aiLL 6/10/23 - The A.I. Tsunami - What's Coming & How You Can Ride The Wave

Transcript

1. 0:00foreign
2. 0:18with the AI learning Labs getting set up here welcome welcome welcome happy Saturday is it Saturday I think it's Saturday
3. 0:30yeah it's Saturday Saturday I can answer your questions hey Emilio's wife what's happening um I can answer your questions about AI
4. 0:39I can't answer your ask questions about what day it is apparently I can't talk either I need to get some water uh
5. 0:48should I just walk away while people are coming in I probably shouldn't but I do need some water [Laughter] all right um
6. 1:00let's see my name is Kyle Shannon this is the AI learning lab uh I'm here to wax poetic about generative AI get water
7. 1:09yeah I'm gonna go get water came for the hairdo come for the hair stay for the jokes about how I look like
8. 1:15Rosie O'Donnell I'll be right back
9. 2:16all right I'm back I'm back I'm back you've been waiting so long for me to return [Music] saying oh the Rosie O'Donnell joke is so
10. 2:30my trolls are creative at least so so in the past week I was told I looked like uh Rosie O'Donnell um oh there's a lot of people here let
11. 2:41me tell you who I am my name is Kyle Shannon this is the AI learning lab I'm a long time entrepreneur um I talk about
12. 2:49what I'm learning in AI in generative AI so I'm relatively new to it I've been deep diving for about a year but I started a um
13. 2:58AI Salon I have a newsletter called everyday AI I've got this Tick Tock Channel and I'm rolling AI into my business which is uh it's called
14. 3:06storyvine it's an automated video platform and I'm completely reimagining the product because of generative AI if
15. 3:13you've got questions about AI pop them in the comments and then you can watch as I comically try to keep up with them but
16. 3:22don't and then new people are like why won't he answer my question I put a question in 17 times and it didn't answer it
17. 3:30and then as I'm scrolling through the questions there's this question 17 times because the guy that you know it's usually a guy
18. 3:38um what else I was a I I did some pioneering stuff back in the early days of the web I started one of the first online magazines and one of
19. 3:49the early digital agencies called agency.com and um why I mentioned that is the parallels between what was happening in the early
20. 4:00days of the web and what's happening with this when in the early days of generative AI are like fascinating and
21. 4:06amazing and and it feels like a very similar um major shift that's happening that we're at the very beginning of
22. 4:15um uh so so back in the early days of the web the internet had been around for a bunch of decades but it was mostly
23. 4:21used by scientists and researchers and universities and then the World Wide Web came along and all of a sudden us
24. 4:28regular people got to play with it and was like oh that's kind of interesting and similarly with AI it's been around
25. 4:35for a bunch of decades and um Google invents this thing called the transformer in 2017 and you know it was a lot of years ago and
26. 4:46it's it's a scientific paper called attention is all you need and that ends up becoming the pivotal technology that
27. 4:52is that lays the foundation for mid-journey and stable diffusion and church EPT and all of this super sexy stuff we get to play with
28. 5:01um and so it's a very similar thing and and one of the differences is the speed of adoption so um uh the World Wide Web took about six
29. 5:11years to get from zero to 100 million users Chachi BT six weeks so anyway all right let me go answer some questions
30. 5:20stop babbling Kyle answer my question I've got questions I've got questions all right let's see what your questions
31. 5:28are best AI app the reason I put these URLs up on screen in my in my fancy little double camera window is that I think right now
32. 5:41and probably for a lot of years just start learning chat GPT the the depth of its capability is insane so
33. 5:50chat.openai.com is the official chatgpt website you don't need an app most of the apps out there are fakers there's
34. 5:58one official chat GPT app on iOS there's not one for Android yet so don't download any of that [ __ ] just point your browser at that
35. 6:06I don't get paid for this like I should hey chat EBT if you're listening hook me up baby you know what give me code interpreter
36. 6:15that would be payment enough okay and I'll talk about that in a second so so that's that bing.com is gpt4 but for
37. 6:24free it's Microsoft's chat bot that uses open ai's gpt4 and it's connected to the internet so if you don't want to pay 20
38. 6:32bucks a month to open AI to get their version of chat jbt4 then you use bing.com and then poe.com gives you access to a bunch of different
39. 6:41models and I did something really cool today so one of the models you have access to that uh chatbot just like chat EPT
40. 6:50except it's got all these different models so Sage is pose model then you got gpt4 Cloud plus Cloud instant 100
41. 6:58000 I'm going to show you that in a second you got Claude instant which is ripping fast and you got chat gbt these
42. 7:04three that say subscribe you have to pay for but this one in particular well I don't don't look at that yet okay we're
43. 7:11gonna clear this context so here's what happens with um with a hundred thousand okay here's what that means the 100K in this case means a
44. 7:25hundred thousand tokens what's that mean well with chat GPT you have 4 000 tokens which for whatever reason they they call them tokens and
45. 7:35instead of words so four thousand tokens translates into about three thousand words it's a geek thing the Geeks
46. 7:42decided to call them tokens because technically that's what's going on but realistically it's three thousand words
47. 7:46so a hundred thousand tokens means seventy five thousand words so with chat gbt if you have a really long document
48. 7:54and you like paste it in there chat gbt goes like yeah I can't do it I don't wanna it just can't deal with it but a
49. 8:02hundred thousand you could take say a novel or a screenplay and pop it in there so what I did earlier was I went to
50. 8:10a site here's the script of The Godfather so I'm going to copy that and then I can just go in here into into this little chat window this is
51. 8:24mind-blowing to me all right um based on this screenplay and then I'm going to hit shift return to make some line breaks I'm going to paste
52. 8:38and what I pasted in was all of The Godfather right so this is Mario puzo's screenplay so pages and pages and pages right
53. 8:49um chat EVT would absolutely choke on this wait okay now I'm at the end now based on this okay um please create a table of the main
54. 9:03characters in markdown and provide uh or n and tell me what happens
55. 9:20to them and on I don't know if this has got Pages let's see what happens to them and how far into the story
56. 9:36uh let's let's see what that does so it just read all of The Godfather and now it's thinking those three little
57. 9:45dots waiting waiting wait for it calm down people still waiting still waiting room
58. 10:00and it is building us a table this is [ __ ] insane it's just insane all right let's see Don Vito Corleone
59. 10:10was shot by celozo but survives Michael eventually takes over the family Don Vito passes away later in the story
60. 10:16Michael Corley Corleone initial State initially stays out of the business but he's dragged into it after Sunny is
61. 10:22killed Sonny he's killed in an orchestrated hit right so you can take incredibly long documents and just start interacting
62. 10:32with it and the other cool thing about this is um what Claude does here is it kind of senses that these are main characters so
63. 10:40oh you can't really see it these are different colors um like uh Michael Corleone is a hyperlink and if I click on that
64. 10:50it says tell me more about Michael Corleone so it writes a prompt for me and now it's gonna go tell me about Michael Corleone
65. 10:59wild huh all right that's that so the question was what are the Cool Tools so so you could just start with chat gbt
66. 11:10or Poe 100K or uh what is it Claude 100k from anthropic you could probably spend the next decade figuring out amazing [ __ ] to do with it
67. 11:21in terms of images um mid-journey is the best in my opinion if you like geeking out then stable diffusion lets you do like the developer
68. 11:32Community is just you know nuts unstable diffusion but you got to geek out a little bit to use it those are the two image things
69. 11:39um I just did my first Runway ml Gen 2 if you haven't heard of that so that was my first attempt I said I described it as uh
70. 11:56what I don't know woman in a flowing white dress walking through a field of flowers with a gentle breeze and a cinematic tracking shot
71. 12:05and that's what it generated in about I don't know 30 seconds so that's pretty cool to play with um yeah I don't know there's lots
72. 12:16there's lots start with chat GPT start there and and then uh then all will be revealed to you all right what's the best AI that's already
73. 12:25trained you can install on a server honest to God there are so many there are so many open source large language
74. 12:34models they're being released like every three days and like each one's topping the one before it um there's a thing called GPT for all
75. 12:43and I think what it does is it it it installs a bunch of [ __ ] on your server and gives you access to multiple
76. 12:49different models so I think a project like that where someone else is keeping track of all those large language models
77. 12:57is probably the way to go so so look up GPT for all do you do Twitter spaces I have not yet but I probably could and should I I
78. 13:06don't know I I like being able to see [ __ ] and show [ __ ] because but that's probably not a bad idea I don't know I
79. 13:12might also put together a um what do you call it a workshop to actually teach people how to do stuff but I don't know
80. 13:20I got I gotta I got a company to run people I can't just be sitting here on Tick Tock being a goddamn influencer I
81. 13:27mean look at me I'm I'm like a handsome Rosie O'Donnell foreign oh God all right thanks Kyle I'm using gpt4 but
82. 13:41it's not connected to real-time data yes it is what am I missing okay let's go turn this [ __ ] on um what opening I decided to do was
83. 13:53make it so convoluted and complex to get to that you just can't figure it out okay so you're at chat gbt if you've
84. 14:04paid for your 20 bucks a month and have chat GPT you'll see jet gpt4 there but what you probably haven't done you have
85. 14:12to go to your settings so go to settings and then within your settings the second choice down here is
86. 14:18beta features turn on the beta features and you have to turn on you have to explicitly turn on browse with Bing and
87. 14:25plugins once you do that now when you roll over gpt4 you've got three different options you've got the default
88. 14:33option which is not connected to the internet and frustrating as [ __ ] because all the data ends at September 2021 then
89. 14:40you have browse with Bing so this one is who won the Nuggets game last night and if you've been watching my lives you're
90. 14:51probably thinking come up with another example Kyle that's dumb you keep saying the nuggets but it knows it doesn't know that
91. 15:01so it goes to browse the web and it rewrote the prompt for me and it says the Denver Nuggets one in in the
92. 15:10game 108 to 95 they held Miami to less than a hundred I'm in Denver by the way don't give two shits about basketball but hey I'm I'm a
93. 15:21fair weather fan where you're in the finals I'll root for you and I don't care about Miami so yeah
