aiLL 8/30/23 - How AI Will Transform Every Job and Industry

Transcript

1. 0:00foreign
2. 0:26Luna Bev what's shaking what's happening it's going down it's going up what day is today Wednesday the 30th
3. 0:35I know that because I had two two two speaking gigs today I spoke to a large pharmaceutical company hey Tobias was shaken welcome
4. 0:48and this evening I had a like an elevator pitch workshop for a bunch of first-time founders that was cool it was Nate
5. 1:00we came up with a new kind of elevator pitch I called the belief brief so it was it's sort of based on the
6. 1:08ideas behind Simon Sinex uh lead with why like people don't want to buy your product they want to buy
7. 1:15you know why you why you do your product and uh it's cool that's neat hang on one sec
8. 1:46it was Vada warm here in Denver today that I've got a warm here in Denver today okay farmer and Tobias rocking out
9. 1:56you guys in the same room are you sharing a bowl of nachos I would assume the nachos are long gone by now I'm a little late today
10. 2:03I just got back from my uh from my little speaker gig uh hi trying lots of your learning stuff suggestions today awesome awesome to Gus
11. 2:14that's great yeah I noticed you were you were making some comments on some of those awesome very fun yeah it's fun
12. 2:19when you start playing with this stuff isn't it um let's see that was the presentation I was doing there's a cool thing I made for
13. 2:31the little class I just took or taught and there was another one these are from uh these are from ideogram
14. 2:41so these are AI generated word pictures word pictures the dude's a little a little distorted like his face is all smashed in like the
15. 2:52the quality of the image rendering with this tool is um is uh like stable diffusion 2022 like mid 2022 but the uh but the the
16. 3:06letters are good they're not bad so and like I had like five minutes to knock this out so you know if you need something that's just
17. 3:15good enough it's good so that's that's one of them there's another these are all yeah learning lab things
18. 3:22aren't they cool Kiel aren't they all right welcome everybody my name is Kyle Shannon this is the AI learning lab is
19. 3:29evidenced by all the AI learning lab paraphernalia in the little window here um welcome Irregulars we got some Irregulars in the house um
20. 3:44so so this channel we talked about generative AI I answer as many questions as I can I try to do this every night
21. 3:51um I'm going to give you a big long list of hold yeah those things URLs if you're if you're just trying if if you stumbled
22. 4:00on this and you're like what's this old guy talking about um or you've heard of AI and you're like well yeah it's all right but I haven't
23. 4:09tried it yet I'll wait for a year or two I'll wait for it to get better um I would encourage you not to wait
24. 4:15that's what this channel is all about get curious start playing now um if you haven't played with chat GPT
25. 4:21that's that's the official URL right there uh encourage encourage more viewers to share yes Emilio's wife has it right if you'd be so kind
26. 4:31um share this live even if your friends think you're weirdos because you're sharing a live of an old man talking
27. 4:37about Ai and they're like oh you're not falling for this AI crap are you just do it anyway do it anyway you're gonna get
28. 4:45sucked in you're gonna get sucked into the lab here at the lab we bring people in they're Irregulars they're they've
29. 4:51named themselves that they're the Irregulars and and they're awesome and you'll be one of those too but go to chat GPT
30. 5:00if you don't know what to say there go to prompts.chat that will teach you um how to talk to these large language models
31. 5:08um share or not share share or not share that is the that is the vision all right and then if you have questions
32. 5:17about AI pop them in the comments below but let me let me walk you through have played with chat GPT there's a couple other things here to
33. 5:27look at we got um Claude and we got why is this not aligned looks like I moved it we got Claude and we got
34. 5:37Bing and uh Claude and Bing are I think of these three together because they each kind of do some slightly different things um chat gbt
35. 5:47has gpt4 this is the best one it's also got code interpreter which is really quite remarkable Bing is Microsoft's version of gpt4 chat GPT
36. 6:00um it's connected to the internet you can make images with it like you just say make an image of whatever and it
37. 6:05will it's based on the dolly 2 model which is kind of shitty but it's it's still free and it's it's fine um
38. 6:14it can also analyze images you can upload images there and talk to them Claude has a really big context window
39. 6:20meaning you can upload a really large document and read it and interact with it which is kind of cool like big like 50 000 words
40. 6:29you have a novel 50 000 word novel you can literally copy and paste it into the prompt and tell it to write you an epilogue and it will
41. 6:38um okay what else do I think is interesting to play with Bard that's Google's entry perplexity.ai is a good research tool
42. 6:46and labs.perplexity they're hosting meta's llama 2 models and their uh code llama model that are worth going and exploring
43. 6:56because those are open source so if you want to roll your own large language model you can do that with that Pi dot
44. 7:03AI That's chatty GPT if you want to have a conversation with a GPT model that's your one Pi dot AI or the PI app and
45. 7:11then po.com is lots of different models you can play with a bot maker things like that what I like too much it's freezing stuff
46. 7:21all right it looks like my pronouns are turtles oh the kids the kids are spicy tonight I like it I like it a lot I like it a lot
47. 7:35um what else pictures you want to do pictures I got pictures for you here's what you're gonna do for pitches and
48. 7:41feel free to screenshot all these by the way I'm not affiliated with any of these I don't give a [ __ ] if you go use them
49. 7:46or not I'm not making any money this is these are like public service announcements mid Journey's the best
50. 7:52ideogram or ideogram I'm gonna figure out how we pronounce that at some point but that's the one that can make um
51. 8:01images with letters in them like this and this and this and this and this all these were made in ideogram it does
52. 8:09letters it does words now does It screw them up about 75 of the time yeah it's just gobbledygook 75 of the time but you
53. 8:19try getting mid journey to do this good luck good luck so this does that now its pictures are not anywhere near
54. 8:28Mid Journeys but with a little creative prompting and photoshopping you could probably merge some stuff all right um what else Leonardo
55. 8:39really cool website sitting on top of stable diffusion clip drop is from stable diffusion from stability I see
56. 8:45the uh the Irregulars are apparently their carpal tunnel has healed they're feeling a little bit better and they've
57. 8:52taken their Advil so all of the hearts that you see you're like what do they like this guy this much what he's not
58. 8:59even saying anything that this is a game they play to try to try to get the likes up to whatever their record was the last
59. 9:05time we did this so I appreciate it because it's fun seeing all the hearts fly um and they appreciate it because it
60. 9:11takes their mind off whatever the [ __ ] stuff I'm talking about and then uh openart.ai promptbook that will teach
61. 9:19you how to talk to these things it's designed for stable diffusion but the ideas in there really apply to anything
62. 9:25at futurepedia.io is a big old uh directory of goods and services um we used to have a scammer but that scammer has been booted
63. 9:36so yay Irregulars for thank you for helping me uh figure that out and get that [ __ ] tossed hopefully it doesn't just spin up a new
64. 9:48one but you know I think you got the I think he got the hint that there was a it wasn't just me he was battling and I
65. 9:56think those those scammers go for the path of least resistance so anyway so pop quite my best to answer them getting
66. 10:03let's start tonight so not going to be here super late but uh but we'll get we'll get to it so let me flip this flip
67. 10:12that AI learning lab is turning on the AC now yeah it was like 100 degrees in in Denver today so my little home office
68. 10:19here the the lab the lab she is pretty close to Melting glass there we go there we go there we go
69. 10:36lovely all right fantastic Rob tell him what he's won you've won a new car
70. 10:53if you have questions about how to make money with chat EBT I'll tell you but I will make fun of you okay
71. 10:58fantastic let's get this going you know he gets in here late he said he's going to come here the same time every night
72. 11:05by the way I never said that he says he's kind of come in here same time every night and then the last two
73. 11:12nights it's been he's been like I don't know I'm a little sleepy I'm not coming I'm not doing it that's true what do you want
74. 11:21am I late no I just got here um why do children oppose us I never did
75. 11:35that to my oh right never mind I know Thelonious what are you gonna do kids these days get off my lawn your
76. 11:44little whipper stampers there's probably some youngster in here right now getting ready to go oh isn't a cute go screw yourself children
77. 11:57this is this is this is a Gen X friendly Zone and that basically means you give us Gruff we're gonna slap you
78. 12:05when we were kids we were allowed to be slapped and people took advantage of that even if they weren't your parents
79. 12:15you got Lippy with an adult the community took care of it none of this flipping off your parents thing you flipped off your parents in my
80. 12:25day you get a whack upside the head from someone you didn't know you're like the [ __ ] was that well you wouldn't swear because you just
81. 12:31got hit in the head and you were a kid but you know you get the basic idea all right let's see survived oh idalia
82. 12:39in Florida what's up Kyle good uh M Phillips what's what's I hope you're well um I've got family in Tampa that reminds me I gotta call them
83. 12:49um how bad was it I didn't I didn't hear a ton on the news so I don't maybe it did it not hit a major populated area
84. 12:56I know there was flooding but I didn't see a ton on the news but then again I haven't been on the news so
85. 13:01USA loves tick tocks I like turtles uh loving your short videos that pop up on my feed thank you I appreciate that yeah I posted some
86. 13:12um some of the old some of the original videos that I did talking about how chat gbt works and how uh diffusion image generation works um
87. 13:22I should probably redo them because there's there's a couple of things in them that that I know more than I knew then
88. 13:28um but but I also feel like there's a lot more people on the channel now that um that weren't there when I made them
89. 13:35and so I figured it would they would be good so I'm glad you appreciate that buffering for others probably how many hours do you spend on AI
90. 13:46um I mean I've got AI I don't have it formally Incorporated in my workflow but I kind of use it for every like whenever I'm doing
91. 13:59any kind of project like it's just a normal part of my workflow now it's it's kind of like I don't know like
92. 14:09you know you've got your Adobe suite you got Photoshop you got InDesign you got illustrator you got word or Google Docs
93. 14:16you got Excel like depending on the project you fire up whatever tool you need and knock that [ __ ] out
94. 14:23and I just feel like it's another tool in the toolbox and and so I would put image generation in that category thank
95. 14:29you steva and I would put the you know this these large language models in that category like what did I do today I um
96. 14:38oh I did I did some things for the uh for the talk that I just gave um and what else was I doing I was doing something else on it
97. 14:52I came up with a flash card kind of thing that I'm really excited about I did that with it so yeah I'm using it so hours a day
98. 15:01I don't know probably two to three using it and then maybe an hour or two surfing Twitter and Linkedin
99. 15:11and Tick Tock and just trying to see if anything's discovered anything I haven't and uh and then two to three hours a night doing this
100. 15:25and while while I'm not necessarily learning here sometimes I end people come in here and they'll tell me stuff that I didn't know
101. 15:31um and and you know figuring out what to put on these lists that I show people that's that's uh it sort of keeps me on my toes so
102. 15:41that's it keep it for us you saved me I was saying the moon does not turn Below 100x on Sky life oh just yeah I know
103. 15:51I know supposed to be uh it's supposed to yeah the moon was really beautiful when it came in but it wouldn't it we're not blue
104. 15:5810 friends wait what Nicholas shared the live with 10 friends thank you Nicholas you rock all right my pronouns are turtle listen
105. 16:09harsh sweet I think that deserves a little little gift s I should probably show you this uh
106. 16:25see the Army regular right here I'm irregular guess who did that yeah yeah boy yeah boy
107. 16:40uh um let's get this party started shall we
108. 16:57are here live at the waterfront village with my friend the zombie Jonathan you're looking good Jonathan just got an
109. 17:03awesome face paint job what do you think I like turtles all right your great zombie good times here at the waterfront
110. 17:12Village good times here and it still cracks me up still does it it just don't grow old it don't grow old
