chat how much
1068. 2:56:08sort of can it remember or how much can you input and and that's the limit so so sometimes if you take a long document and you paste it in
1069. 2:56:19the chat GPT it's just like I can't do it it's too big with Claude you can paste it in and it it does it and it's you know the way it
1070. 2:56:26works is it's a combination of your prompt plus the response it's that whole thing has got to be under 4 000 tokens
1071. 2:56:34or 8 000 tokens or in the case of Claude a hundred thousand tokens so that's that's what that that means it's basic it's it's basically
1072. 2:56:43the components of a set of words that get processed if I wanted to make a video game developer assistant that knows free
1073. 2:56:54software where should I start hmm well you would want to start with some of these tools that are connected to the
1074. 2:57:04internet so you might want to start with Bard or Bing but
1075. 2:57:19I I if if you want to start building something that's sophisticated probably something like Lang chain where
1076. 2:57:28you're going to start embedding you're going to take data sources um and embed them and get them in a vector database so you can GPT them
1077. 2:57:37um so I would start exploring the Lang chain project um and then and then I would think using the Llama 2 models um within Lang chain
1078. 2:57:46just because they're new and commercially viable that's where I'd start all right I'm gonna do one more question and I'm gonna go Gotta Go
1079. 2:57:53Silver Fox see you later uh um I want to create the same thing can we use AI to help correct students
1080. 2:58:04essays and compositions yes and they're probably using AI to help write their essays and compositions so yes you can I
1081. 2:58:13sent you a short AI created on Tick Tock yesterday well yesterday okay cool all right I'll check it out all right everybody um
1082. 2:58:20I gotta take off thanks for hanging out with me on a Saturday night my eyes are bloodshot my voice doesn't work anymore
1083. 2:58:27my ass is numb that means it's been a good night I hope you enjoyed this it's been great hanging out with you peace
1084. 2:58:33out everybody I'll see you tomorrow