94. 15:37come on calm down calm down gonna be okay lay down lay down all right hey
95. 15:51all right so thank you 25 years ago I had some really poorly behaved dogs and we got a dog trainer and what I learned when we got a dog
96. 16:03trainer was that they don't train the dog they train the humans and you have to do [ __ ] like don't let
97. 16:09them jump on the furniture and you know it's basically on you for for having a dog that's not poorly behaved and so I
98. 16:16made a decision about 25 years ago oh I'm just gonna have shitty behaved dogs all my life because I don't have
99. 16:22boundaries certainly when it comes to pets all right oh no Tick Tock you just jumped me down this is this is one of the uh when I
100. 16:30when I do my lives most of my time is spent trying to find myself in the comments um oh it looks like there's some new
101. 16:35people if you're new here my name is Kyle Shannon this is the AI learning lab I'll answer your questions about
102. 16:40generative AI I'll make jokes I cuss um uh what else do I do um I don't know these URLs right here are just there
103. 16:52kind of as a public service announcement the top one was chat gbt the next one down is Bing which is gpt4 that you can
104. 16:58use for free and Poe gives you access to multiple models that you can compare so that's super cool and fun okay
105. 17:09how do you think AI will change research fundamentally fundamentally it will change it I think forever because it's kind of
106. 17:18instant like like the thing I just did I don't know if you're here a minute ago where I used um po the Claude hundred thousand
107. 17:28um there and then it automatically created a table of all the characters and what happens to them in the
108. 17:35screenplay like like you can read and summarize a screenplay so so think about there are a lot of human beings on a
109. 17:44planet whose job it is to read screenplays and and and create what's called coverage for a screenplay so they hand it to the execs you could
110. 17:52absolutely automate that so a screenplay comes in you run it through this thing it creates it creates automatic coverage
111. 17:57it will probably even score it based on quality uniqueness what genre it's in things like that um instantly see you know same thing for
112. 18:07research and now that um excuse me now that gpt4 is connected to the internet you can basically just point it at big data sets
113. 18:18and then you've got um this new thing coming out called code interpreter which I don't have access to
114. 18:25and I'm very bitter about that but it's a new mode of chat GPT and you can upload giant data sets to it and it does
115. 18:34full-on data analysis including creating um you know graphs and charts and things like that it's amazing so it's going to
116. 18:43fundamentally alter it so like here's the thing a lot of the work that we do in today's world is doing
117. 18:52kind of low value processing right so so if I read screenplays for a living like I have to take a number of hours to just
118. 19:00basically get to understand what the story is about who the characters are what the main arcs are well if I can do that in three minutes
119. 19:08or five minutes or if I can automate it to just do it when a screenplay comes in it generates that report
120. 19:13now I can use my brain for higher level stuff so now I might rather than considering you know 10 screen plays I
121. 19:21might consider a hundred screenplays and I might really sort of mix and match them and compare compare them in a different way
122. 19:28so it's not like research goes away or you know screenwriter coverage goes away it's that the job completely changes and
123. 19:36I think that's what what all of us are going to face in in all knowledge work all creative professions
124. 19:43it is not that you know the AI is coming to take our jobs it's that as human beings learn to use these incredibly
125. 19:50powerful tools the jobs will change and then people that so Paul rateser who runs the marketing AI Institute says
126. 19:58um AI will not take your job but someone using AI will right so so the purpose of this channel is to encourage you to like
127. 20:07go there start playing get curious about this stuff if you do a particular job if you're a researcher start figuring out
128. 20:14how these tools can augment your work and then that will serve you incredibly well sit still no I have add this is how it is
129. 20:23you're not babbling that's what ADHD meds do I'm not on ADHD meds mid-journey versus stable diffusion mid-journey mid Journey mid Journey
130. 20:34however stable diffusion has a lot more development going into it because it's all open sourced so if you want to geek
131. 20:41out like if you're technical and you don't mind going to GitHub and installing things on Google collab and
132. 20:48you know virtual machines and [ __ ] like that stable diffusion is amazing there's a lot of tools there I still don't think
133. 20:54the quality of stable diffusion is as good as mid Journey both of them however have been trained on copywritten images
134. 21:05so using mid-journey and stable diffusion for commercial work especially if like you're an agency doing it on
135. 21:12behalf of a client there's a risk that at some point someone's going to get sued for using that by one of the
136. 21:19artists whose data was trained so Adobe their Firefly products while it's not anywhere near as good from a just
137. 21:28raw quality standpoint as those other ones it's only been trained on open source public domain and Adobe stock
138. 21:36images that they own all the rights to and they actually two days ago announced Adobe for Adobe Firefly for Enterprise so fireflies
139. 21:47their generative AI product so they're now making it safe for companies to use that so if you're in a commercial
140. 21:52environment I'd use Adobe Firefly just because it's it's safe and that's going to be their intent they're allowing
141. 21:59artists to opt out opt-in they're going to be doing attribution and compensation so how different is three to four
142. 22:10um for for GPT the difference between 3.5 and 4 3.5 is much faster so if you need speed um 3.5 is like wicked fast
143. 22:20if you need Nuance in language gpt4 is staggeringly good it's just it's it's just way better apparently it codes
144. 22:30better as well if you're doing coding apparently gp24 is much better at coding but gpt3 is not bad like my um
145. 22:39my big Epiphany with chat GPT was riding a python application in like 90 minutes I've never written a line of python in
146. 22:46my life and I I created a functional very sophisticated application in 90 minutes with GPT 3.5 so it's good to pay
147. 22:56the 20 bucks have you used the new GPT plugins yeah I have um they're good they're they're so let me let me show you how to do the plugins
148. 23:06there so so basically the way it works is you you have to pay for GPT Plus you have to enable the beta features in
149. 23:16your settings and then you have access to browse with Bing and plug-in so if you go to plugins you now have this drop
150. 23:22down menu and and here are the plugins I've got installed so you can install as many as you want but you can only
151. 23:29activate up to three per session so the way I've kind of thought of it is you you basically group together related
152. 23:37plug-ins it might be one or two or three and what you're doing is creating kind of a custom application so let me see
153. 23:44okay I did something last night so I'm gonna I'm gonna turn on experiences and then I'm gonna turn on get your guide
154. 23:55and I'm gonna turn off ask your PDF and I'm going to turn off link reader and I'm going to turn on trip.com Okay so
155. 24:03I've got three wait no I don't now I have three okay I have three travel related plugins installed right
156. 24:14okay so now I can go down here and say um um I'm going to be in um Florida
157. 24:28in the Tampa area for five days in September um make me a detailed
158. 24:46um itinerary of day and evening activities I like to fish listen to music
159. 25:02and eat like a pig all right so now it's thinking and it should yeah
160. 25:16so it says using get your guide so it fires off one of the plugins and then each of these plugins you can open it and see what it did so it
161. 25:25said Tampa Florida fishing tours so it pulled fishing out of there so so it got some results and then it did
162. 25:35Tampa Florida music activities cool and then it did Tampa Florida food and Gourmet tours
163. 25:51and so it took all of those responses and it said here's your detailed trip you know start with a private in Source
164. 26:00inshore fishing in Tampa Bay book here it says let's click on book here sometimes it will give you hyperlinks that are broken
165. 26:08no that's not broken so I can go book that trip right so I didn't have to Google finding some dude or dudette that
166. 26:16you know takes you out on on fishing trips um afternoon after a morning of fishing enjoy a leisurely lunch at a local
167. 26:24seafood restaurant and that I can tell it to book me that um day two day three day four day five so and there's the you know little links it
168. 26:33has so that's plugins for you so so like what I've just created in in whatever that was a minute or two was a little
169. 26:42specialized travel application and and this is I think this is what we're gonna start seeing is that is
170. 26:49that really specialized and personalized applications you'll be able to generate them for yourselves kind of kind of in
171. 26:55real real time and I've even heard that um kind of the next iteration of plugins is you won't have to manually turn them on
172. 27:04like I just did like it'll just figure that [ __ ] out it'll it'll it'll basically know the entirety of the
173. 27:09plug-in universe and when you ask it to do something if it finds a plug-in that can do it better than its large language model it'll use it
174. 27:17what book do I recommend for a good foundation on AI if you want to learn about the underpinnings the machine learning and
175. 27:25how all this stuff works um I would probably go to MIT or Stanford um like online courses and take an online course because they're going to
176. 27:35be the most up to date um if you if you're just talking about this generative AI stuff there's not going to be a good book on it right now
177. 27:42it's too new it's too it's too early so just get curious go there and start playing start digging in start watching
178. 27:50YouTube videos start watching Tick Tock videos of people that you um trust Rachel Woods she's got a thing
179. 27:59called the AI exchange and she does tick tock she's really good um there's a guy named Mackenzie Bose who's got an app called GPT boss which
180. 28:08is a kind of overlay on top of gpt4 it's pretty slick he's got a good Tick Tock Channel if you want to geek out there's
181. 28:15a guy on YouTube named David Shapiro who's who's building autonomous agents and he's got a whole open source project
182. 28:22doing that he's [ __ ] Wicked smart when I want to feel humbled and stupid I watch David Shapiro I'm like oh I could do this and then I
183. 28:31watch him I'm like I can't do this did you not have a degree in acting yeah
184. 28:46it's processor measurement okay I don't know what that means what book would you recommend holy cow I was looking for something like this
185. 28:54like this hi nice to meet you Victor all right I have paid version and wrote basic code to help automate time consuming part of my job yes
186. 29:04do that here is my feeling there is a very real possibility there was a paper put out by Mark and Drayson
187. 29:14who created Netscape he was one of the early pioneers of the internet of the interwebs he just wrote a really long paper on how