111. 17:22it don't go out it don't you know what I'm saying all right well let's see not wait we buffer now good who said Thelonious that what
112. 17:36what random collection of words did you type into that comment not wait we buffer now good [Laughter]
113. 17:52polishing up my English course coming laughs ideogram oh per their Discord it is ideogram huh interesting
114. 18:07ideogram yeah I wonder if Ideo is gonna Seasons assist them because if I were them I'd call it idiogram so that Ideo doesn't go on ahead and
115. 18:20sue me we got our tappers from the song Oh that's very cute uh hey Uncle Kylie Troy what's shaking uh nice shirt Choice thank you very much
116. 18:34yeah that was I I gave a a talk to a large group of marketing people in a uh pharmaceutical company today a rather large one which was a
117. 18:47little nerve-wracking but it went really well and hopefully I blew some minds and inspired some people I showed them
118. 18:55what did I show them I showed them um I showed them that slide the November 30th slide I don't know showed him a bunch of stuff
119. 19:06it was cool it was good comma [Music] I am friend Jesse
120. 19:21hello friend Jesse Jesse logsdon the name is Jesse logsden I am friend Jesse logsdon you see me regular
121. 19:36hmm curious you should check with others here there are others like you they're not regular all right got an idea of
122. 19:51of his location will provide first letter okay Molly just got here too welcome Molly Stevo what's up buddy what's shaking
123. 20:00while shaking our Stevo I love watching you while I'm working out nice nice this yeah you're just working out going I
124. 20:09gotta get skinnier than him well that's not gonna be that hard laughs good Lord yeah I guess I'm the I'm the before picture
125. 20:20I'm your before picture hopefully I inspire you Matthias thank you thank you I'm just learning thank you
126. 20:28we're all just learning if you're just learning this stuff and you haven't done it go to chatgpt chat.openai.com
127. 20:35and then in another tab or window open prompts.chat that will teach you how to talk to these things it's a good primer
128. 20:42there's all sorts of stuff you can do Beyond this this is a pretty basic document but um it's a really good starting point
129. 20:50welcome and if you have any specific questions pop them down below Steve oh hey regulars hello hello Aristophanes 23 I like the name a lot
130. 21:01thymelonious Ted looks at life if big Tech continues to dominate Enterprise for Gen AI due to expense what will startups need to do
131. 21:21so it depends if you're talking about the frontier models if you're talking about actually developing large language
132. 21:28models like gpt4 gpt5 Bard Gemini like those those kind of lower level base models that's always out of partner you
133. 21:38know life is tough but it's tougher when you're stupid um I'm such a trainable idiot
134. 21:52um so so I think these the the big large language models are just just by the nature of how expensive they are to make
135. 22:01thank you togas I appreciate the pizza well you reached level 10 to gas is at level 10 he's at level 10. this channel is very much like
136. 22:12Scientology if you level up you can reach the next level do you know what I mean [Laughter] um
137. 22:26so so I think at that level the big companies are gonna are gonna be they're gonna be the players right um
138. 22:38meta seems like they're going to keep open sourcing stuff at some point they're probably going to have a model
139. 22:42strong enough that they won't open source it and then people will have developed [ __ ] for their other models and they'll start paying them
140. 22:51um I think that in terms of usage of AI I think large companies are going to be preoccupied with their own data for a
141. 23:03lot of years right so how do we get all of our data you know into these models the the chat gbt Enterprise that just launched
142. 23:12um they said the the sort of next thing that's coming is they're going to make it easy for you to ingest your own data
143. 23:18and ingest permissions and all that sort of stuff which is super exciting but I think that's going to be a big
144. 23:24complicated you know if you think about the past 50 years of data sitting in some of these big Fortune 500 companies
145. 23:33they're going to be spending lots and lots of times just dealing with keeping up with this stuff figuring out what to
146. 23:40put in the models but not to all that sort of [ __ ] um so I think there's massive opportunity for smaller companies it's
147. 23:48going to be on the application layer right you got your base model down here you got chat CPT here you got things
148. 23:53like copy Ai and Jasper sitting on top well actually copy Ai and Jasper are sort of fancier cousins of chat gbt
149. 24:02um so so at that application layer there's there's massive opportunity the the challenge the challenge for
150. 24:09small businesses and quite frankly the challenge for big businesses any business that you come up with and any
151. 24:15technology that you come up with someone can knock off like that like trivially like give me a weekend you get someone
152. 24:26that's decent with coding and with chat GPT and and couple of the other tools for doing interface [ __ ] and images
153. 24:34like no business is safe from from just getting cloned getting knocked right off that I don't have any sense of how you
154. 24:46play that game I think we're gonna need to figure it out whack hell yeah Irregulars I miss those days they won't
155. 24:57understand our childhood days they will not Swift Warrior Swift Warrior they do not understand the drinking from the hose
156. 25:07why would you drink from the hose because we weren't allowed in the house get out
157. 25:22oh man it was fun times it was simpler times in my day stopping by to say hi hello and what's happening congrats your LinkedIn post
158. 25:37was awesome by the way um congrats on what you're doing it's really like it was really well articulated really well put out there it
159. 25:44was really generous um it was good solid kudos Kudos Kudos Kudos my first floor apartment near Saint Pete Beach only
160. 25:55leaked two inches that's not bad Silver Fox sorry you're having to deal with it at all but you know floor Florida hurricanes mm-hmm
161. 26:06but it's weird now like this is now two two or three in recent time that are you know coming golf side in into Tampa so it's bizarre um
162. 26:22your work is so epic why thank you which work this stuff this groovy stuff [Music] yeah some of this stuff's pretty cool
163. 26:38I like that one a lot Hey look whoops isn't that groovy I mean it's not bad you know
164. 26:55it's the other faces are just just off it looks like a bad sci-fi movie poster like back back when you know they
165. 27:05couldn't figure out how to actually get photographs on movie posters so they were sort of semi photo realistic illustrations of people
166. 27:12hahaha um that one's cool too
167. 27:27you know the these image tools are really really fun when like like I didn't really care like what it came up with I just put like
168. 27:3690s graphic poster you know that says AI learning lab and then I did I did 70s poster 80s poster 90s poster 2000 poster 2010 poster
169. 27:51and I just did a pile in a row and it just did all this [Laughter] so pretty cool oh let me
170. 28:05it's Ideo now we know it ideogram.ai and then you know it's just like it's just a very simple uh image generation tool it's web-based
171. 28:21you don't have to go to Discord and it does text it does good text it also apparently does like NFL like logos so let me try let me try um
172. 28:35football with um Denver Broncos logo on 50 yard line
173. 28:51and then we'll do photo cinematic we'll do horizontal see y'all uh uh all right let's try that
174. 29:06and so you can you can pick your aspect ratio one of three aspect ratios and you say generate and it's decently quick
175. 29:16I'll be nicer when you're smarter that's very AI learning web kind of smarmyiness I like it it's solid that's 10 out of 10 and on that
176. 29:29um oh yeah that looks like a Broncos logo look at this one to the right here that looks like it's gonna be one yep oh and oh this is cool
177. 29:42oh that's funny it took football to mean soccer but it sort of put the Broncos logo on that one but this that's actually pretty close it's a
178. 29:53little it's a little distorted and weird but so so now that I've got that image I can remix it and let me throw I don't want typography in
179. 30:04there we'll do vibrant generate and so now it's going to do form work kind of based on that as the seed I guess
180. 30:17oh this is hilarious Zuckerberg and Elon Musk kissing kissing during their boxing match oh internet you're so inappropriate well
181. 30:31you make me uncomfortable in am I offended or excited I just don't know um it didn't really do too well with the Broncos logo this time but
182. 30:44anyway um that's interesting it can sort of do it that's a really good post a really cool typography that's really cool
183. 30:54so let's see so one of the things that I've been doing here is you can either click on this and you can just remix it
184. 31:01or you can just grab the prompt like many of these tools and just I'm going to get rid of gonna get rid of the remix I'm going to
185. 31:13paste the new thing in there so what's it say I'm not weird I'm limited I'm limited edition okay so I'm gonna do I'm irregular
186. 31:31typography funny sticker cool dragon so this is going to be cool turtle with glasses white background conceptual art 3d render vibrant
187. 31:42so let's do see y'all 3D render get rid of that get rid of that um where's my pen there it is
188. 31:59all right let's give this a shot oh I should wait what was it doing here I probably should have done Square [Music]
189. 32:19there's I'm irregular there's I'm irregular that's pretty good these are gonna be all right oh that's good actually this kind of
190. 32:28inspires me I think I think if we do I'm a regular it's got to be misspelled I think it's Gotta Have a turtle on it that's pretty good um
191. 32:40let's remix this one I got a marketing call from an AI data firm Hocking a product called Med charts
192. 32:55offering a demo oh there's gonna be the amount of inbound marketing inquiries we're all gonna get from
193. 33:04AI powered this and that is just going to be staggering by the way black coffee because I ran out of milk um I'm irregular I mean ruler
194. 33:20I'm ERG L lap wait I'm a I'm a ERG L LARP LARPing yeah it didn't do too well without me regular when can we delete these images
195. 33:34that's pretty funny so anyway yeah go play with this actually you know what would be fun is we did this last night at the AI salon go to
196. 33:47ideogram.ai and then go to if you're a member of the AI salon Discord go there and and um
197. 34:01post some of your AI Salon or or just whatever some of your ideogram images in the show and tell Channel if you're not
198. 34:09in our Discord go to the salon.ai it's a link tree I think it's the third link down is the link to our Discord go join
199. 34:16our Discord go make some images and ideogram and then go to Discord and drop them in the show and tell Channel that's
200. 34:24cool it's a neat group that group we uh we meet every other Tuesday so there's also a link to the meet up there there's a link that
201. 34:32explains what the group's all about it doesn't cost anything it's just you know it just it's just there it's just there
202. 34:40to support people you know we're just trying to help we're just trying to heal that's all we're trying to do
203. 34:46we're just trying to help it out you know what I'm saying you know what I'm saying I don't know if you know what I'm saying
204. 34:53yeah I'll leave that one there for now that one's pretty cool AI learning lab yes she's very sexy Kyle that is a salty
205. 35:05little picture of a robot hmm sexy robot sexy robot all right how do you run a Python program from Windows prompts um just go to replet
206. 35:20togas I go to uh when I when I first used chat GPT to write python for me so this was my process uh write me some Python and then
207. 35:34the realization I don't know how to run python how do you run Python and it told me how to do it on a Mac
208. 35:40and then I said no I just want to do it on the web and it gave me replit was the first site that it recommended so in
209. 35:46replit you just paste the code in the left window you hit the Run button and it'll run in the right window and so you can just
210. 35:54pop back and forth between chat GPT and replit and just if it gives you an error in replit just paste it over in chat GPT
211. 36:03and say you gave me this error tell me what to do and it'll walk you through it took me like 90 minutes and I made an
212. 36:09application it was absolutely mind-blowing Nicholas has 10 friends Nicholas may not be an irregular yeah like you gotta you just just the
213. 36:19Irregulars alone you're gonna have 10 friends I like totals uh um Play Frankie bit my finger what
214. 36:34what is happening oh Tick Tock jumped me to the bottom thanks for the licky Pop I like licky pops licky pops are tasty
215. 36:43I know they're not called that but you know I'm old we got to give the young people something to make fun of
216. 36:49uh how was your morning Pharmacy speech for the marketing department well they sell more drugs now um it was good it was for for a
217. 36:58pharmaceutical not a pharmacy department but a Pharma company um it went really well I mean you know it's
218. 37:07a big Corporation it was it was a big meeting and a big Corporation so you know people aren't going to stick their
219. 37:13necks out and ask questions that might make them look stupid but we got a couple of questions people said they were inspired the woman