188. 29:22he thinks that um generative AI or AI is going to be the the salvation of humanity not the just the the death of it and um
189. 29:32things like that things like using AI to automate the shitty parts of your job imagine a future where all of the jobs
190. 29:41that are the soul-crushing you're a cog in a wheel and you can't believe you [ __ ] do this for a living and you live for
191. 29:48Friday and you live for the weekend and God damn you hate your job what if all those jobs were automated out
192. 29:55and you got to choose what you wanted to do that's a possibility of a future that we're living into and so the sooner that
193. 30:03you learn what these tools can do and start to do things like what's what's your name um Superfluous honey awesome Name by the way um
194. 30:13if you start doing what she's doing um okay here's the downside everything's gonna [ __ ] change everything's gonna be disrupted if you
195. 30:27think you're immune from this you're not the upside of it is every single one of us has an opportunity to reinvent
196. 30:35ourselves in dramatically positive ways if we choose to do it but you gotta step in and do it you got to get curious you
197. 30:43gotta you gotta be willing to go to the uncomfortable place of oh [ __ ] I gotta learn something new really yep
198. 30:50because if you don't someone else will and it's not necessarily just going to be the young kids that know how to use
199. 30:55the Internet it's going to be old [ __ ] like me gen xers that are like this is [ __ ] cool man reinvent themselves and
200. 31:02all of a sudden have super powers again that's coming think we have synergies cool connect with me my my contact info is in my bio
201. 31:11and you can DM me and I've got contact info up there and you can join the AI Salon that I do there are so many opportunities
202. 31:19based on tokens you said on this platform to download this thank you yes I don't quite know what I said there and
203. 31:26I don't quite know what you mean but awesome amazing thank you what book do you recommend I don't for good
204. 31:32foundation on AI uh I I don't know of any I'm I have ADD I can't read [Laughter] all right kind of concerning too yeah so
205. 31:43here's here's my here's my thoughts on all this AI stuff I am more excited that I've been in decades and I'm more
206. 31:51terrified than I've been in decades and that feels like the appropriate balance to me that if you're too excited about
207. 31:58it or if you're too scared about it both of those are bad um you should have a healthy balance like life is about balance
208. 32:06total excitement and joy and shocking horror that's what AI is right now for me and I think that's the appropriate thing all
209. 32:17right uh Duke dude stop asking questions um I love these lives thank you llms need a lot of Graphics memory won't work
210. 32:26on my laptop oh that's changing real quick um just go just go to Twitter and look up open source llm there are yeah it's
211. 32:36come they're coming out like every three days um increasingly optimized um large language models you can install on like an M1 Mac
212. 32:46um there's one that someone did that can be installed on a phone um so so the developer Community is taking the Llama models that meta
213. 32:55accidentally released to the world they might have been accidentally released them but they did they released them to
214. 33:00the world so now the entire developer Community has these core models that cost millions of dollars to create and
215. 33:06they're fine-tuning them and optimizing them and making installs for Macs and PCs and things like that so
216. 33:13um just go look at the open source stuff there's there's a lot there you may need to geek out a little bit to get it
217. 33:18installed but it's getting easier every day handsome Rosie O'Donnell thank you all right Duke what book would you recommend
218. 33:26Duke I've already answered and used I'm asking the question oh Duke I use Bing for basic AI yeah Bing's pretty good
219. 33:40thanks been pulling out my hair and I don't have enough as it is yeah exactly my I look much smarter now
220. 33:48because my forehead is so much larger it's just the hair just keeps going and then yeah the glue's holding pretty good today
221. 33:59it's not it's not a total disaster um after the Apple launched Tim Apple went on morning TV to talk AI Zuck went on
222. 34:08Lex podcast on VR yeah Zuck did one of those yeah well whatever Apple introduced wasn't very impressive our
223. 34:16howdy partner how you doing today um Zuck is in a little bit of a defensive posture because Apple successfully they
224. 34:26didn't use the word metaverse once they didn't use the word AI once and they introduced some really revolutionary
225. 34:33technology and somehow mysteriously they took a thing that is form factor wise and from a technical perspective
226. 34:40essentially identical to every other Goggle out there and they somehow successfully redefined that as this new
227. 34:46thing called spatial computing It's amazing And there's apparently 5 000 patents in in in those uh in those classes
228. 34:55um so they're 3 500 bucks so they're really expensive but but what they do and the level of integration of AI and
229. 35:02machine learning and all of that stuff within them is remarkable the other thing that they do is they do
230. 35:07all their AI processing in the goggles themselves they're not sending it off to some other servers so they're going to
231. 35:12win the Privacy game I think they're going to win the the uh well spatial Computing that will be a new thing
232. 35:19that that they will have invented and then um and then I think they're gonna they're gonna win the um like the the autonomous agents where
233. 35:29within the Apple ecosystem it will just learn all about you and and you'll just say hey oh you don't say Hey Siri
234. 35:36anymore you'll say Siri book me a trip to Florida give me a five-day itinerary and get me the cheapest tickets and make
235. 35:42sure that they're cancelable and it'll just go off and do it for you I think they're going to be brilliant at that
236. 35:49I need a notebook I've written too many stick sticky notes with the info you've shared go get a notebook they sell them
237. 35:55now that you can get them at like drugstores you can get them you go to Target or we call it Target because it's
238. 36:01fancier than Kmart used to be um they sell them there there's uh there's office supply stores they sell them they sell pens
239. 36:10it's remarkable it's all AI generated too I am a smart ass by the way uh some might call me an asshat and give me another hat
240. 36:23I'll talk like a cowboy all right I lost my place in the questions again trying to find where I was this is one
241. 36:36of my hobbies scrolling through Tick Tock okay there's Emilio's wife talking about the need for a notebook that's really funny and awesome
242. 36:44I use creative Tick Tock hat yeah creative Fabrica spark for my AI artwork it's free and it's easy cool
243. 36:54Built My Own autonomous agent on gpt4 but keep hitting token limits any tips nope uh if if anthropic so if Claude 100K is available via API
244. 37:09which I assume it is because pose using it I would use anthropics Claude 100K or there's a MPT 7B has a 65 000 token
245. 37:22a model called story something story maker or story I don't know look up MPT 7B 65k here I'll go do it now for you I
246. 37:37m p t Ash 7B 65k or I did SK okay mptb Beats gpt42
247. 37:53foreign this is also commercially usable so that's an option and then the Claude Claude instant 100K is an option and
248. 38:09it's got much larger token limits the other thing you can do if you're hitting token limits is um install Lang chain
249. 38:17and you'll have to do some programming but basically you take large documents and you break them up into into sections
250. 38:24and then you embed them and then you put those vectors into a vector database and it gets really geeky really quick but
251. 38:32you can either use larger token limit models like the ones I just talked about or you can install Lang chain and start
252. 38:41to learn about doing your old all your data or you can wait three months is my guess this is total speculation
253. 38:53and Microsoft will roll all this GPT stuff Within um Office 365 and Google's Gonna Roll barred across all of their office things
254. 39:05and I would be dramatically shocked if you couldn't just point your those things at your Cloud stuff and query them so so it's coming
255. 39:16I currently work in QA but I want to get in AI any advice on what routes I should take um I that's a sweet little heart hello thank you um
256. 39:30I think my advice for everyone in any profession is learn as much as you can about how AI will be used to replace
257. 39:42your job as a QA person right so it's going to be able to do things like create testing regimens and in a lot of
258. 39:49cases if not all of the cases it'll actually be able to go test them um so what I would do is I would go to
259. 39:56chat and I would start writing your own like like have it help you create testing Frameworks just start start
260. 40:04using Chachi BT to augment your job and amplify what you already do and figure out how quickly can you build tools that do what you do
261. 40:16because if you're able to do that with that I guarantee you there's a couple of punks sitting in their mom's basement
262. 40:22right now writing that tool so at some point someone's going to release a tool that's like the QA you know testing
263. 40:30button and I'll be able to upload my IOS app on my Android app and go test this and it will not only test it it will
264. 40:36analyze the code and it'll say okay here are this three race conditions you have here are the here are the 16 syntax
265. 40:44errors you have in The Code by the way I fixed all the syntax errors um I fixed two of the three race conditions and you have a decision to
266. 40:51make on the third race condition do you want to give up this feature give up that feature or do you want to
267. 40:56completely refactor that section of the app like that tool is coming it ain't here yet but I I would strongly
268. 41:03encourage you to start to figure out how to replace yourself in doing that you'll get much much better at your job and um
269. 41:13and you'll be able to identify when the the tools are coming at you that um are going to hit everyone else by surprise
270. 41:25do I partner with others um sure I I'm the CEO of a company called storyvine so I'm happy to talk about anything I'm in a
271. 41:33well I mean less of one for the first three months of this year I was in a just just say yes to everything mode now
272. 41:39now I have over committed pretty significantly on a lot of stuff but um I'm I'm very open to projects because
273. 41:47there's so much going on am I a robot uh I am gpt5 thank you very much you thought you thought gpt4 was cool check this out yeah yeah
274. 42:01all right um hello from West Palm Beach Florida hello hello Isabel favorite plugins I haven't played with a ton of them I mean
275. 42:11the ones that are just sort of clear and obvious to me like link reader and um chat with PDF like PDF readers and and
276. 42:19Link readers are really powerful that's just super cool but you can kind of do that with just gpt4 connected to
277. 42:27the internet but I like I like those plugins they're just very very specific um I mean I don't do a lot of trip planning
278. 42:37but if I did do a lot of trip planning I'd like all of those um I don't know right now I I haven't used them enough I haven't used them
279. 42:45enough and there's a lot of them and it's like like I said I mean it is daunting enough getting your head
280. 42:52around what gpt4 makes possible much less what gpt4 plus 260 micro applications that you can wield within a context of some sort of
281. 43:05app you've conceived so I've been watching all your videos and reading your newsletter in the past two
282. 43:12days mind blowing thank you Jen that's awesome oh yeah I've seen your comments I appreciate the uh I appreciate the
283. 43:17engagement that's cool I'm glad it's making a difference I think you said last night didn't you say like you were
284. 43:22really hot you're like I'm really high and this is blowing my mind it blows your mind when you're that High
285. 43:28um I think there'll become black spots in human knowledge when trading data is selectively excluded the training data being exclusive
286. 43:42selectively excluded kind of lives in the conspiracy theory that two or three Mega corporations are going to control everything um
287. 43:54because the the open source Genie has been let out of the bottle because there are so many open source models now and
288. 44:02understanding how to train these the the the large data sets um is is much more broadly understood now it's possible that that um
289. 44:15that that can't happen now could a country say it's illegal to use open source large language models yeah possibly
290. 44:25um will that stop people that really want to use them no so I thought I I don't see that happening now but you
291. 44:36know listen every everything's up for grabs because this is I mean here's the deal about these Technologies these are
292. 44:42by far the most powerful Technologies ever put in the hands of regular everyday people like the the level of
293. 44:49power that you have in your hand hands even with a safety guard rail gpt4 is significantly beyond what I think any
294. 44:59government is comfortable having their people have access to and then if you take the large language models or you take off the guard rails
295. 45:07um there's a lot of people in power there's a lot of power centers that are about to be upended so I'll give you a
296. 45:14an example of of an area that I know pretty well I I started an early agency and I sold it to Omnicom so I know the
297. 45:22agency holding company model pretty well um I think that the billable hour business model the time and materials model that
298. 45:35all holding companies are based on I think that's about to blow up I think clients are going to start demanding
299. 45:42fixed pricing which is going to destroy the Top Line and profits of holding companies they're going to have to
300. 45:52completely reinvent their business model um so when big Power centers get disrupted like that you know powerful
301. 46:01people are going to do some [ __ ] to try to control to control us plebes so you know you're not necessarily wrong that the big Power
302. 46:10centers are going to try to do that now whether or not they can successfully do that that's a different thing I I mean
303. 46:16the only way that you can really stop this AI stuff in in my opinion if you actually wanted to stop it is throw a uh uh an EMP an
304. 46:28electromagnetic pulse throw some EMP bombs that destroy all the computers but they're not going to do that because
305. 46:34that destroys all the economies right it it it so I don't think it's going to go away all right just like a marketing agency
306. 46:42will there be a need for an AI agency teaching businesses how to use AI certainly short term this is so so all
307. 46:50of these new technologies go through go through the um go through the same phases like I went through this with the World Wide Web we
308. 46:57went through this with smartphones we went through this with social media now we're going to go through it with AI
309. 47:02so it initially starts out like it's a fad it's a phase we're gonna get over it oh the.com bubble burst the internet's
310. 47:10over we're never going to use the internet now don't invest in the internet it was all a waste of time nothing to look at here and then it
311. 47:16[ __ ] transformed everything right um same thing's happening here so so the first phase is they dismiss it the
312. 47:23second phase is they try to destroy it the third phase is they try to protect from it right so universities using chat
313. 47:30GPT detectors what if the students use it the right papers oh my God well it's going to be in Microsoft Word it's going to be in Google Docs
314. 47:40everyone's going to [ __ ] use this for everything so so so that's that the next phase is we need a czar of AI oh you know how to
315. 47:50spell AI you're the VP of AI they put it in a corner and they make someone go figure this out for us and it's off to
316. 47:55the side and like social media like we were talking about integrating digital in 1996 and a lot of big agencies and big companies
317. 48:07are just now starting to integrate digital fully into business right it's it's fully a part of business but but a
318. 48:14lot of them still have these digital centers of excellence where it's off to the side so you're going to absolutely see that
319. 48:21but the tools are going to very quickly I think much more quickly than um than just digital did they're going
320. 48:28to inundate all of all the processes of businesses because it's so much more efficient it's so much more powerful so
321. 48:36I think short term probably for the next three to five years you could make you could have a very good business
322. 48:43um teaching businesses how to incorporate Ai and not just how to incorporate it how to deal with the implications of it like I I think that
323. 48:54might be an area there's sort of tactical teach them how to use it but there's how do you deal with if you're a Holden company
324. 49:02and you rely on the time and materials business model and you can con you can scenario plan out it's not going to be that many years
325. 49:10until we can't do that anymore well what do you [ __ ] do do you go first do you say [ __ ] it we're gonna we're gonna go to fixed pricing
326. 49:19and we're gonna beat all the other holding companies to the punch that's that's a possibility so AI will be used as a tool not
327. 49:26something that will take over jobs well so AI absolutely AI is a tool Microsoft is branded their AI copilot
328. 49:34which feels like the right branding to me like that's the right positioning it feels like the right definition humans
329. 49:39are the pilot AI tools are the co-pilot they happen to be incredibly [ __ ] capable co-pilots like I guarantee you
330. 49:47that AI can can land a plane in wind shear at some point maybe not right now but they're gonna it's gonna be able to
331. 49:53land a plane in wind shear better than a human pilot but as long as human lives are at stake human Pilots will still be in charge for
332. 50:02many years to come right it's going to take many many years for that to happen um but it's a tool AI will not take jobs
333. 50:13people using AI will take jobs and jobs will transform think anywhere where you've got incredibly complicated processes and you
334. 50:24have job people's whose job it is to deal with the complexity of the thing so think Insurance underwriting think um legal research think
335. 50:38um I don't know anywhere where you've got incredibly complicated systems oh you know like medical paperwork you know
336. 50:46patient records right now incredibly complicated because the insurance companies tell you you have to put in
337. 50:53all these codes right and that's mostly manual and getting more digital but but that's going to quickly move to fully AI automated right
338. 51:03um so that's where you're going to see the biggest disruption is jobs that are ultimately low value to the company but have to get
339. 51:13done right so what's going to happen some companies will just fire all those people right and that's what will make the Nightly News
340. 51:22the smarter companies will say oh I don't have to pay these 10 000 people to to to manually read faxes anymore right I can
341. 51:35just have ai do that what if I train them up to be super high level customer service people we get rid of our
342. 51:44offshore customer service call centers because they're they're clueless anyway like no matter how much we train them up
343. 51:50it's a horrible experience and they never have the answers anyway we replace them with AI and then we take the people
344. 51:56that were doing this low level um you know document processing and we train them to be high level customer
345. 52:03service people so the AI will take the initial call you'll never be put on hold you'll always be talking to a robot
346. 52:09that's way more capable than the current customer service people are and it will know when it needs to escalate your
347. 52:16issue to someone up here if it can't solve it it'll escalate it so you'll call someone will pick up the phone
348. 52:22you'll talk to them it'll be useful if you have a truly unique problem you'll get escalated immediately and you can talk immediately so like
349. 52:31we don't know right now what the new jobs are going to be but there's going to be all sorts of new jobs created but
350. 52:37the short term two to three year window where all of this automation happens is going to be like radically disruptive
351. 52:44it's going to be really painful it's going to suck um how will this affect electronic Discovery can it be used
352. 52:53for early case assessments yeah I think so I I mean like I just did if you weren't here earlier I did so let's go let's go get a
353. 53:08let's go get a legal brief I don't know anything about the law legal brief um examples okay whoa lots of them okay um
354. 53:23I don't know are any of these long I assume some of these are long I don't know what this is like how wait legal brief
355. 53:35um let's see legal brief um um let's do Supreme Court um PDF let's see what we get there
356. 53:53is this a this is bad interface Bing visit site oh sorry just looking at my nose okay is this
357. 54:07long uh yeah it's pretty long okay so I'm gonna do select all there I'm going to copy now what I'm going to do is I'm going to go to PO
358. 54:18so Poe's got this thing called Claude instant 100K so this is a hundred thousand token um 100 000 token large language model so
359. 54:30it's like chat GPT except you can paste in up to 75 000 words so I'm going to go over here and I'm gonna and I have to I have to
360. 54:38this is for the subscription version of poe.com to get access to this but I'm going to clear my my context history
361. 54:46then I'm going to say based on this legal brief colon paste so I just paste it in all of the brief and then I'm gonna go please
362. 55:02summarize um this I guess case let's say this document because I don't know anything about law as if I were a
363. 55:16seventh grader writing a rip writing a report give me mark down formatted
364. 55:30um bullets if appropriate all right so now it just read that whole document and
365. 55:47didn't jump down to the bottom so we're going to scroll all the way to the bottom we'll get there oh it's going to be so
366. 55:53good Jesus this was a long document okay so here's a summarized version of the legal brief in bullet points ice IFC is
367. 56:04an international organization owned by 104 member countries in 1945 Congress passed a thing I IFC was designated
368. 56:12designated as an international thing Indian Nationals who live near an IFC Finance power plant are claiming that
369. 56:20the plant caused them harm right so so this wrote a summary of of that thing so I don't know enough about the world of
370. 56:28Law and legal but the fact that I just took a supreme court legal briefing and summarized it that quickly yeah I think that changes things
371. 56:37I could be wrong but but if if little me could understand what the hell was just talk going on there um it looks like the the uh the the