220. 37:21that organized it thank you very much oh dumbbells oh yeah working out pumping iron um yeah she said it was awesome and she
221. 37:30says she thinks it thinks it's going to inspire you know a lot of activity across the organization I'm hoping part
222. 37:37of that activity is they reach out to storyvine and go we want to tell more authentic videos too he's not just about
223. 37:44AI it's about authentic human beings authentic human beings that are important they're more important than ever
224. 37:50you with your dang robots you're all just doing Ai and they're and the robots are stealing everything from everyone what about the people
225. 38:00I say we strike we go on strike and we say to help with these robots let's throw a party I'll take a old-fashioned blantons thank you
226. 38:16so yeah you know that I am Ascension potato Jess Jesse logsdon I am a fan
227. 38:35a good lordy Shimmer what's happening welcome welcome welcome play Ralph me fail English that's unpossible hello hello from Jesse crazy
228. 38:49yooper okay crazy yooper 9.5 out of 10 on the name I don't know what take it to a 10 but 9.5 solid name crazy you per
229. 38:59thumbs up love it Firefly great to hear the meeting went well thank you very much I'm Ascension potato we got that
230. 39:05Mitch McConnell froze again not being political no yeah I so they've actually figured out why Mitch McConnell's freezing he's got Xfinity
231. 39:14[Laughter] joke cast sucks All Right Moving On
232. 39:31I got chat and Bard working together today pretty cool they came up with a fantasy frog football that's awesome
233. 39:37that's very cool I you know it's funny I find that that idea of like flipping back and forth between um Chachi PT Claude and either Bing or
234. 39:49Bard for me those three I I use them if they're I don't use them interchangeably I I basically what I do is I I'll use them until one
235. 40:01of them frustrates me and I'm like oh [ __ ] you chat EBT and I'll go to Claude and then it'll be good for a while and
236. 40:06then it'll rate limit me and say you've got three things left and I'm like oh [ __ ] you and then I'll go to Bard so I
237. 40:12can find myself just bouncing back and forth because they're all some version of good right they can all do something
238. 40:17so that's how I've been using them but yeah that's interesting when you get them working together like that one of
239. 40:23the things that I've heard is really good is um I haven't done this myself but it's a really good idea because Claude has a a big context
240. 40:33window it's got a hundred thousand token context window you can put really big documents in it so if you've got a big
241. 40:39document but you want to do some like good writing on it you want to whatever you want to write copy out of it or you
242. 40:45want to whatever it doesn't matter go to Claude with the big document have it summarize it and sort of reduce the
243. 40:52word count basically and then take the summarized version of the long document into gpt4 and do all your all your work
244. 40:59uh in gpt4 because it's just a better writer I thought that was a a clever thing I didn't come up with that but I
245. 41:06thought that was a clever use of mixing these two things uh if you're new here my name is Kyle Shannon this is the AI
246. 41:11learning lab if you haven't used chat GPT that's what that URL is right there chat.openai.com prompts.chat will teach
247. 41:19you how to use it if you're sitting on the sidelines going I don't wanna with no not not in here if you're in here you're curious you're
248. 41:30adventurous you're like okay Kyle I'll use Chan gbt and then like five minutes later you'll be like all Kevin Mcallister on us going hey
249. 41:45Kyle I don't know if you knew this but these things are pretty [ __ ] powerful we'll be like yeah duh
250. 41:51we could have told you that like an hour ago because that's how we talk here foreign we're neuro spicy we're irregular it's
251. 42:05not normal in here so if you're gonna be in here get curious start playing tell it to write you a country song about a turtle
252. 42:15like a good irregular wood tell it to make you a mid-journey prompt about a turtle that isn't right take it to Mid Journey make a picture
253. 42:26take it to ideogram make a picture with text on it and just get weird just get weird just like a dead concert without the music or
254. 42:36the talent it's just like that or the drugs we're we're all hopped up on declining hormones [Laughter]
255. 42:59what does he talk about I thought this is the AI learning lab he's just talking about being old and what's with the nose hairs
256. 43:08all right let's see I can't pay for Claude yet and I don't really like Chachi BT what can I use um well so so Scott
257. 43:18I'll give you seven so Chachi PT is one of them I don't I don't know I don't get why you don't like chat GPT if if it's because
258. 43:28it's not connected to the internet I understand that but you can use the plug-in web pilot to read web documents
259. 43:35so there's kind of a hack for that code interpreters quite remarkable and gpt4 it just it's just better than all the
260. 43:44other large language models but I don't know what you're trying to do with it um Claude is not bad wait what did you say
261. 43:52you can't pay for Claude yet claude's free so cloud.ai is free Bing is free Bing is free and connected to the internet the
262. 44:01other four that you can play with Bard is also connected to the internet it's also free labs.perplexity.ai I think has a free
263. 44:10and a paid version um but well no labs.perplexity.ai is free that's the the Llama models I think perplexity.ai
264. 44:21the research tool is both free and paid Pi is free and is connected to the internet but that if you don't like uh
265. 44:32chat gbt you're gonna [ __ ] hate pie like pie is if you want to have a conversation with a bot like if you want to know what AI
266. 44:40dating a robot's gonna be like go talk to pie if you want to get [ __ ] done go don't talk to pie and then Poe lets you play with multiple
267. 44:49models and it's got a whole bot system and you can make your own bot so there's seven four here and three here um and most of them are free
268. 45:01hey Kyle hello Robert Ross here what is happening good to see you good to see you sir um lots of side conversations going on I
269. 45:13love that hey Kyle cowboy hat to you is like the Bell to Pavlov's dog it really kinda is it's it's very
270. 45:22um it doesn't matter how deep I am in a conversation the cowboy hat just demands me do something and I still haven't like
271. 45:32quite I haven't done the homework to get the John Wayne accent and quote right it's like that's like homework and me and homework oh
272. 45:43I had that I had that um that Pharma talk today at 7 50 is when I went on and I'm like I'll get up at five
273. 45:53and I didn't get up at five I got up at six I got to the office at seven and I'm like I'm like cramming like the half
274. 46:00hour before I'm I'm going on I'm like in chat GPT writing stuff I'm doing research I'm screenshotting [ __ ] I'm
275. 46:07updating the deck I'm making sure all my builds are correct I was done with seven minutes to spare it was fantastic
276. 46:18the nice thing about working like that when you have to do a presentation is if you finish it right before you go on you know everything
277. 46:28because you just finished it it's all still in the Noggin um did you hear about the AI app called talk to Jesus where you can talk to him
278. 46:37or anyone in the Bible you could probably you could probably make a PO bot for that um Bradley from the AI Salon the other
279. 46:47night oh this was cool so what Bradley did he's got a he's got a an agency called Redline which is an AI Centric
280. 46:57like virtual agency where like all of the employees are going to be Freelancers it's really cool um but he wrote
281. 47:07a Persona of himself like here's my educational background here's my this he basically wrote like a whole thing of who he is
282. 47:14like how he thinks how he writes all that sort of [ __ ] and then he condensed it down and he made it into a bot a pobot poe.com
283. 47:25bot version of himself and so when he wants to ideate now he's got a version of himself that he sends raw ideas to
284. 47:32and and the the bot version of him spits back all these ideas basically the way he would create them it was such a cool idea
285. 47:45um did you hear about the AI app oh yeah we just did that one will there be a need for a data analyst I don't know Pam
286. 47:52so Pam if you haven't played with um code interpreter you need to play with code interpreter so let me go a lot of
287. 48:01people are talking about this largely because it's like I don't know a ton about data analytics but I know that this is
288. 48:10[ __ ] remarkable oh they changed the name look look people we did this we did this Pat yourselves on the back
289. 48:22so you may not know this but Sam Altman Andre carpathi are constantly in here they never speak but they're constantly
290. 48:29in here they're listening to us the whole time they're they're trying to figure out Kyle how should we run open Ai and like they
291. 48:37haven't emailed me yet but I know they are so anyway like two months ago I said the name code interpreter was a poo-poo
292. 48:47name for this feature and they changed it Advanced Data analysis which is what it is so so Pam 39.72 the the bad news
293. 49:01for your question is that this tool is a data analyst the good news about your question is that I firmly believe that someone who
294. 49:13knows something about data analytics is going to be able to do [ __ ] amazing things with this me as a person who doesn't know [ __ ]
295. 49:22about this I can still get it to do stuff that blows my mind but but you know I'm mostly Flying Blind with it but the way it works is um
296. 49:33you go into what they call it now Advanced Data analysis that's a little dry but you know they're doing this for
297. 49:39corporate right this is this is a corporate tool um so you now have an upload button here and I don't I don't know what the limit is on
298. 49:48files to upload it used to be one but now you can do more um but I'm gonna go grab a zip file so this is
299. 49:56this is a ZIP file of some data set that I grabbed off of uh kaggle so if you're a data analyst you probably know kaggle
300. 50:04and then I'll say please let me do it with this please tell me what's inside this zip file and what's interesting about the data
301. 50:21I'll just leave it there oh wait so so let's start by extracting the contents of the zip file so so I tell it
302. 50:35I want it to look inside the zip file and tell me what's interesting about the data so the first thing that it does is
303. 50:40it fires off this plug-in like thing and if I open that plug-in like thing this is python code where did it get
304. 50:52that it wrote that so what happens with advanced what's it called Data analysis formerly code interpreter when you launch it it's
305. 51:01firing up a virtual machine and then it's got access to certain python libraries and it's got python no
306. 51:08right now it only knows python in the future I assume it will be more sophisticated but you give it a goal and it starts writing
307. 51:18its own Python and instructing itself what to do so so it's this is a rudimentary version of an autonomous agent
308. 51:27so so it went and pulled that out it said the zip file contains 15 files 14 of which are csvs seemingly random uuid like names
309. 51:37the 15th appears to have a name suggesting it might be related to education statistic all indicators right
310. 51:43so so it's starting to learn what's in here so to better understand this data I'll take a look at the first few rows
311. 51:49of one of the CSV files and determine the nature of the 15th file right so it assumes 14 of them have the same
312. 51:56structure so it's just going to go look at one like you saw what I typed I just said like look what's in here tell me what's
313. 52:03interesting about it so there's some more python it brings in pandas figure some stuff out looks at what's in the in the file
314. 52:14so it says the 15th file contains data related to education statistics country name country code series series code gear based columns
315. 52:25so so say to understand the data's breadth and depth we should check if all the CSV files have similar structure
316. 52:32it would be beneficial to know the unique metrics series available across these files shall we proceed with these steps I'll go yes
317. 52:42and now it's going to go off and do some more [ __ ] so as a data analyst this is doing a lot of what you probably do now having having
318. 52:53tried to use this a fair amount of times now for actual work products this thing can get really [ __ ] frustrating so so
319. 53:01it demos really well but then it seems to get confused like it it you'll give it a data set and you're doing stuff with it and
320. 53:11everything's going fine and then all of a sudden it just seems like it must knock something out of the the data
321. 53:15frames in its virtual memory and then it just is lost and you can't seem to get it back so it can be very frustrating so
322. 53:22I don't think this in its present form is threatening any jobs but this is just the beginning like this is
323. 53:33the worst it's ever going to get right um and you'll see I'll I'll let it do something here and then we'll have it
324. 53:40we'll have it generate some charts and graphs because it can do that too it had an ISO error so it's going to change ISO types
325. 54:02I'm going to say do what you recommend so it's basically saying we've got a couple of files with different ISO types and it's causing
326. 54:11issues and but we can merge the data so I'm just going to say you tell me what to do so it says given the nature of the