372. 56:49number of people here are kind of dwindling so I'll either go but like tap the screen a bunch and um if you can
373. 56:55share this uh share this live I don't if you know how to do that I don't know how to do that um just so we can get some more people
374. 57:02in here uh otherwise I'll just I'll just take off um openai has gotten about 75 of my coding questions wrong I would say um
375. 57:11go to GitHub copilot which is gpt4 that's that's dedicated to coding that's probably going to be better and if
376. 57:18you're not using gpt4 I understand that gpt4 is way better than um GPT 3.5 you might also try Bard so bard.google.com is really [ __ ]
377. 57:31horrible for it for writing for language stuff but apparently it's strong at code so I would also say give Bart a shot
378. 57:40um work with time 100 true not sure what I said I know open AI but what are the other sites oh so bing.com is
379. 57:48Microsoft's search engine but it also has button you find a little button and you push the little button it's so exciting all right
380. 57:57so you go to bing.com and it has this new little button and the new little button is called chat and chat is that is gpt4 that's like
381. 58:08chat GPT but it's gpt4 and it's free and it's connected to the internet and then it's got like a couple of modes it's got
382. 58:15more creative more precise so if you're doing research go to more precise if you're doing you know writing a
383. 58:21screenplay or the Great American novel go to more creative so that's Bing and then Poe is the thing I just showed with that
384. 58:29legal document Poe gives you access to a bunch of different models Sage is pose model and and the company behind Poe is
385. 58:37quora right so the question and answer site that that's been you know doing Q a for decades they took all that data and
386. 58:44so Sage is is pose large language model you got gpt4 you got Claude plus which is anthropic's version of gpt4
387. 58:55you've got Claude instant 100K which is the thing where you can put in giant documents um Claude instant is like GPT 3.5 and
388. 59:04chat gpt's GPT 3.5 the other thing you can do on Poe is you can make your own bots so like I made a Shakespeare bot
389. 59:11and a Yoda bot and a sling blade bot and a Emily Dickinson bot so you can make your own Bots which is kind of fun it's
390. 59:18it's a good way to practice your prompt crafting and then the other two URLs prompts.chat is just a document that
391. 59:27will teach you about prompting and then gives you like hundreds if not thousands of prompts that you can try like act as
392. 59:34a legal assistant or a paralegal act as a novelist act as a mathematician which surprisingly actually makes uh chat GPT
393. 59:45work better and then futurepedia is just a directory of massive massive amounts of tools AI tools so if you want to go
394. 59:54see all the video tools and the music tools and the whatever it's there's just there is right now an infinite amount of things to explore
395. 1:00:05and so what I've personally done is I've just kind of locked myself down to this [ __ ] plus the image generation tools
396. 1:00:12that I like which right now is mid-journey um all the other stuff I'm just I don't have the time to play with but you can
397. 1:00:18if you want to run down a rabbit hole like I I do strongly suggest like if you're a musician go find all of the
398. 1:00:25music tools out there figure which ones are good which ones are [ __ ] and then like just start learning that corner of the AI industry
399. 1:00:33it will serve you well are you a gamer on Steam at all no well I'm a Sim race I like Sim racing it's it's good for my ADD
400. 1:00:44um it's like left right you know Stomp the gas um but but any any sort of games that require attention and learning and not
401. 1:00:54so good uh but I've I've not been on Steam um how do you see AI affecting education schools in particular if they don't um
402. 1:01:04if they don't quickly get their [ __ ] together and stop resisting this and start embracing it they're gonna quickly find themselves out of
403. 1:01:11um out of students because well not quickly these are Big institutions they're going to take for [ __ ] ever to change but um
404. 1:01:20the the whole resisting that this is coming is is a feudal resistance um and and there's a real opportunity for teachers
405. 1:01:32to reinvent how they engage with students drive a lot more critical thinking um so rather than saying you can't use
406. 1:01:39chat GPT to to write a paper why don't we bring chat EPT into the classroom have it write a paper and then have the
407. 1:01:47students analyze what chatgpt wrote and figure out what did it do good what did it do wrong where did it hallucinate how
408. 1:01:54would you do it better really demonstrate your knowledge of The Thing by analyzing a thing that the tool made
409. 1:01:59and in doing that you'll get better at using the tool right because these tools are going to be everywhere
410. 1:02:04so so I think the effect is going to be dramatic if you want to see um a really uh cool thing so so this was cool before AI so Khan Academy
411. 1:02:16um salcon's Khan Academy where you know he's got this idea of flipping the classroom where like you do the learning
412. 1:02:22part out of the classroom you do the engaging part in the classroom like I think we're moving there they got access to gpt4 back in August
413. 1:02:30of 22. so they spent six or seven months refining the experience and they created this thing on Khan Academy called
414. 1:02:39conmigo which is a clever name but it's basically a teaching assistant that will use Socratic reasoning like if you just
415. 1:02:47ask it to answer a question for you it won't it will query you and go back and forth with you and teach you the
416. 1:02:52concepts and it won't move on until it has done you have demonstrated to it that you understand the concepts sitting
417. 1:02:59underneath the thing so so I think the um I think the impact on education is going to be quite profound and I think
418. 1:03:06Educators that are resisting it need to get their heads out of their asses um have I run into any tools for
419. 1:03:13instructional design such as creating how to's gpt4 let's let's go to the video the masterful at uh instructional design
420. 1:03:24I've done a lot of it but I don't know [ __ ] about it um I I don't know [ __ ] about it as as a as a profession but that's one of the
421. 1:03:36things that that chat ept's [ __ ] masterful at so here's what we're gonna do I want you to act as an instruction instructional designer
422. 1:03:52and help me design some course work the first thing I need to know are
423. 1:04:06um no is what are the main Frameworks for creating instructional
424. 1:04:21materials let's just call it that creative creating them requires a systematic approach so there's the addy model there's the
425. 1:04:34Sam model successive approximation model well of course there's that one the Sam model everybody knows that
426. 1:04:41well you know dick and Carrie right they came up with a model Bloom's taxonomy you know I haven't played with Bloom's
427. 1:04:48taxonomy in like a week there's Kirkpatrick's four Level Training evaluation okay it apparent there's a lot more of these than I had a
428. 1:04:59[ __ ] clue about so there's nine or there's seven of them here Ganges nine events of instructions I'm gonna say
429. 1:05:06okay so I'm gonna say okay let's go with gagny period um I want to teach a one hour
430. 1:05:20generative AI overview for people in the marketing department that will
431. 1:05:36that will move them from fear to curiosity please give me that outline
432. 1:05:53sounds like an interesting session that's one of the things about chat GPT is it it responds in a very human kind of way I find myself like
433. 1:06:03um acknowledging it like hey you did a good job there um I I find myself using please and thank you as if I'm talking to just like
434. 1:06:10a really capable intern but okay so what's it done gain attention five minutes start the session with an intriguing question or a
435. 1:06:18statement such as imagine a world where AI can generate creative content for your marketing campaign show examples okay that's cool
436. 1:06:25inform Learners of the objectives we're going to State the objectives of what you're about to learn stimulate recall of a prior learning
437. 1:06:35um what do you already know about AI in our marketing marketing department that that the boss said we're not allowed to
438. 1:06:40use chat GPT right whatever present the content and then the content introduction case studies benefits challenges provide learner guidance
439. 1:06:49explain some possible first steps illicit performance provide feedback assess performance enhance retention and transfer right so
440. 1:06:58there's the nine steps of the gagny approach which I knew nothing about they sound reasonable to me they've blocked
441. 1:07:04out the time for the hour and I'll just say um What If This Were a three-day session what might the agenda look like
442. 1:07:19boom expanding a one-hour session into a three-day session allows for more in-depth exploration of generative AI day one
443. 1:07:31so so it's doing the same nine hour or the nine segment gagne approach but it's just splitting it into three days and expanding it
444. 1:07:43case studies and Hands-On session so now are there other tools there might be other tools for that are specialized to
445. 1:07:58um to instructional design and if you want to find those go to futurepedia.io and just look type in instructional design
446. 1:08:06there might be some created for it what you'll probably find is that you probably know more like the tools that
447. 1:08:14are going to be dedicated to this might not be all that exciting I I would go teach yourself how to do it with gpt4
448. 1:08:21okay enough of mentioning creative fabric of spark this isn't don't spam in here please leonardo.ai I think is
449. 1:08:30really a web front end to stable diffusion um I think it's still in preview though leonardo.ai but I would I would
450. 1:08:41encourage you if you want to do image stuff I would encourage you to um get on the wait list for this because it's um
451. 1:08:56cold please so so it gives you lots of different models so you can create images in these different models across the top
452. 1:09:10like here's a vintage photography style so I want to generate with this model I don't want to tell you what role I am I'm a professional
453. 1:09:22what ah pop-ups make me right so I'm going to say a um 70s muscle car oops not a 702 muscle car 7ds muscle
454. 1:09:39car on a winding Mountain Road you know going fast and so I so I picked a model that does vintage photography so this
455. 1:09:55look should look like some vintage photograph of a 70s muscle card doing something um I think it's decently fast you're
456. 1:10:06currently on the free plan yeah there you go so there's there's a bunch of you know cars on Windy mountain roads that one looks pretty good
457. 1:10:18but I don't like that color so I want to do um 70s lime green muscle car on a on a Mountain Road going fast I'm
458. 1:10:37going to say um cinematic motion blur so sometimes adding things like that will help but anyway there's lots and
459. 1:10:47lots of other things that Leonardo does you can easily train if you've got photographs like I'm a photographer so I
460. 1:10:52uploaded a bunch of my sky photographs to it it learned them and now I can just call that that style so let's let's get our green car so now
461. 1:11:02here's our green car right so Leonardo is really good I think it's I think it's awesome that's a little weird looking
462. 1:11:10that sort of looks like a GTO Judge that looks like an old Mustang it looks like a 7D Mustang 69 Maybe 70.
463. 1:11:20that sort of looks like a cross between a Camaro a judge and a Mustang so anyway um yeah I think it's good good call good call on the tool
464. 1:11:32all right how I feel about faceless videos using AI how do I feel about faceless videos using AI you mean like text videos I
465. 1:11:40don't quite know what you mean I think all videos are going to use AI I have a video uh technology company that has
466. 1:11:46human beings tell their stories and and um I am I am deeply exploring all of the Avenues of AI including
467. 1:11:55um what if we didn't need people which my initial response to that was go [ __ ] yourself Ai and then the more I thought about it the
468. 1:12:04more I thought we're gonna have a competitor that comes up against us like where we go into a client we can say Hey you could have
469. 1:12:11like 20 of your employees talk about how awesome it is to work here someone's going to come along with a product that
470. 1:12:16says I can take a pic picture of your employees their their employment record and I can generate a video per person
471. 1:12:25um talking about how awesome it is to work here and then they just all they have to do is sign off on it yes I'm
472. 1:12:31okay with you having my likeness say that it's great to work here that's going to be one of my competitors so
473. 1:12:39I should probably figure out how to do that [ __ ] so it here's the thing it's almost irrelevant how I feel about it this
474. 1:12:47stuff is coming whether you want it or not I think that's the thing that that maybe why this channel exists is that
475. 1:12:54the disruption that this generative AI stuff is going to cause is coming whether you want it or not what I'm
476. 1:12:59choosing to do is stand on the surfboard on the tsunami rather than be on the Cozy beach cottage which is where where
477. 1:13:06we all are in our current jobs right now which is terrifies terrifying how can I beat AI detection programs there's lots
478. 1:13:14of ways to do that there basically it's basically you you take something that was generated with chat EBT you throw it
479. 1:13:20into something that rewrites it and rewrites it in a way that's not detectable but that's a war that's going
480. 1:13:24to go back and forth here's the deal I think that's a very short-lived thing if you're dealing with it you're going
481. 1:13:31to need to find ways I don't know what the tools are because I don't give a [ __ ] about them because when Microsoft
482. 1:13:36includes generative AI in Office 365 and then uh you know in Google does it in Google Docs you're it's gonna it's just gonna be in
483. 1:13:46everything like those those writing detection tools are going to be um just very short-lived in my opinion
484. 1:13:53um who said that quote hey I want to take your job but someone using it will Paul rateser r-o-e-t-z-e-r from the marketing AI
485. 1:14:02Institute you should follow him on LinkedIn um he posts good posts and then you should listen to his podcast they've got
486. 1:14:08a weekly podcast where they sort of catch up on the week of AI what's cool about Paul and the marketing AI Institute it's like
487. 1:14:16five or six years old so he's been on this train since like the early days of like gpt2 gpt3 like he's seen this
488. 1:14:24coming all of a sudden you know since chat EBT you know he's he's way ahead of the curve the thing about Paul and what
489. 1:14:33they talk about there is I just like every time I think huh I kind of think it's going to go in this direction I'll
490. 1:14:39get on the podcast and he'll say I think it's going to go in this direction like I just I feel like I'm personally very aligned
491. 1:14:46um just in terms of kind of analyzing what's happening with what they're doing there so I I think they're he's a really good guy to follow
492. 1:14:55I'm looking for an AI to create gauge graph like fuel gauge um I assume we can just go back to Leonardo well let's go to well yeah we'll do
493. 1:15:07we'll do Leonardo um um let's see classic um fuel gauge
494. 1:15:24face with black backup and yellow tick marks
495. 1:15:40um modern well I call it classic so let's see um updated design I don't know whatever that's called design we'll see what that generates
496. 1:15:54like any of the any of the image generation tools you know so so what I would start teaching yourself is
497. 1:16:00prompting because the the like where we're heading is what the what you effectively have to do with these tools
498. 1:16:07is you have to be good at critical thinking and good at asking for stuff so there's some fuel gauges right
499. 1:16:14um if that's not what you want then you basically just describe what you do want now you'll notice that the numbers and
500. 1:16:21the words are messed up the these tools are not good with with words and numbers yet like they're out of order but but you
501. 1:16:29know you you get the the basic idea here right so what did that take me 30 seconds so go just go start experimenting that the
502. 1:16:42thing about these tools is they're so much more capable than than I think they are like every time I try
503. 1:16:49something new where I'm like well it probably doesn't oh [ __ ] it does that good too best used for the link reader plugin um
504. 1:17:01have it go read the pages of your website and give you SEO well there's SEO analysis tools out there um what I've been using it for is I'll
505. 1:17:12take scientific white papers or I'll take like a really complicated article and I'll turn that into a Twitter thread
506. 1:17:18or I'll turn that into a LinkedIn post um where it's not writing something new from scratch it's like read this long
507. 1:17:25document and then summarize it for me in some interesting way and as as I get the prompt styled in like the I have a I
508. 1:17:32have a Twitter thread maker right now that's [ __ ] brilliant like it's just it's not what I did isn't brilliant but
509. 1:17:38the output of it is consistently like 95 percent complete like I don't need to edit much in it it's and you know and it takes me
510. 1:17:4930 seconds rather than having to read it for 20 minutes to four hours because of my 80d and then go in and highlight the little pieces of it
511. 1:18:00what were the actual Salient points here like like I did a um I did a 15 segment Twitter thread on Mark and drayson's you know AI might
512. 1:18:10save us thing it was just it was just brilliant at it all right won't this only give you results through
513. 1:18:20a previous year well so the the core model for chat GPT stops at September 21 but gpt4 connected to the internet or
514. 1:18:29bing.com connected to the internet um do both so there they are it does not stop there how do I see AI affecting education we
515. 1:18:41covered that one instead of organizing of course I'd prefer Zoom class where folks can ask questions and try ideas that's
516. 1:18:52not a bad idea I might do that Amelia's wife I gotta figure out how to set that up but I can go ask chat GPT to help me set that up
517. 1:19:01I did a I had a um a podcast for a year where I used stream yard I just read that paper today yeah it's
518. 1:19:09good paper salvation like cleaning the planet from cancer um that kind of stuff's coming genetic diseases cancer it's going to
519. 1:19:19start solving it's going to start so Sam Altman was being interviewed once and he said he thinks it's plausible that in a
520. 1:19:26relatively short amount of time we're going to have a single year where we accomplish as much scientifically as we have in the past 500.
521. 1:19:37and and if you want to do an interesting exercise go to gpt4 and have it make you a table of the top 30 or 40 scientific
522. 1:19:46achievements in the past 500 years and just look at all the stuff on that list and you'll be like oh yeah oh that was a
523. 1:19:52big one oh that was a big one oh the polio vaccine right you just go down through them flight um and then imagine that all happening
524. 1:20:01within a one-year period like crazy how can you vectorize large chunks of text and pass pass it as context to gpt4
525. 1:20:12so you have to use something like Lang chain the good news about Lang chain and these tools is that the developer
526. 1:20:18Community is is making it easy to do that like I think you can install Lang chain now in like I don't know 20 lines
527. 1:20:25of code or like it's some ridiculously stupid small amount of code but you'll probably need to set up a virtual
528. 1:20:31machine and and then you know you'll you'll have a module that will chunk up your large pieces of data it will embed
529. 1:20:38it it will vectorize it then it'll put those vectors in a vector database and then you'll be able to just interact it with it like chat GPT
530. 1:20:46um I saw a thing on Twitter like two days ago that someone's created a a company that is basically putting a front end on top of Lang chain
531. 1:20:57um so Lang chain is the thing I would go um explore there it's but it's going to be geeky is there any AI that can grade
532. 1:21:05my algebra students worksheets probably just chat GPT so um yeah it it knows math and and chat gbt with the Wolfram Alpha plug-in like
533. 1:21:17if it's not doing accurate grading of the algebra worksheets um oh I guess one question I have is if it if you need it to like
534. 1:21:28scan the visual and then grade it there you're gonna have to wait for it so gpt4 is multimodal it can it can use
535. 1:21:38pictures but the current implementation of it it can't so you're just gonna have to wait a little bit if you if you
536. 1:21:44wanted to like look at like handwritten algebra um but if it's if it's the kind of thing where it's in PDF form where you can just pop it in
537. 1:21:53um you can probably do that right now and and there may be a tool like I would say go to futurepedia.io and look up
538. 1:22:00document analysis or or like um computer vision or something like that there's probably apps right now that can do that S3 photographer
539. 1:22:11and already trying to learn AI applications loving Firefly yeah Firefly is amazing and so is generative fill in
540. 1:22:17Photoshop if you have not seen generative fill let's see oh no that's not a fun one hang on let's go get this one this is a fun one
541. 1:22:29so here we have some dude some some sailor dude in Times Square and so I can just select this bottom white area oh sorry
542. 1:22:43so there's the dude in Times Square this is something I made in mid-journey and so in this so to get to this if you've got Photoshop
543. 1:22:54um Creative Cloud you go to Beta apps and you download the Photoshop beta and and it's got this in it so I just
544. 1:23:01selected the bottom half of this image and it pops up this contextual menu here and on the left this button says
545. 1:23:08generative fill and so I can put in here um Lakers basketball shorts and I'm going to say generate and it should put shorts on this dude
546. 1:23:24that are Lakers shorts now it didn't do it but like it gives us three options let's see they're not Lakers okay so it doesn't
547. 1:23:37know what the Lakers are so we'll say generative Phil will do um yellow basketball shorts oops oh dang it yellow basketball
548. 1:23:52shorts with purple stripes and so it's going off it's thinking it's thinking we're thinking here paintball
549. 1:24:08give him some yellow shorts oh those are funny where's my purple stripes there's no purple in there yeah we don't we don't have I could keep
550. 1:24:18messing with that but that's funny and then you know up in the sky we could do um we could do generative fill um large
551. 1:24:34um Godzilla attacking City and it's thinking it's thinking genius takes time like at least 30
552. 1:24:49seconds and it might not be genius it might be funny all right there's no Godzilla there let's see if it gave us a Godzilla it
553. 1:24:59did not oh there's one there's a monster so it's on top of the sign so anyway that's generative Phil if you
554. 1:25:06haven't played with generative Phil it's just amazing and then um who was it that came out with it is it stability
555. 1:25:13I think stability came out with uncrop so let's go play with that um stability uncropped so stability has a a site called clip
556. 1:25:28drop so it's clip oh dang it clip drop dot Co and then one of the and then you go
557. 1:25:43slash it's clipdrop.co slash uncrop which is a good brand name and then what you do it's it's it's exactly like what we just looked at but
558. 1:25:53this is it's a more simple implementation it's pretty smart um let me go to finder what's this picture oh that's the video
559. 1:26:07I made okay here's one so I'm gonna just grab a picture here and I'm gonna drag it yeah I'm gonna drag it into the little window
560. 1:26:18and so it gives me that and so now I can grab these blue handles and I can just you know expand them to like if I want
561. 1:26:28this little version of me to be in a city I can just sit hit next and it will now uncrop that cropped image