327. 54:17data we're going to merge the data data from all the files we're going to fill in missing values with not available or similar similar
328. 54:25notation so it's cleaning data it's identifying that some of these things don't have the same data sets or it's
329. 54:32got incomplete data so it's going to fill it in with you know some option again work work that a data analyst
330. 54:38would do let's proceed with this approach the data from all the CSV files has been successfully merged so 14 csvs or I
331. 54:47guess 15 csvs were now merged into a single data frame there are 775 000 rows 54 columns given this comprehensive data set we can
332. 54:59now explore various aspects of the data such as Trend and educational metrics um give me the three most interesting
333. 55:09trends for the United States over the past 10 years see what it comes up with for that
334. 55:26we're going to filter the data for the United States we're going to focus on the last 10 years we're going to analyze
335. 55:30and identify significant changes fluctuations or patterns it's working let me start by filtering the data
336. 55:41so it's writing all the python code to filter the data I mean listen as a as a non-python writing you know business product guy
337. 55:52I I I find this pretty [ __ ] mind-blowing but again as a data analyst you're going to look at this in a very different way
338. 56:03so it might be brilliant it might be less brilliant the videos that I've seen of data analysts looking at this their
339. 56:10jaws slap the floor pretty hard I mean the the thing that that potentially does take away well so here's an interesting thing
340. 56:24like even if I had a hundred person company I probably wouldn't hire a data analyst so so a tool like this for small to medium companies
341. 56:35doesn't mean that they're um they're gonna get rid of a data analyst they probably don't have one what this
342. 56:44tool means for smaller companies is that they can now do stuff that they wouldn't have done otherwise because they're
343. 56:49probably not going to hire a data analyst if they're that small big companies will do it um here are three metrics so I'm just gonna say
344. 57:00can you visualize those three trends in a visually stunning way let's see what it comes up with like I
345. 57:10didn't even look at what what it came up with of course a good way to make visualizations more engaging is to use
346. 57:17clear and distinct colors labels and possibly interactive elements here I'll plot these Trends using line plots
347. 57:25to show the changes over the past 10 years I'll use distinct colors and labels for clarity so we'll see how it does there you go
348. 57:36it looks like there's only two or one of them is the same yeah they've missed one so this is this thing where
349. 57:46like you're gonna have to like go through this stuff with a fine-tooth comb and you know make sure it doesn't
350. 57:51hallucinate make sure it doesn't miss stuff um but it made that chart and graph right so if you get the if you get your
351. 57:57prompting dialed in and understand your data set and know what to ask it for and what success looks like uh it can
352. 58:04essentially just do this [ __ ] right so that's formerly code interpreter now Advanced data analytics so will there be a need for data
353. 58:14analysts I think short-term yes long term they'll still be called Data analysts but I think the job will be very very
354. 58:22different right you'll still be asking data for insights but how you actually get those insights done I think a lot of that will be
355. 58:33automated okay [Music] Claude is free but when you're working
356. 58:47on a lawsuit it shuts you down after a while I yeah I know I know um yeah that's a pain I mean the the pain with Claude right now is that
357. 58:59um it's or the the the the pain with the other Alternatives is they can't do large data files although you know what you could use Scott is
358. 59:11um the code interpreter that I just showed the advanced Analytics um you can upload a PDF or you can upload a text file of the big documents
359. 59:22that you're working with in Claude and just work within Advanced analytics because then you can do analysis on it
360. 59:30you could you could pull insights out of it you could do math if you needed to do that you could you could do um
361. 59:38you can do writing analysis like I've done it I've uploaded three or four screenplays at a time and and pulled out
362. 59:44you know similarities and differences between them so um I would revisit gpt4 with code interpreter if I were you um for that
363. 59:56particular use case the Pharma speech went great thank you for asking Matthias Luna stick what's happening welcome
364. 1:00:09we didn't have those we had a puddle all right remember like turning turn in the hose faucet on you didn't have the hose
365. 1:00:17and you just stuck your you like whacked your teeth against the metal thing that had it was like zinc coated steel so
366. 1:00:24like you got the taste of metal in your mouth for for like hours after that we weren't allowed in the house when
367. 1:00:32company is coming oh exactly and then the living room was you could you could you could exist in the family room but
368. 1:00:39don't you ever [ __ ] go in the living room that's where the hand-painted elephant is [Laughter] exactly I live in New Hampshire and
369. 1:00:52winter in Florida oh that's nice so you're okay oh okay cool good at some point I'll be able to pay for
370. 1:00:58Claude but until then I need something besides chat gbt if possible try bard I I would say Bard I like I don't think Bard has good um
371. 1:01:09big context window so you may need to break [ __ ] up with it but but it's pretty decent um and um code llama so if you go to
372. 1:01:21labs.perplexity.ai code llamas apparently quite good so I've tried I tried these two Scott I tried Bard and lab stop
373. 1:01:31perplexity and choose the code llama model that seems like it's doing pretty well um hour north of Tampa and only dealing
374. 1:01:41with downed palm trees a few blocks away Riverbanks flooded that's not too bad Luna stick thank you for joining in the in the
375. 1:01:50aftermath of a hurricane must must get through the water AI learning lab is going live soon must get to Wi-Fi [Laughter]
376. 1:02:04uh uh whatever happened with that Jackal scammer we got rid of him I think thanks to some people in here
377. 1:02:15um Becky Rue told me that she reported him for a copyright violation which I hadn't tried that we were doing the
378. 1:02:24celebrity and and I was doing it as me and other regulars were reporting him as with reporting me as a celebrity which I
379. 1:02:32assume that didn't go anywhere um I'm wondering if Becky Rue's copyright thing is what got it but um anyway he's gone
380. 1:02:43he's been gone for like three days peace um I've been having a blast with ideogram this evening yeah it's crazy isn't it Kyle
381. 1:02:54um thoughts on training with synthetic data versus real data due to cost of training llms so conceptually conceptually synthetic data
382. 1:03:06is absolutely amazing um one of the guys in the AI Salon a guy named Eric Roland he and I have been talking a lot he's doing a project for a
383. 1:03:15bank right now where he's doing something where employees are going to be asking lots of questions and
384. 1:03:22they've sort of the the bank has given him here's our anticipated questions and he said but it was a very limited set so
385. 1:03:29he wanted to really push the boundaries so so he's created a whole set of personas and then he's got those
386. 1:03:35personas you know asking a bunch of questions and then he's doing all this all his testing against that he said
387. 1:03:42it's he's not seeing any problems like one of the one of the things that some folks are talking about is when you
388. 1:03:52use synthetic data there's a thing called Model collapse where as you're training the models on synthetic data at
389. 1:03:57some point they just collapse and they're just not good anymore um he said he's been finding the opposite true that it's getting really
390. 1:04:05really good um so I'm guessing that it's fairly um use case specific um I don't think scientists know enough
391. 1:04:16about how these models work to know what the real impact of synthetic data is um yeah I mean at this point I only have
392. 1:04:27the instincts my instinct is because it's generating new original [ __ ] um that it might be just fine the the
393. 1:04:36handful of times I've used synthetic data it's mind-blowingly good but I'm not doing I'm not doing like real
394. 1:04:42rigorous ongoing training or fine-tuning or anything I'm using it more in an ad hoc kind of way it's [ __ ] brilliant for that
395. 1:04:51any ideas on where I can get an answer my question is pretty basic but Kyle's busy now I've answered your question like four times now
396. 1:04:58that's so Scott one of the problems with this channel is I try to answer all the questions and I just don't get oh yeah
397. 1:05:05Thelonious Scott record patience he'll get to it in about 20 minutes yeah I've already answered it so these three I I
398. 1:05:13think Scott your answer is Bard and labs.perplexity uh um see when you get an old man using new technology everyone's frustrated just
399. 1:05:27click the [ __ ] button Grandpa where's the button I don't see it but this phone doesn't have a button where are the buttons
400. 1:05:36it's a virtual button grandpa just click it click it I don't where's my mouse where's my mouse here it is [Applause] [Laughter]
401. 1:05:53in on a tick tock old man on a tick tock old man on a tick tock yeah the Irregulars I'm irregular oh irregular I like that
402. 1:06:08Irregulars Assemble okay so that's one of the things we need to figure out if we're going to do t-shirts is it Irregulars assemble I'm
403. 1:06:16irregular I I'm a regular good maybe that's a better hat right because people just think you work for some diarrhea company uh
404. 1:06:26you got that at the corporate meeting um I'm a regular a regular assembler Irregulars unite or just irregular
405. 1:06:36it could be irregular misspelled because I don't know that's two on the nose I don't know let's figure that out
406. 1:06:44so we can still share in the show and tell cool yeah exactly so the the the salon Discord is open you can share in
407. 1:06:52any of the channels ask in any of the channels if you find news posted in the channels um that's the whole point of the Discord
408. 1:06:59is that we can continue the conversation in between live sessions or the the salon meetings um why go to a therapist switch at gpt's
409. 1:07:1020 bucks or pie pie's free Pi is very therapy like oh I'm sorry that you're feeling that way Kyle oh I can imagine that must be
410. 1:07:22very upsetting fight does a lot of that stuff AI learning lab sponsored by Metamucil for the Irregulars
411. 1:07:29see that would be a funny t-shirt like I think that see I think that's the T-shirt like hey I learned sponsored by my enemies
412. 1:07:39air learning map for the Irregulars sponsored by metamucili I'd love to get cease and desisted by Metamucil you're taking our our name in
413. 1:07:50vain there AI learning lab uh you know I know you think this might be humorous but uh irregularity is a not a laughing
414. 1:07:57matter if you know what I mean okay yeah yeah yeah that's what our product is so uh no disparaging uh folks that suffer
415. 1:08:04with that condition it's really not funny right you understand you understand what I'm saying here yes Mr remodium [Laughter]
416. 1:08:27if you're asking why am I not talking because Tick Tock jumped me to the bottom of the [ __ ] questions again
417. 1:08:35people that haven't been here before like but Kyle why is that such a big deal it is just this it just is it's very upsetting
418. 1:08:43they don't listen to my they were they weren't Banning my scanner but now they did so I can't [ __ ] about that but they
419. 1:08:50haven't fixed this bug I'm still waiting and I think as we all know it's all about me hello tick tock you're making transcriptions of these
420. 1:09:01things you know what I'm saying in real time because you can boot me in real time so fix your bug yeah oh wait I got a Tom Ronald thing you
421. 1:09:11look a lot like tar Monroe Mr Wolf thank you that's my first d-list celebrity sighting of the night Tom Arnold Rosie O'Donnell's cousin
422. 1:09:20Rosie O'Donnell fat Alec Baldwin um uh um the dude from uh Modern Family uh there's a lot of them Philip Seymour Hoffman occasionally
423. 1:09:34occasionally Nicholson if my hair is particularly well not like it is now White rap about irregular Turtles um [Music]
424. 1:09:50what AI art programs do you recommend for beginners let's go to the video tape let's do it I would go I would actually go to ideogram
425. 1:10:05it's super simple it's free for now it does letters it's not as sexy as the other things you can go there so just go to ideogram
426. 1:10:16mid journey is the best but you have to do it within Discord so you got to learn Discord and mid Journey which is a bit of a pain in the ass
427. 1:10:24but if you want high quality mid Journeys at Leonardo is really cool but but it's fairly sophisticated and fairly
428. 1:10:31complicated if you just want simple there's a little too many you know switches knobs and dials for for just getting started
429. 1:10:38um clip drop is easy because it's a bunch of individual discrete tools and the tools are actually quite simple
430. 1:10:44so that one's pretty good and that one I think is free I'm not I'm not 100 certain um the other thing that you can do
431. 1:10:55is you can go to bing.com so let me let me go to bing.com and show you so you get a Bing so when you get to Bing you're here