562. 1:26:48oh that's cool that's really cool so there's one like I didn't have to type anything that put a weird hand there
563. 1:26:56but there's another one and also with a weird hand that's not a weird hand like like it seems to like it does a really good job
564. 1:27:06of inferring that this is like a futuristic sci-fi world like that one's really good right so I don't know people
565. 1:27:20everything's changing everything is changing it's all changing you should be excited and you should be terrified you
566. 1:27:31should be both if you're not pee in your pants then you should be pooping them all right Jarvis from Iron Man is coming
567. 1:27:40I agree yes balance balance between excitement and fear um is there any AI that can grade my algebra I think we talked about that
568. 1:27:47where is he reading comments from way way way earlier I'm very behind because I sit on these things for like two hours
569. 1:27:55I tend to sit on these for two hours until my butt gets numb and I dehydrate myself because it's just how my
570. 1:28:02personality works and I also try to answer everyone's questions so you know that's just the that's just the
571. 1:28:10shakes I'm an old guy on Tick Tock if you're expecting me to behave like a twerking 20 something it ain't gonna
572. 1:28:18happen I'm a bitter gen xer so [ __ ] it this is how it works this is how we roll here at the AI learning lab mm-hmm all right
573. 1:28:33oh it just jumped me to the bottom L.A Lakers maybe yeah maybe it'll do that but I don't know ain't worth it you can
574. 1:28:39go figure out the uh how to get the actual LA Lakers it's yeah that would probably do it but I don't know maybe not who knows
575. 1:28:49I'm here to show the concepts you guys can go do the Brilliance my level of Brilliance is is thin I'm trying to find where I was
576. 1:28:58in the questions which this is one of my hobbies when Tick Tock jumps me to the bottom oh I'm too far I went too far I went too far
577. 1:29:08open AI he's gotten about 75 percent of my coding questions wrong boy I'm way far back hold please
578. 1:29:27what was that article you showed the other day about the biggest Week in Tech AI history it might have been at my newsletter every day AI
579. 1:29:46because we had a we had an issue [Music]
580. 1:30:05AGI timelines so issue 10 was about a category five week of AI announcements and then the following week was also a really big one um
581. 1:30:18I I mean what's been what's been happening is that like I I just I don't think there's been a week where there
582. 1:30:26hasn't been a significant amount of of uh advancement so I don't I don't remember what article it was sorry about that yo
583. 1:30:33man I have a project that I've been working on and most of it is still hand written how can AI fix that just go this
584. 1:30:40is why I have these here so I I did this I'll tell you a cool a cool thing this was another one of my epiphanies
585. 1:30:48so go to just go start using the tool um and start taking your handwritten stuff and and start collaborating with chat GPT so I I
586. 1:30:57had created a card game about aliens um like four or five years ago and because I have 80d like I went through the sort
587. 1:31:07of the burst of of the idea and and basically I got it as complete as the uh for for as long as the adrenaline lasted
588. 1:31:15right so I think I had like 45 cards identified like there were aliens and there were spaceships and there were
589. 1:31:21areas like Area 51 and S3 and all you know all that sort of [ __ ] and then there were like government
590. 1:31:27cover-ups there was all sorts of stuff right so I had I had like 45 cards written just the title of them I had
591. 1:31:33about a third of them descriptions written and then it all it hit me in putting that together that doing the research to like
592. 1:31:44write all of the cards I would need was a daunting task and I was like oh [ __ ] it and so when I started playing with chat
593. 1:31:50EBT I'm like I wonder if it can do if it can help me with this so I went into my Excel thing with my you know
594. 1:31:59radically incomplete list of cards and I copied and pasted it and I described what the card game was and I said here's
595. 1:32:06the beginnings of a list of cards and I pasted those in there and I said now give me 15 categories of cards that could be
596. 1:32:16part of this game and Boom Out came the 15 categories and then I said now now um like give me a bunch of you know government locations
597. 1:32:26right as as cards with the description and what happened there boom out they came within like 60 minutes I had 300
598. 1:32:36well-written fully researched cards for this card game like and and then and then I even put one of the cells and I you know I pasted
599. 1:32:46it all back into uh Google Sheets and and I did an open Ai call in one of the cells so I used the description of the
600. 1:32:53card to go to to Dolly and generate the image for the card now Dolly sucks so the images sucked but like I got a proof
601. 1:33:02of concept of taking like a poorly written poorly researched project and within 60 Minutes I totally blown the doors off it
602. 1:33:14I'm a gpt6 model very nice to meet you gpd5 here all right hitting token limits build your own exactly there someone just I just saw
603. 1:33:26that this past week um some group of researchers um has created a model that's got a 5 million token limit so so basically you
604. 1:33:37could probably put in all your company's data or you know some large chunk of it um they said it's not very good right
605. 1:33:43now but it can take five million tokens so do you have a like list of shortcuts or blog anything yeah everyday AI so it's if you go to um
606. 1:33:56everyday AI news on Twitter There's the link to the subscription there it's free where's my Twitter where's my Twitter I lost my Twitter
607. 1:34:12so profile so if you go here wait see the uh right there at everyday AI news is the Twitter handle go there and that's we do
608. 1:34:29we do a weekly um wrap up of the the big important stuff going on in AI um also Rachel Woods here on on she's
609. 1:34:39got a newsletter called the AI exchange that's pretty good um there's a bunch of them just start following people on Twitter that that
610. 1:34:46are into this AI stuff you'll get up to speed pretty quick is there a business opportunity to teach small businesses
611. 1:34:52how to use it yeah I think there is how would you go about it um start giving it away for free teach yourself teach yourself enough
612. 1:34:59about how to use these tools to do if if you're good at marketing then you can Target like you know companies doing marketing stuff if
613. 1:35:10you're good at process you can have it do that so I'd say figure out what you're good at learn how to use the tools and then give away
614. 1:35:17three engagements you know reach out to people that you know that have small businesses and say hey can I come in and
615. 1:35:23teach you how to do this for free so I can just get practicing and then once you do it for free then you just come up with a price
616. 1:35:29500 bucks 10 000 bucks whatever it is what's a good business to start using AI the one I just described is one um
617. 1:35:39like right now things are changing so fast that I wouldn't get too married to any solution or any um tool right now because things are
618. 1:35:48moving so fast so I think I think the the best thing to do in business right now is just figure out how to solve an existing
619. 1:35:57problem faster using generative AI because that's that's going to be the like every business is going to have to
620. 1:36:05level up like everything's going to get more efficient it's going to do this so just take something that you know and
621. 1:36:11figure out how to do it better faster or cheaper using Ai and then start putting that out in the world and just you know
622. 1:36:17like keep iterating because something that seems remarkable today is going to see it's going to be a big yawn
623. 1:36:25six six months from now so don't get too married to anything but start putting stuff out in the world and start talking
624. 1:36:31about it on LinkedIn like as you learn stuff like seriously I would start like join I created a thing called the AI Salon the
625. 1:36:40salon.ai go join that come hang out with us we we meet every other week we've got a Discord channel one of the women that that started with
626. 1:36:50um the AI Salon she put that she was an organizer of the AI salon on her LinkedIn profile and got a job a week later doing AI um
627. 1:36:59and she didn't know much about AI before that like it was a new thing so like start talking about hey I'm teaching
628. 1:37:06myself chat gbt you can make money um I'm starting to teach myself um chat TBT and here's what I learned
629. 1:37:14and here's you know I tried this and it didn't work but I tried this and it worked great all that sort of stuff
630. 1:37:20how would I generate passive income I don't know I I mean like like all the passive income stuff most of those
631. 1:37:27people are making money teaching you how to make passive income that's how they're making their money not on
632. 1:37:33passive income they're making their money on teaching you how to right it's all [ __ ] it's you know you got my
633. 1:37:38money today make money while you're asleep it's all [ __ ] um if you can find a passive income business
634. 1:37:52like selling T-shirts or [ __ ] like that you can use generative AI to make the images and write the copy and do all
635. 1:37:59that but use and and help you do the marketing and the marketing plan and all that but you still have to do the business and and
636. 1:38:06the quality of the stuff you create has to be good enough to rise above the noise right so there's no magic formula
637. 1:38:13it's all just you know are you solving a problem that someone else is willing to pay for and if you are doing that then
638. 1:38:21you have a business and if you're not there's there's there's not a Magic Bullet when it comes to that stuff
639. 1:38:27unfortunately and and the problem is if there is one and it's really that easy then everyone will start using it and it'll stop
640. 1:38:35working right so there are blips in the radar where someone will crack the code on something like you know I made fifty
641. 1:38:41thousand dollars in three minutes you know with this thing and then they put out the course on that and then
642. 1:38:46everybody does that and all of a sudden that thing stops working and at that point they're just making money on selling the course
643. 1:38:53I heard a good argument about AI learning that human Innovation comes from mistakes oh that's interesting yeah
644. 1:38:58and it's like what AI does is it gets you to 80 really quickly and and like if it gives you shitty stuff and you keep
645. 1:39:06working with it it'll give you increasingly better stuff that's it that's that's really interesting Delta gummies and GPT can be crazy oh
646. 1:39:14that's funny um uh that sounds like crypto it's a fad now now destroy then Mass adoption uh yeah although crypto kind of screwed
647. 1:39:26the pooch because they had too many crypto Bros with not actual businesses that were just scamming people out of
648. 1:39:34money so crypto and blockchain have some catch not some catching up they have some reputation some well-earned
649. 1:39:41reputation repair to do before people are going to trust it again I think AI is actually going to drive adoption of blockchain and nfts
650. 1:39:49um in in very specific use cases like authenticity um so so we'll see flight computers already land yeah I know they do but you
651. 1:39:57know you know there needs to be the override and all that sort of stuff like we're not going to have um
652. 1:40:03pilotless airplanes anytime soon as long as insurance companies are involved they won't have it we'll always have elevator operators
653. 1:40:11exactly that's true judges have biases let's use AI to process the law yep and if you're interested in that Ampersand
654. 1:40:19um check out the company anthropic they've got what's called a constitutional learning model where their AI they write basically a set of
655. 1:40:26rules like don't hurt humans increase Prosperity uh whatever whatever their rules are and then when it processes an
656. 1:40:33answer it runs it through the Constitutional model and iterates it until it passes so so that's a company
657. 1:40:39that's based on you know some form of automating the the um analysis piece so as not to introduce bias and and I I assume what you'll be
658. 1:40:50able to do is is opt into the kind of Constitution that that matches your values I had a help desk interview last year
659. 1:40:57told them I was working on python for AI in the future cool because AI would have our help desk jobs no offer well