432. 1:11:07you're happy you're like oh yeah it's sort of like Google but crappy but then you click on this little chat button
433. 1:11:15and now you have chat GPT you can make money with chbetai and you can put it in creative mode balanced mode and precise mode and those
434. 1:11:26are if if you have ever played in the openai playground those are the those are basically temperature settings which is kind of cool
435. 1:11:35but the one nice thing that um Bing does is I can just say or can I make me an image of a turtle surfing
436. 1:11:52here are some images of turtles surfing from the web oh Siri she gets so jealous when I talk to the other models oops
437. 1:12:06searching for turtle surfing searching for turtle surfing images I'll try to create that here are some
438. 1:12:15images of turtles surfing from the web why is it talking oh because I told it to talk oh is that new I guess it's not new I don't
439. 1:12:24think Bing used to talk like that all right so there it made did it make four yeah so there's four turtles surfing I like turtles foreign
440. 1:12:38Ty and it's image generation and it looks like you can just talk to it now let's try this can you tell me the latest
441. 1:12:53music activity in Denver this weekend searching for music activity Denver this weekend
442. 1:13:08sure here are some music events happening in Denver this weekend till dawn after hours open bar featuring Dead Space dark
443. 1:13:17I I think that talking is new I've like I haven't I haven't been I haven't been playing with Bing recently but I think
444. 1:13:25the fact that you can actually have a conversation with it now Pi can do that and it's connected to the internet but
445. 1:13:30now binkin and I bing might have been able to do that for a while but that's new to me that's pretty cool
446. 1:13:38um Mercy right replet for my own screen shooting yes what air programs we just did that um and hello today to that old-fashioned party Avec
447. 1:13:50plasam France wait Translate old-fashioned party with fish not fresh took us you're funny
448. 1:14:06good night from Europe I don't want to miss the AI Revolution but I need to sleep a bit good night Matthias thank
449. 1:14:12you thanks for hanging out for a bit um we do have people from all over the flipping world come into this thing which is kind of cool
450. 1:14:25what time are it 10 27 10 27 we're doing okay we're doing okay people I got an eight o'clock call tomorrow but it's
451. 1:14:32with a friendly it's with a friendly [Music] is there an AI that can trade crypto or Forex um most of the Forex and crypto sites now
452. 1:14:48have ai Bots and and things like that you can also take the code from those Bots pop them into chat GPT or bard
453. 1:14:57uh or probably Claude and tell it to analyze the bot and maybe you tell it to change some things in the bot and then go run those so
454. 1:15:09um what I would say for that stuff is go to futurepedia.io right there and go into the financial section I can
455. 1:15:21only imagine that there are a pile of AI tools thanks AI videos appreciate it what is that oh it's a little microphone
456. 1:15:28a little Mr microphone remember those you gen xers little Mr microphone I do I remember those that was before they had the karaoke you
457. 1:15:40had to learn your songs in your brain they didn't play the damn lyrics on the TV and by the way who else was the remote
458. 1:15:50control in your house we didn't have remotes Kyle channel four dial channel seven Kyle channel two go to UHF 43. no
459. 1:16:09I needed a button on the back of my head click click click click go click do you have concern about combining AI
460. 1:16:19like chat CPT with Quantum processing like Google's Sycamore from what I understand um they're very very very different kinds of math that
461. 1:16:31are done on quantum computers and um like current GPU um the machine learning stuff with the neural networks and Transformers and all
462. 1:16:41that [ __ ] so I made a video I made a video about Sycamore that said because like the the the um the benchmarks are like staggering
463. 1:16:53um Sycamore did in six seconds what it would take today's fastest supercomputer 49 years to calculate so the the Improvement is dramatic but
464. 1:17:09when I made a video that said imagine what this will do for AI a bunch of math nerds came in and go well it isn't dumb
465. 1:17:16dumb it's a very different kind of mathematics and it's not really appropriate to predict that the best of at the math math
466. 1:17:23so so I don't think they mix but just being a common sensical kind of feller if you got a tool that can do something
467. 1:17:34in six seconds that would take today's fastest supercomputer 49 years yeah they're gonna find ways to do
468. 1:17:41something that's going to augment you know whether it's on the training side or it's on the processing side um
469. 1:17:50but I I would assume listen you know quantum computers they've been adding a qubit what is it a cubit a month
470. 1:17:59and every time they add a qubit it's some exponential insane [ __ ] amount of power increase and Sycamore is what
471. 1:18:0677 or 79 cubits or 70 cubits some it's getting up there so um I don't know I don't know but but you know
472. 1:18:19sitting underneath your question is the is the right question like is there a point at which these things
473. 1:18:27are gonna get [ __ ] radically smarter yes and that's why if you look at the like the the big players in this space put
474. 1:18:40together this group called the frontier the Frontier Model Consortium or the Frontier Model Group whatever it's
475. 1:18:47called they're basically saying if you've got enough money to buy 10 000 gpus and build a large language model
476. 1:18:53you need to join the frontier group and we're going to agree to make sure that we got these things aligned in alignment and
477. 1:19:04if you haven't heard the term alignment what that basically means is are the imperatives of these things aligned
478. 1:19:12with the imperative of these things because if they're not that's really [ __ ] bad like if the imperatives are aligned and
479. 1:19:23these things are like hey let's make sure we take care of these things and increase prosperity for them and make
480. 1:19:29sure they're still around and everyone's healthy and happy and prosperous if they're not aligned and they go well
481. 1:19:35our imperative is to get bigger better and stronger and you know what's the real bottleneck with us getting bigger
482. 1:19:42better and stronger these things and if it decides these things are counter to its imperatives then it will
483. 1:19:50find a way to make these things go bye-bye that's the that's the thing that the that the Sci-Fi people in the in the
484. 1:19:58you know the Jeffrey hintons of the world and the you know the people starting this Frontier group are saying
485. 1:20:03um if if we're going to make these things dramatically powerful let's make sure we got our [ __ ] together because it
486. 1:20:08could go south pretty quick um I think they'll do it I mean there's enough awareness now that that they're
487. 1:20:14they're starting to do it I got to tell you one of the things that I think could really [ __ ] up the works is meta
488. 1:20:20releasing their llama models just out in the open they're just like [ __ ] it here you go have fun there's no safety guard rails on them
489. 1:20:29there the weights are out there so all of the talk of you know these these companies have to be responsible with AI whatever
490. 1:20:40and then meta's like here want one seems a little premature to me but you know I'm just an old fart on tick tock
491. 1:20:54I'm not staying long Vicky no running on two hours sleep from last night this all this all this stuff keeps me
492. 1:21:01awake I know you know what it really does it's I woke up one night at like four in the morning with just an idea like burning in my head
493. 1:21:13I flipped open my laptop I went to chat gbt then I went to um mid-journey made an interface for a piece of software like like in 45
494. 1:21:24minutes out of a dead sleep I woke up and within 45 minutes I'd mocked up a piece of like an interface for software
495. 1:21:31and had like a whole business plan for it this stuff's weird it really is weird that like I don't know like you can just have
496. 1:21:41like a really small idea and then once you get in the habit of knowing that you can get that small idea
497. 1:21:46out of your head in a powerful way it is kind of like crack man it's like it it's it's super addictive in a in a compulsive kind of way
498. 1:21:57which you know for ADHD or ADHD is a neuro spicy people this stuff it really is a drug it's it's it's super compelling but
499. 1:22:09like I rarely get overwhelmed with input I've been overwhelmed for a better part of a year so yeah I get that one
500. 1:22:24write a rap about a regular tone Turtles you look like Tom Arnold uh deep machine learning can it go too far yeah Destiny Troy um
501. 1:22:36like any good thing like any good things in life it's all about balance um sure and and any bad actor right now could take the meta
502. 1:22:50llamas llama models and the the code llama model and they can do some pretty bad [ __ ] with that right now and I'm sure there's
503. 1:22:58Bad actors building that [ __ ] as we speak but there's probably also white hat hackers building their countermeasures to that
504. 1:23:08[ __ ] so I don't know I think that I think that the higher likelihood than you know someone's gonna do some wacky hacking attack with AI is
505. 1:23:23oh there's a doggy doggy wanted in here chippy you want to lay down yeah hey buddy he was locked out Champion was locked
506. 1:23:35out now he's back in um hey buddy let's see um what's it talking about deep learning going too far yeah [ __ ] I lost my train of thought oh
507. 1:23:48it'll come back hey cousin of funnel you're so funny I love your show thank you meta Victory Cosmo funnel
508. 1:23:58I think that is a reference to Rosie O'Donnell cousin o'funnel listen cousin o'funnel you look like Rosie I with this hair
509. 1:24:07tonight this guy this is a Rosy vibe to it doesn't it this is a little more Sears Roebuck 1978 carpet salesman of
510. 1:24:16the year but this has got some Rosy action going on it right now I demonstrated the Kevin McAllister look for my big boss
511. 1:24:24who now knows I haven't closeted AI freak so that's that's actually good cabrino because you know again
512. 1:24:32that idea of at some point and you work in a big Corporation at some point in those big corporations you know someone somewhere is gonna
513. 1:24:42get it in their head that we gotta figure this [ __ ] out who knows anything about this and your boss is gonna go I got I got the woman
514. 1:24:51for you so that's pretty cool I support that catch EPT goes to web pilot on its own for answers sometimes surprise me does
515. 1:25:01it really even if you're not in even if you're not in plug-in mode is that true Ted looks at life which by
516. 1:25:08the way I love I love the name Ted looks at life that's solid um which which one AI does free videos
517. 1:25:19none of them Runway All the video things are pretty pricey um you got Synthesia did do like face animations I think there's
518. 1:25:33a third one that I can't remember right now well there's probably there's probably five or six but those are the ones I know
519. 1:25:39you got Runway ml I I don't see I think the video stuff still is It's gotta It's gotta still be really
520. 1:25:51um GPU intensive to do that because you got to figure to do a video they're doing 30 frames a second so they're generating 30 frames and
521. 1:26:01and I think they're doing it Runway does it in four second chunks but then you can extend it up to 18 seconds
522. 1:26:08so that's a lot of compute power so none of them are cheap but I would just go to futurepedia.io and look in the video
523. 1:26:15section and see see what you can find there because there may be there may be some like generally what happens is like like a a
524. 1:26:25a tool like ideogram comes out and it's it's initially free because they're like hey we're New Kids on the Block come use
525. 1:26:33our tool and if the tool is awesome then peop too many people come and use it and their server Bill shows up the next
526. 1:26:41month and they're like we're gonna go out of business unless we start charging so so you can probably catch some video
527. 1:26:48Tools in those early Wonder Years stages where with where they haven't gotten the server bill yet that might be free but I
528. 1:26:55think they are they're all pretty pricey right now perplexity ai's free tool has essentially been replaced oh it has
529. 1:27:01essentially replaced search engines for me I use it all day every day that's interesting Kevin you know perplexity
530. 1:27:08for me was perplexing because when I heard about it and it's called perplexity.ai I went there kind of
531. 1:27:15thinking it was chechi BT and the first thing I did it like didn't give me what I expected and I and I haven't really
532. 1:27:23like dug into it and learned it I've I've now played with it a bit to at least understand kind of what it is but but um
533. 1:27:31yeah I keep threatening to learn how to use it so it's interesting you're using it all day every day I I gotta I'm gonna
534. 1:27:36dig deeper on it I think I'm thinking perplexity might need a place on the list I'm thinking I'm thinking Poe has to go
535. 1:27:45because I'm always sort of I the PO feels like it's got to go want to do free stock Runner uh I don't know what that means
536. 1:28:02do you have an AI to trade Forex we did that tired of scams on bitcoin AI oh tired of paying for scams on bitcoin AI um
537. 1:28:16neuro spicy work tips that's funny Vicky love thing yeah Bing's funny but Bing seems very polarizing as far as these tools go some
538. 1:28:26people are like Bing's awful other people like I love Bing I want to love Bing but they've got such [ __ ] shitty safety guard rails on it
539. 1:28:39like when it when it when it refuses to answer something for me and I give it a smarmy smart ass answer and then it goes well you're just
540. 1:28:47too angry we're gonna have to just stop this conversation right now little mister I'm like I don't need to be
541. 1:28:54[ __ ] [ __ ] slapped by a search engine it's like shut that [ __ ] off if I'm gonna cuss at a search engine I'm
542. 1:29:03gonna cuss at a search engine foreign what are you gonna do what are you gonna
543. 1:29:18do Google that [ __ ] you are amazing thank you Byron Brad was red seam creative not red line oh thank you Amelia's wife I know that
544. 1:29:30I've been I I've talked to Brad like four or five times today I was actually giving him input on his brand but you
545. 1:29:36know add Red Line red seam you know so red seem creative thank you that boy I am I'm way behind on comments so
546. 1:29:45here's what I'm gonna do there's 80 folks in here they gotta all be pissed off he's not answering any of my
547. 1:29:50questions I'm very upset about this I've come in here I've been in here for at least 10 minutes I've asked my
548. 1:29:56question 17 to 18 times 19 times 20 times and he hasn't answered it I don't understand what's happening I'm
549. 1:30:05just as important as those other people all right so what I'm going to do I'm going to jump to the bottom
550. 1:30:13oh there's so many questions all right and I'm just gonna scroll up a little bit and so if I missed your question
551. 1:30:20sorry but all right so now I'm at uh Ted looks at life in news archive market research biz for 30 years
552. 1:30:29even no bullying search this is Light Years leap yeah exactly um uh remember Boolean search any of you
553. 1:30:39remember bullying search till Google got their [ __ ] together and figured out how to do some version of natural language
554. 1:30:45parentheses plus I mean minus he it's like a math equation although not for nothing but prompting chat GPT and
555. 1:30:56all these large language models is is like programming in English it's like English is the programming
556. 1:31:03language because you're still thinking in logical constructs so yeah it's crazy um I remember when I was the family remote
557. 1:31:13Thelonious channel four exactly best AI I saw this week was Drake trying to diss Eminem and Eve's response was
558. 1:31:22epic don't mess with Gen X nope Gen X don't Gen X is like there's a couple of things we were already bitter and smarmy
559. 1:31:32and sassy and now we're tired but we're still ambitious and we still have got our faculties about us and we can use technology
560. 1:31:42so just don't [ __ ] with us because we're kind of done we're done with it all we're done with people being [ __ ] to
561. 1:31:49one another we're done with all the polarization we're done with all the [ __ ] snowflakes we're done with on both sides we just done
562. 1:31:58so let us be calm down take a chill go to Chachi PT have it write you a song and if you're not gonna do that just get the [ __ ] out
563. 1:32:10it's not quite that bad you can be here if you don't use chat GPT come in here troll me tell me I look like Rosie O'Donnell
564. 1:32:20ask for my autograph oh I'll give it to you I'll do it all right
565. 1:32:36if you could commit one crime that no one would ever be able to commit again what crime would you do a crime that no one would be able to do
566. 1:32:48again oh that's a really interesting question it's got nothing to do with [ __ ] AI it's a good philosophical brain teaser isn't it
567. 1:33:09you know a weird one that I think I would probably do like like fraud Computer Fraud because I think the Computer Fraud with his AI stuff is
568. 1:33:24gonna get really sophisticated and really bad so if I could scam one person and then no one could ever scam again
569. 1:33:36I just picked someone who was douchey I've got like two or three people that I wouldn't mind [ __ ] doing that with
570. 1:33:45like that's one that feels like it just feels like it's going to be this ongoing rolling disaster of just
571. 1:33:52nice little little ladies that well well my daughter called me and she said she was in trouble so I wired her the money
572. 1:34:01like I just feel like that's gonna be like [ __ ] news story after news story of that kind of [ __ ] and like our scammer the scammer that
573. 1:34:12was in here that was scamming me I mean he was reaching out to people to scam them which sucked but yeah
574. 1:34:19yeah that I think it would be that it's not a it's not a sexy crime but it's like it's really ubiquitous I think it's gonna get worse
575. 1:34:27two guys mind extension tools these things are absolutely these are mind amplifiers ability amplifiers is is how I think of these things
576. 1:34:37a lot of things happen from Deep Sleep double helix for example yeah interesting how many programming
577. 1:34:44languages do you know not many I know I know I knew I knew basic back in the day that's what I learned on and then I knew HTML and then
578. 1:34:55I can look at CSS um we created our oh no we didn't create our own XML we use XML for the video assembly for storyvine so XML isn't
579. 1:35:06really a programming language it's more of a scripting framework um I have written a python application but
580. 1:35:14well no I directed chat GPT to write me a python application um I don't know I'm not I'm not really a
581. 1:35:22coder I'm more of a product guy business guy um I'm more interested in what you do with the tools than tactically building them and and you
582. 1:35:35know me holding out because of my ADD I just I can't be an engineer I can't program I don't have the attention span
583. 1:35:41for it or the focus for it I can hyper Focus like if I wanted to take the weekend and learn a programming language
584. 1:35:48I could do that but I've resisted it because I had a feeling [ __ ] like this was coming or maybe I was in denial but now it's
585. 1:35:55here and now you don't have to [ __ ] know programming languages so I know enough about how systems work programs where databases work
586. 1:36:07data hooks work all that sort of [ __ ] that I'm fine I can either get one of my developers to do it or I can have this do it so
587. 1:36:17I don't know much my focus vava blue which is a fun name is kind of November 30th 2022 forward right so the launch of chat GPT for me
588. 1:36:29is a seminal event because that marks the day where before that day it's all Engineers researchers scientists programmers
589. 1:36:39mathematicians after that day it's the rest of us and what I'm interested in is now that these tools are here even though they're
590. 1:36:46in this janky ass state that they're in or as I like to call them janky pieces of [ __ ] even though their janky pieces
591. 1:36:54of [ __ ] oh thanks for the doggo Stevo um we're now all trying to figure out what does this mean for work what does this
592. 1:37:03mean for project Managers from data analysts for creatives for copywriters for Consultants it's gonna [ __ ] up a lot
593. 1:37:15and it's not just gonna [ __ ] it up for programmers right it is going to [ __ ] it up for programmers but it's gonna [ __ ] it up
594. 1:37:22for everyone if I were a consultant right now if I were a consulting company full of data analysts and Consultants
595. 1:37:31that rely on those data analysts I would you know it's funny I would not only be nervous I would be actively trying to reinvent my business
596. 1:37:43because your business model is about to get flushed down the toilet and and I I I don't sense
597. 1:37:57a grand panic and a lot of the people that I talk to um but I I think it'll it'll come that's that's why I'm doing this like I
598. 1:38:09own as many people as possible to go play with these things share bear thank you play with these things have your Kevin
599. 1:38:19McAllister moment freak out and then start to internalize the tools and go okay what can I really do with
600. 1:38:25this how can I transform my business how can I transform my role how can I transform who I am what I do how I contribute to the world
601. 1:38:36so I don't know that got a little philosophical I don't like it when I get all philosophical and serious like that
602. 1:38:42I don't know why you people are doing asking me serious questions why would you do that can't have fun when I'm up on a soapbox
603. 1:38:51dancing around twerking it up futureize them the information overload is real jrc honest to God man I just it's it's so
604. 1:39:06exhausting it's like it's both exciting and exhausting but it's just like the the AI Salon this last Tuesday wait yesterday oh God
605. 1:39:19the AI Salon yesterday like right before it I talked to Leah my co-host and she was like I'm kind of done with
606. 1:39:27AI and I was like I'm just exhausted and then what the salon was was we had five members each talk about a project they were working on and
607. 1:39:39I was assuming people would just come in and like do what I do here like just sort of wax poetic for 10 minutes and be done
608. 1:39:46and they all prepared PowerPoint presentations they had their [ __ ] together um and and like I left that meeting like
609. 1:39:57really inspired so like I'm like oscillating between I'm just [ __ ] exhausted I can't keep up and oh my God
610. 1:40:03this is so exciting I gotta lean in more and just she's like back and forth it's just like a it's like you know like a
611. 1:40:11bad tennis match or a good tennis match Champy's here yes llama you were talking about llama you were talking about go on
612. 1:40:26maybe this would help us use more than 10 of our brains you know cavruno yes so and and it could be the reason that
613. 1:40:38we're all feeling so exhausted I think so so these tools are doing two things for me on that front
614. 1:40:54one is because it's because there's so much less friction in getting ideas out of your head and into some sort of tangible format
615. 1:41:04I'm having more idea like I'm putting more ideas in the world and that's kind of exhausting because each like idea that you put out in the
616. 1:41:12world is some version of your baby and then you like you know have to deal with them and care for them or you know let
617. 1:41:18them drop off and then you're like oh [ __ ] I had those ideas and that's exhausting and the other thing that I think is
618. 1:41:25happening is and this would actually be a really interesting thing for for some Neuroscience people to study is if these things do
619. 1:41:41the low level problem solving like writing the first draft of a document takes a certain amount of your brain
620. 1:41:48energy and a certain amount of your Creative Energy to block that out and it might take you an hour to write a first draft
621. 1:41:55and now with these tools that first draft happens in 20 seconds and then you spend that hour that you
622. 1:42:04would have been writing the first draft refining a thing that was generated and I think it uses a different part of
623. 1:42:12your brain so so there it might be that like historically I would have spent 10 hours in a week doing kind of blocking and tackling creative
624. 1:42:23and 10 hours a week doing refinement and now I'm doing three hours of work doing blocking and tackling creative and 17 hours doing
625. 1:42:34refining and synthesizing and so I don't know there might be you know these things are going to change our brains I think
626. 1:42:47and there's also there's a third one which is like multimodal switching like if I were writing historically I would never think about
627. 1:42:57oh I gotta go make an image for that but now because you can switch so quickly between these things you're like oh
628. 1:43:04here's a cool thing oh give me some headlines oh those are cool headlines oh this should have a logo let me run over
629. 1:43:10and pop out a logo real quick and then you're over in logo mode and then you bring it back and now you're in design
630. 1:43:15mode then you're in typography mode then you're back to writing mode like I think this sort of shape-shifting creatively
631. 1:43:23has got to be just taxing our brain in a very different way right right yeah exactly so yeah no wonder we're [ __ ] tired
632. 1:43:35exactly I need I need a little blue board I need a little blue board I need to take a blank break with a little blue
633. 1:43:43board I'm gonna go out in the yard and cry now yeah yeah it's crazy he likes it hey Mikey those are the best commercials and
634. 1:43:55remember Mikey what was the rumor that he drank um Pepsi and pop rocks and his stomach exploded and it killed him was it wasn't
635. 1:44:04wasn't that the rumor at least in my in my high school or wherever it was [Laughter] um so are prompts similar for
636. 1:44:16mid-journey and ideogram yeah they're very similar I don't think ideogram does any of the dash dash modifiers like
637. 1:44:22aspect ratio and things like that um ideogram's much more simple but things like um you know I want a painting or a
638. 1:44:31photograph here's the subject here's the lighting conditions or Styles here's the medium ideogram also has like you can click on
639. 1:44:40the boxes so photo illustration 3D render those sort of things those are basically the mid-journey modifiers or a
640. 1:44:48handful of them that you can just click on instead of typing stable diffusion question mark I don't know what you mean Molly oh oh when you
641. 1:44:57were asking about um image tools to play with if you're geeky stable diffusion is great um the developer communities all over it
642. 1:45:06if you can run a Google collab notebook and get automatic 11 11 up and running stable diffusion is awesome there's all
643. 1:45:13sorts of different models out there there's all there's Laura's there's just there's a ton for it if you just want to get pictures done
644. 1:45:21it's a complicated [ __ ] nightmare um especially if you're not technical if you're technical go for it it's my
645. 1:45:29challenge with stable diffusion like it took me nine times I failed nine times before I got automatic 11 11 up and
646. 1:45:38running and it for me it was just understanding Google collab and like the whole framework and how it's connected
647. 1:45:44to Google Drive and it was just a lot so I finally cracked that but then my problem was every time I would log back
648. 1:45:52into automatic 1111 The Notebook that I was using had been updated and it would break something and then I would have to
649. 1:45:58troubleshoot it and go fix it and then like 16 new features came out and you had to turn on extensions Within
650. 1:46:05automatic 11 11 and it just it was just like I'm I don't I don't I'm [ __ ] exhausted back to cabrino's point it's
651. 1:46:12like I've only got so many hours of Creative Energy in my brain and if I spend them debugging a [ __ ] Google
652. 1:46:19collab notebook that I don't really understand in the first place um that's like two hours that I could
653. 1:46:26have been making images so so but but it's great if you want to geek out on stuff and have a lot more control
654. 1:46:35um how about the stock Bitcoin you keep asking about that I don't know what that means um bitcoin's not a stock it's a crypto
655. 1:46:44should we truly worry about the singularity if an an attempt to evolve any AI First Step would be to strip any human emotion
656. 1:46:57or any human traits I don't why would that be the first step I I mean I I find the opposite is happening like in theory I think
657. 1:47:12that's rightly that like like if you listen to the Sci-Fi movies the machines are gonna they're gonna strip human emotion the
658. 1:47:21way these things are being programmed if you go play with pi if you go play with this that is not programmed to strip away
659. 1:47:29human emotion at all and Chachi PT quite frankly is more polite and more it displays more empathy than most humans I know it shows nuance
660. 1:47:45it doesn't get worked up so I I don't know that I mean my experience hasn't been what you're talking about there
661. 1:47:54but you know if you're talking about the singularity and these things getting you know exponentially more powerful which they are um
662. 1:48:04I don't know that's why the frontier working group was put together to make sure that they're working on alignment
663. 1:48:09as much as they're working on improving the capabilities um yes you do I don't know what that's in reference to but I'm sure I do turtle
664. 1:48:21Wrangler Turtles totals uh um Tick Tock just jumped me to the bottom again I'm gonna go up a little bit and then I'm just going to pick a
665. 1:48:30question I'm not going to answer all the questions I'm just not going to do it because Tick Tock has got a dumb dumb interface
666. 1:48:36dum dum tick tock what was the Malto meal Maynard tagline um I don't remember
667. 1:48:51I remember Carnation Instant Breakfast but that's not Malta meal not the bottom welcome to AI Boomer talk love
668. 1:49:01this is Elder talk how can I help you today we're going to talk about computers the little round thing sitting next to
669. 1:49:12your computer is what we call a mouse when we say click on the mouse it means press down on the top of the mouse
670. 1:49:22I'm a librarian still use bullying search more power to you bully I mean what was nice about bullying search is like you
671. 1:49:31knew exactly what you were getting and it you know it's funny the problem with natural language search before these things before the large
672. 1:49:41language models if you if you use Boolean search in search engines okay I'll just speak for myself because
673. 1:49:56I use Boolean search and search engines when Google said oh it's okay you can use natural language now we'll figure it
674. 1:50:03out I never trust Google to know what the [ __ ] I'm talking about so I end up sticking Boolean operators within my
675. 1:50:12natural language searches because I don't trust Google and for the most part I'm right but these things now actually do natural
676. 1:50:20language which is just insane when will the Apple tool be released I really hope [ __ ] clawed that it's in um
677. 1:50:33in September when they do their iPhone 15 announcement I'm hoping that there's a one more thing moment where they say oh one more thing
678. 1:50:45and here's our AI tools our AI agent or whatever it is um apple is willing not to be first right
679. 1:50:59they Apple doesn't want to come out here's what Apple's not gonna do Apple's not gonna launch a tool that's at parody with chat GPT
680. 1:51:14Apple's gonna want to launch a tool that redefines the market like like division Pro for me was a significant announcement because
681. 1:51:25I didn't think it would be possible to take us out of ski goggles that look exactly like every other [ __ ] AR and VR headset
682. 1:51:35and successfully separate that from the crowd and they did like in in my opinion they they made an embarrassment out of all
683. 1:51:45the other goggles right they didn't use the word metaverse once they didn't use the word the phrase AI once but they made it very very clear
684. 1:51:56there's significant machine learning horsepower right in front of your eyes you know the R1 chip is sitting right
685. 1:52:04there doing real-time analysis of those 16 [ __ ] cameras and sensors all over your head um that's all AI right so it's clear
686. 1:52:13it's deep AI they successfully created a new category spatial computing um you know they did just weird amazing
687. 1:52:26[ __ ] like put displays on the outside of your eyes so that they could project the images of your eyes forward so people
688. 1:52:35could see that you're there like the minute you see it you're like why isn't everyone else done that well
689. 1:52:42because that's the kind of [ __ ] Apple does right that was nine years of development in that thing so they're brilliant at that so I think
690. 1:52:54they're going to do the same with AI so they're not going to do what Google did which you know when Chachi BT came out
691. 1:53:03and everyone lost their [ __ ] and then Google declared their Red Alert like a month later they had this big hootenanny in Paris
692. 1:53:15and they announced we do AI too we invented this [ __ ] and then it was an embarrassment the tech didn't work the one woman got up to demo
693. 1:53:25something on a mobile phone and they didn't have the mobile phone there and she went where's the phone and people
694. 1:53:31were like we don't know just skip it and she goes oh I okay I guess we're not gonna do the demo like they put stuff
695. 1:53:37out there that wasn't ready to be put out there Apple won't do that so I'm hoping that they announced
696. 1:53:44something in September but I wouldn't be surprised at all if they wait until the developer conference next June so June of 24.
697. 1:53:54because Apple would rather be late and and just of another level then play catch-up and be a parody so I don't know I don't know
698. 1:54:08I'm excited about it I mean the one thing I heard is that they spent their their research budget at Apple is 24.6
699. 1:54:16billion dollars which is bigger than the market cap of chat GPT of open AI um and and Tim Cook said they spent the
700. 1:54:25bulk of that money on generative AI in the past year 24.6 billion dollars and we haven't seen [ __ ] from them
701. 1:54:36so they're cooking up they're cooking up they're cooking up some chitlins and some gravy you're gonna be tasty down in
702. 1:54:44a bayou they got some crawdaddies and put it in there cook it up and they're gonna come out and they're
703. 1:54:53gonna be like would you like some tasty gumbo and everyone's gonna be like what the [ __ ] is gumbo oh my God this is good I hope
704. 1:55:06although I do gotta say that their voice synthesis technology that's in iOS 17 is [ __ ] horrible it's horrible
705. 1:55:17wouldn't have thought it but it is AI will run your washing machine with no soap you on one Kyle I don't know what that means but you
706. 1:55:27were laughing what even is chappy GT exactly what is it what even is internet uh uh you actually do look a little bit
707. 1:55:41like Rosie thank you thank you it's those are my favorite trolls I know it wasn't intended as a troll it's a very pleasant compliment right
708. 1:55:56listen at least I'm still here you forgot Gen X being snarky oh yes snarky smart ass bitter you betcha we're Gen X get off my lawn
709. 1:56:08and you millennials Gen X zero F's left well whatever never mind yeah whatevs uh whatever yeah Molly would you ever
710. 1:56:22do debates on Tick Tock I see a lot of live stream debates about prove God exists well I wouldn't do it about
711. 1:56:28proving God exists I do debates I do debates about education yeah about this [ __ ] I'd be happy to debate yeah that would be fun
712. 1:56:38I don't know it you know I think I don't know the problem with debates is they get a little too serious but at the same time I I just
713. 1:56:54I so much is changing so fast and these things are so profound that I think anything we can do to talk
714. 1:57:06about this [ __ ] is valuable right now so if it's a debate if it's my stupid [ __ ] Tick Tock live idiotness if it's
715. 1:57:14the AI Salon if it's the videos I make if it's Rachel Woods doing the AI exchange I think all of it's valuable
716. 1:57:23like like MacKenzie Bose with GPT boss like he's all over the place now like like GPT boss didn't take off like he
717. 1:57:29wanted it to so he's like all over the place trying to figure out new businesses and new models and services
718. 1:57:34and and it's like I think that's brilliant it's like the transparency there of watching him he's a guy that's so capable and he
719. 1:57:44built this whole tool and and you know the interface and he took all this input and and it was like this is going to
720. 1:57:50full-on be his business and then it didn't have the metrics that he wanted and he's like [ __ ] it it's like I'm gonna do other [ __ ]
721. 1:57:57and it's like I think that's I think that's where we all need to be for the next three years is just you can't not participate in my opinion
722. 1:58:08I think I think not participating in this stuff is career and professional and personal I was gonna say whatever I was it's not good
723. 1:58:24I'm so sick of the word police um I'm so [ __ ] sick of the word police um I I think you got to be in it and at the
724. 1:58:39same time this stuff's changing so fast that it's almost like it's irrelevant what you work on because anything that you work
725. 1:58:46on and build and are proud of today is going to be just obsolete three months from now or three days from now
726. 1:58:53someone's gonna come out with a feature that obliterates entire businesses that were created over the past year and I
727. 1:58:59like that's like I just feel like that's the next three years and if you're at that place like I am right now and it sounds like a lot of
728. 1:59:08people are like this is [ __ ] exhausting it's like yeah I I think it's gonna be exhausting and I and and I
729. 1:59:14still think you got to be in the game and and you know the game is whatever it is for you like it doesn't have to be
730. 1:59:21all in like me like I'm you know cracked out on this stuff right now but I think yeah I don't know we gotta play gotta play to win
731. 1:59:35ocean 11's AI scam versus Casino so what did I do with my AI learning lab under t-shirts you gotta do you gotta do something
732. 1:59:46where the penalty will be half of what you made uh where now you're changing the rules I can't keep up with you
733. 1:59:56all right what can they do in Hollywood to stop the AI that Netflix and others are doing so James they they can't they here's the deal
734. 2:00:13everything every profession that is computer mediated every profession that is computer mediated in any way
735. 2:00:27anyone who uses a computer in any way all of those jobs are going to be transformed writing filmmaking TV video film uh social media
736. 2:00:44knowledge work Consulting project manage all of it all of it all of it is going to be reinvented this [ __ ] isn't getting uninvented
737. 2:00:58[Applause] what the writers and sag I mean this this is my hot take as as an
738. 2:01:12as a trained actor and a screenwriter but I'm not in the wga and I'm not I'm not the Screen Actors Guild
739. 2:01:23I think the most sensible path forward is to say let's assume that all this [ __ ] is going to take over
740. 2:01:32the industry let's assume that every actor is going to be synthesized let's assume that yeah studios are going to
741. 2:01:38use these tools to do the writing and they're going to use tools like the the advanced version of runwayml to do the
742. 2:01:44filmmaking let's assume all of that happens and make sure we get contracts in place so that storytelling professionals still
743. 2:01:55participate in the game but that's not what they're doing what they're doing is they're saying [ __ ] AI ai's evil we should never ever
744. 2:02:06use AI it's exactly what photographer said film photographer said when digital cameras came out every single new technology the existing
745. 2:02:19professional class that is going to be disrupted tries to resist it every single time in all of history here let me show you something
746. 2:02:41from today painting is dead that was French French painter Paul de la rouge upon seeing the first daguerreotype so the first photograph
747. 2:02:52came out and this guy said from today painting is dead painting obviously hasn't died filmmaking nuts is not going to die in
748. 2:03:00the future but it's going to be different so so we like it's an appropriate metaphor we've seen this movie before we know how this ends
749. 2:03:16ai's not going to get uninvented companies like Netflix or it might not even be Netflix see Netflix and and the
750. 2:03:24streaming services may say okay we're gonna we're gonna agree with sag and what the writer's Guild that we're never
751. 2:03:30gonna use this stuff and we're gonna we're gonna keep the old way of doing things and if they actually commit to that and
752. 2:03:36stick to that someone else is going to come along that doesn't know the history and Legacy of that it's going to be some
753. 2:03:4120-something that's sitting in his mother's basement that whips up some AI streaming service that's generating
754. 2:03:49Dynamic content that people all of a sudden love and start using and they're going to cut Netflix and all of these contracts out of the loop
755. 2:03:59because they're fighting a an unfightable battle it it's literally like the town of Tampa three days ago going we're not going to
756. 2:04:11accept that hurricane nope nope we're striking we're striking we want to in our contract that uh we no
757. 2:04:20longer accept hurricanes and hurricanes are bad and we're not going to accept them like that's what I feel like it's like right now so
758. 2:04:30I don't have a good answer for it but I I just think that the AI is evil [ __ ] AI we're never going to use AI I just think
759. 2:04:37it's incredibly short-sighted it's I'm a writer I've written seven screenplays I've written a one-man show these
760. 2:04:47are the most remarkable writing tools in the history of humanity and the writer's Guild the group of professional writers
761. 2:04:55is saying nope we don't want to use the most professional writing tool ever written I don't it doesn't make sense to me now
762. 2:05:03what does make sense to me is holy [ __ ] these things right good enough that if they get much better what we currently do what we currently
763. 2:05:12call writing is at risk that I absolutely agree with but you're not going to uninvent it not gonna uninvent it
764. 2:05:22is it sad is it tragic is it all [ __ ] up yeah but you know they're going to get blindsided five years from now by some company that
765. 2:05:35doesn't think like they think because it's going to be people that aren't from that world it's literally from another world
766. 2:05:44like think about you know the kids that are in ninth grade right now they make it through and especially the
767. 2:05:54kids in ninth grade right now in schools that are embracing AI so they've got three four years of high school
768. 2:06:05and then they got maybe four years of college but that might not even be a requirement there they got three four
769. 2:06:11years of high school if you think about regenerative AI was a year ago and where we are today and then you go out three or four years
770. 2:06:19these tools are going to be [ __ ] remarkable so you're gonna have people coming out of high school that have quasi-native AI skills
771. 2:06:32that don't know the history of anything they don't know they don't know the world all they know is they got these
772. 2:06:37tools oh cool I can make a movie oh yeah this is you know Grandpa watches Netflix all the time this is kind of like
773. 2:06:44Netflix but we're making original content for everyone everyone gets their own personal movie that's kind of cool Click Boom launch
774. 2:06:51billion dollar valuation Netflix is scrambling to keep up and that becomes the new Netflix like that's gonna happen
775. 2:07:02so how does the writer's Guild account for that they can't and neither can sag you know what they could account for is
776. 2:07:11okay assuming that that's the future what kinds of contracts do we need to put in place now that allow us to participate in that future
777. 2:07:25it's a weird one it's a really weird one but yeah they do be janky but a lovable Yankee yeah had an argument with my high
778. 2:07:34school soccer no social studies teachers Hub hubby teach kids to use AI critically or forbid it uh you know what channel you're in [ __ ]
779. 2:07:48teach them to use it critically critical thinking so I'm a librarian you know this critical thinking is the skill
780. 2:07:56that's required for these tools forbidding it is just like the whole just rant I just went on about sagging
781. 2:08:05and the wga right now the education system is worse forbid it forbid a thing that is going to be in everything okay
782. 2:08:18aren't aren't you there to prepare students for the world they're about to enter how can you do that if you don't
783. 2:08:25understand what are the tools that are going to transform that world this is dumb it's just dumb keep getting
784. 2:08:35in those arguments I fully support you being in those arguments do we need project management I think we're going to need
785. 2:08:42I think project management yes I think project management like that's one of those plumber roles where where it's just going to be like
786. 2:08:53everything's [ __ ] changing how late is it it's late isn't it 11 22. dang time flies when you're in the AI
787. 2:09:01learning lab it flies it flies right by I don't know what you people are doing here how many people 74 people still
788. 2:09:08here folks it's Wednesday night you have work tomorrow my hair is holding up okay tonight I don't know what happened although this
789. 2:09:17this situation we got a little we got a little Rosy situation going on over here but that's hanging out all right
790. 2:09:24uh what's the AI the AI Salon is a group that I started last December you can go to the salon.ai that's a link tree we meet every two weeks
791. 2:09:35um in person here in Denver and then we have an online thing so it doesn't matter where in the world you are we
792. 2:09:41have people from all over the world in fact last night sander who I met in here spoke at the AI Salon from the
793. 2:09:48Netherlands it was like two in the morning his time when he was talking about his new company that was kind of cool
794. 2:09:54um and we have an online Discord server we have a LinkedIn group and it's basically just people that are curious about AI on the
795. 2:10:02adventure trying to figure this [ __ ] out together and uh it's a it's a really neat group
796. 2:10:17yes our brains will have an unprecedented opportunity to rewire right yeah like like this is potentially
797. 2:10:26evolutionary kind of stuff like if we if we just fundamentally shift the parts of our brains we're using actively because of these tools
798. 2:10:36it's wild flipping wild it's exhausting it is it is it's incredible exactly I'm more excited than I've ever been I'm more terrified
799. 2:10:47than I've ever been and I'm now more exhausted than I've ever been as a direct result of this [ __ ] and then you're asking but Kyle are you
800. 2:10:58gonna get so tired you're gonna stop [ __ ] no do you know how lucky we are to be able to play with this [ __ ] you think I want
801. 2:11:06to miss any of this yep it was pop rocks that killed Mikey I remember that what history is AI being taught whose
802. 2:11:23perspective so they were trained on all of the internet or all of the public internet so what's in there is everything
803. 2:11:35but it was everything that was on the internet so whatever wasn't on the internet didn't get in there so there's a lot of that
804. 2:11:43and then within it there's garbage history and real history and pseudo-garbage history and pseudo Real History it's all in there
805. 2:11:53how you fine tune it and how you pull the words out of the large language model um has largely to do with the fine
806. 2:12:01tuning so you could take the same base large language model and fine tune it in different ways to get it to behave in
807. 2:12:07different ways if the foundational information is flawed then you can't fix that but they're aware of that they're working on it so
808. 2:12:17but but it's so so John one of the one of the things about the way these tools work is it's not copying and pasting like these
809. 2:12:31things do not have an agenda these things are calculators they're they're literally calculators that as you type
810. 2:12:39your prompt it's creating a mathematical probability of some single word in latent space in Vector space which is a mathematical
811. 2:12:52thousand dimensional space that has words clustered by semantic meaning as you type your thing it makes a probability that some some
812. 2:13:03Vector in that space is the it has the highest probability of being the next right word it doesn't even know what the
813. 2:13:10word is it just knows its Vector string and that Vector string is associated with a word so when you type your prompt
814. 2:13:16and you hit submit the first word that it puts there was a mathematical formula so is the second one so is the third one
815. 2:13:24so is the fourth one it doesn't know what it's writing so they're not they're not inherently biased other than
816. 2:13:36if there are things omitted from the data set then they're not in there and if how the things were fine-tuned so
817. 2:13:45the way Chachi PT fine tunes their stuff is they put in a prompt it generates something and then human beings go thumbs up or thumbs down
818. 2:13:56that introduces human bias so depending on who does the human reinforcement learning here that could make them more biased Claude
819. 2:14:06that's from a company called anthropic they use a constitutional learning model where it's not humans doing this it's a
820. 2:14:12set of rules it's a set of imperatives so Claude has potentially a higher likelihood of being less biased but here's the here's the
821. 2:14:21the fact of the matter about bias is bias from whose point of view the thing that I think is perfectly unbiased you may think is completely
822. 2:14:33biased and vice versa and then the third person is going to have their own flavor of this so I think in the future we'll
823. 2:14:39have different models that we can choose from I want one that looks and sounds like me because because I'm a petty [ __ ] and I
824. 2:14:48want Petty [ __ ] in my large language models and then other people be like I want tough guys so I don't know I don't know
825. 2:15:01I do not know I read that on average we consume more data per day than one would in a lifetime in the 17 and 1800s consume
826. 2:15:11who uh Peter Kaminsky last night at the salon talked about a um an image organizing tool system that he created
827. 2:15:24because he's generating so much generative AI art that he wants to publish in the world like his backlog is
828. 2:15:31like 13 000 images that he's created in the past three weeks yeah it's not only consuming it it's generating it we're generating we're
829. 2:15:41gonna like oh my just the amount of crap that we're gonna generate it's [ __ ] nuts that's tough it's tougher if you're stupid
830. 2:15:52all right see just like a just like a freaking Pavlov's dog all right it just jumped me to the bottom I think that's a sign I gotta go
831. 2:16:00I'll answer a few more questions they were not trained on everything available on the internet that's very
832. 2:16:05incorrect well yeah they were trained on whatever opening I could get their hands on and tag and do all that but
833. 2:16:15ostensibly a lot of the public internet and they want to get more right so it's all about the data and and the new um
834. 2:16:24open AI bot that allows you to opt out a bunch of the big publishing companies are opting out of that so open AI may
835. 2:16:32have all of these companies may have a tough time training the new models if the big publishing companies are like [ __ ] you
836. 2:16:41and and you know Google's probably going to start cutting deals with companies that AI that openai can't compete with
837. 2:16:48so we're we're starting to the next two years are going to be some corporate Cutthroat Malarkey as Biden would say
838. 2:16:58um so so basically data analysis and sorting uh yes I think so is the internet left-leaning yeah um perfect example of how Neo became the
839. 2:17:13one on the Matrix they uploaded loaded data to him suggestion can we do more lab examples with these tools and less Chit Chat yeah
840. 2:17:22we can possibly do that although people aren't asking for that but yeah like I would say ask specifically for what you want and
841. 2:17:32then I'll do that because I try to do as many questions as I can my Chit Chat is answering questions if
842. 2:17:40the Chit Chat said hey can you show me this tool or that tool then I'd be showing this tool or that tool so sure
843. 2:17:47you are determining what I chitchat alignment make them biased well yeah hopefully when we want them
844. 2:18:03biased in the direction of human prosperity I made a chatty bot site the date of her resources May of 2023 very cool I like chit chat
845. 2:18:17uh once it can reinvent the wheel read patents we are going places it can do that right now in fact I think there's some patent um
846. 2:18:28there's some patent plugins for chat CPT and I would imagine at futurepedia.io there's probably some intellectual
847. 2:18:36property stuff going around all right listen I gotta go people it's 11 30 here in Denver I don't know what time it is
848. 2:18:43there if you're on the East Coast what are you doing with your lives why are you still here I I love y'all I appreciate y'all very
849. 2:18:51very much this is fun I have a blast doing these things um thank you for hanging out uh Share Bear Pate welcome again um
850. 2:19:04multi-dimensional man Emilio's wife jrc Dr J I was in a paddock good night talk to you later nighty night Tobias good night
851. 2:19:17Kyle great job tonight thank you very much thank you I'm a librarian digital Gods thank you so much I see share bear peace out I see River I
852. 2:19:28see paint Pace Pace Pace jrc thank you everybody this was fun I'll talk to you tomorrow bye-bye