660. 1:41:07it's gonna destroy the help desk uh world as we know it it's going to make it much better what operate chat gbt I don't know what
661. 1:41:17that means why no one explains how it works you want to see how [Laughter] hold how it works when you understand how this [ __ ] works it is
662. 1:41:35more mind-blowing than before you knew it so where do I have it here it is okay so we're going to talk about how Chachi
663. 1:41:50BT works and if you are a machine learning person and a scientist and if I [ __ ] this up and butcher it please let me know
664. 1:42:00but I've done this enough now that I think we're we're generally in the neighborhood so if oh I can do this I've
665. 1:42:05got it presenting in a window let me make this smaller hold for a second here wait is that in a window
666. 1:42:23it is isn't it or is it not hang on hang on I'll be right with you [Music]
667. 1:42:38okay all right so I got this thing so now it fits in there okay so we'll talk about a couple of things generative GPT stands for generative
668. 1:42:49pre-trained Transformer so generative means that it's generating stuff let me get rid of this background that's distracting that's good okay
669. 1:43:03generative means that it's generating stuff pre-trained means that it's pre-trained you don't have to bring your own data machine learning
670. 1:43:12um models you used to have to bring your own data and you can only use relatively small data sets the Transformer the T in
671. 1:43:19GPT stands for Transformer that's the technology that allows these models to be trained on really big training data like all of the internet
672. 1:43:28so so GPT Bard you know Claude all of these things the base model is trained on massive amounts of data um so generative it makes [ __ ]
673. 1:43:38pre-trained you don't have to bring your own data Transformer is the model that it's trained on that's what GPT stands
674. 1:43:44for okay the other thing you hear a lot is llm large language model that's just basically this is the the this
675. 1:43:54particular tool that looks at all this language and you can interact with right it's large it uses language and it's a
676. 1:44:02model and there's lots of different versions of that then there's this thing called latent space and this is this is how this [ __ ]
677. 1:44:11works and it's this is a dramatically oversimplified explanation of it but the latent space is basically a mathematical
678. 1:44:18rep representation of words clustered by meaning meaning so you've got sort of think of it like these semantic little
679. 1:44:25canyons and sub Canyons so all those different colored dots represent like on an x y a two-dimensional x y axis
680. 1:44:35Bunches of different words so um you might have like the blue cluster might be animals and the black cluster might be plants and the
681. 1:44:45orangey cluster might be I don't know different types of wood and what whatever now this is just a 2d um representation but basically what
682. 1:44:54they do is they do what's called embedding and when they embed they take these tokens they take these words and they turn them into a
683. 1:45:02mathematical string like like just a series of numbers that plots that word in a particular Vector space right in a
684. 1:45:11cluster of meaning semantic clusters so so an example might be you might have the word dog that there it might be
685. 1:45:19clustered in a group of words that means animal but it might also be clustered in a group of words that mean you know he's
686. 1:45:26a dog right so so you can have the same word can have different meanings and then what happens is you you create your
687. 1:45:34prompt and it basically takes your prompt and it turns it into mathematical stuff as well and and then um
688. 1:45:43the large language model predicts what should be the next word in the response based on your prompt and then it just
689. 1:45:50does that over and over and over again um and and um and so it kind of works like this so I I typed in the prompt the the quick
690. 1:46:01the quick brown fox jumps over the lazy blank and the answer is dog we all know the answer is dog because our brains
691. 1:46:07work like that right well here's how the large language yeah how the large language model does it it basically
692. 1:46:13deselects all the clusters of words that are not the right sort of semantic neighborhood based on what you wrote and
693. 1:46:19then you can't really see it but there's a bunch of light blue dots there maybe I can make this bigger so you can just see the the
694. 1:46:26this a bit better um so all those light blue dots are numbers right these strings of numbers that are kind of semantically related and then
695. 1:46:39one of those strings of numbers will have a higher probability that says 96 so one of the world's words will have
696. 1:46:46the highest probability that it's the right next word to go in that sentence and then that number that that Vector
697. 1:46:54number is associated with the word dog and then the machine inserts the word dog into that sentence right and it's
698. 1:47:01you know there's other words that could have been there but it puts the word dog in there and it does that over and over
699. 1:47:06and over again so when you see it right like a big long um you know a big long response um it's doing it word by word in in this
700. 1:47:16kind of mathematical way so so this is the reason so I hope that helps for one thing um that's how they work and this is also
701. 1:47:25the reason when they say well it doesn't like it says it hallucinates why does it hallucinate because it's not actually analyzing the
702. 1:47:33thing that it wrote It's just saying I think the next word should be this and then I think the word after that should
703. 1:47:39be this and basically it just does that and it writes all this [ __ ] it's not going back and looking at it
704. 1:47:44there are models being developed right now that do go back and look at it right and say it created a citation here was
705. 1:47:51that citation actually accurate but that's a completely separate process when it writes that big long thing it's
706. 1:47:57completely unaware of what it wrote so sometimes these models will act empathetic right they will present an
707. 1:48:04empathetic response they're not feeling empathy they're just saying based on what you said there you could probably
708. 1:48:11use a hug so so the the next series of words that should be there are some kind of empathetic response so so that's why
709. 1:48:18it gets [ __ ] wrong and but but also when you watch it Write a response like the first chapter of a novel and it does it
710. 1:48:25in like 10 seconds like what's happening mathematically like these people are [ __ ] smart that invented this [ __ ] it's crazy all right
711. 1:48:35all right I gotta run in a few I do two thanks for all the great info you're welcome yeah I gotta go too I gotta go mow or do I don't know
712. 1:48:43whatever I gotta go live my my Saturday life and I'm going out to dinner with one of my employees is up from Durango so that's kind of cool
713. 1:48:51what website is this I don't know which one I don't know which one I was on sorry about that e here we come to save the day
714. 1:49:08what's going on uh move that up too high there we go okay all right hi I hope as well did you try mid Journey yes I heard it's a good one as
715. 1:49:17well I think mid-journey is the best one mid journey is really good resistance to AI is like when calculators came out they took
716. 1:49:26um and now look at math Baseline capabilities that's exactly right they're I think there were even I've seen pictures of this now they could
717. 1:49:32could have been mid-journey created but I've seen pictures that professors were actually like protesting the use of
718. 1:49:38calculators right so same exact thing Paul rates are marketing AI Institute thank you for putting up there hey Kyle
719. 1:49:46how are you today good Liz thank you I'm quite new here but enjoying it cool I'm glad that you're helping one another
720. 1:49:54because I don't get to these questions very fast um a problem in AI inventing inventing sources for papers a problem in AI
721. 1:50:03inventing sources for papers and the sources actually do not exist that is absolutely a problem and and you know
722. 1:50:10that's really on the humans this will get better over time so part of the challenge with all this stuff is that it's very very
723. 1:50:18new and it's very very raw um there was a lawyer I think in the past week that created a had chat GPT write a legal brief that he
724. 1:50:30handed to a judge and he didn't check the sources to see if they actually were right or existed at all they didn't and the judge caught
725. 1:50:42him and said what the [ __ ] is this so if you're a lawyer and use chat GPT for a little leg up which you should be
726. 1:50:52doing because why not it's the most powerful writing tool in the history of mankind check your work that's on the human but but you should
727. 1:51:01just know that these things um that these things hallucinate a lot and and you need to you need to know that and and kind of understand what
728. 1:51:11that actually means and how it does it and why it does it all that sort of stuff and I I mean I just explained it
729. 1:51:16like when I explained how these tools actually work that's why why they hallucinate because it's not
730. 1:51:23aware of what it's writing canva bulk at it with chat EBT tables is insane it's it's absolutely insane if you don't know
731. 1:51:31what that is so you could you can have you could have let's go do it we'll go to chat GPT chat GPT you can make money
732. 1:51:40all right we're gonna do a new thing we're going to go to gpt4 because it's just better we're going to say create a table of um T-shirt
733. 1:51:51sayings uh um we did one the other set let's say four um Gen X dad's um
734. 1:52:05who let's see um just want to be left alone um they should be funny and memorable
735. 1:52:20and um short enough for t-shirts um in the table
736. 1:52:37have um a suggested image wait suggested image the saying and a target audience
737. 1:52:54as columns okay so this should just create us a relatively brilliant table of t-shirts that we could go make
738. 1:53:06so the suggested image cartoon of a man on a desert island I'm not anti-social I'm pro-solitude okay the first one is [ __ ] good
739. 1:53:14I'm not anti-social I'm pro-solitude with a dude on a deserted island a cartoon of a dude on a deserted island
740. 1:53:21really that's awesome okay so so so this has now just generated this table you can copy and paste this into a CSV file
741. 1:53:30into a Google Sheets file create a CSV then you go into canva and you could pick a bunch of designs and you
742. 1:53:36basically just say swap out these text elements with what's in the CSV and it will spit out 20 or 30 t-shirt designs I mean or
743. 1:53:47Twenty or thirty thousand there there was some jerk on tick tock he was he was making I think he was making a hundred
744. 1:53:55thousand t-shirt designs a day and listing them like he he had figured out generative AI it wrote a hundred
745. 1:54:03thousand you know t-shirts generated designs for them and uploaded them to some T-shirt shop or I think it was just
746. 1:54:10designs I don't think it was sayings like this but yeah that stuff's absolutely amazing um all right I've gotta I gotta run so let
747. 1:54:22me do a couple more questions I'm kind of down at the bottom I skipped a bunch of questions so Tick Tock told me I did a great job I
748. 1:54:29went live for 90 minutes I I tend to get that we're down to 19 people here so this is just the hardcore the the
749. 1:54:35hardcore AI learning lab fans so thanks for hanging out with me what's the name of the weekly group it's called it's called
750. 1:54:42um the the AI Salon it's the the there's a link to it in my bio and it's the salon dot AI is a link tree that's got
751. 1:54:52the Discord it's got the Meetup info for the actual meetings it's got a LinkedIn group so yeah it is on Discord and you can do
752. 1:55:00mid journey within our Discord it's pretty cool saw someone call themselves an AI expert you better catch up Kylie yeah I know
753. 1:55:07experts everywhere here's the deal anyone that tells you well there are AI experts that are like machine learning
754. 1:55:14scientists they're experts but anyone talking about this generative AI [ __ ] if any of them tell you they're an expert at this right now
755. 1:55:21they're wrong no one's an expert right now everyone's trying to figure this [ __ ] out I sell a legit literal step-by-step
756. 1:55:28course on how to digitize any business and has mrr that's awesome it has over 90 modules crypto a fad no
757. 1:55:35digital currency is the future yeah I agree with that but they they have some uh some well-earned PR [ __ ] they got to dig out from
758. 1:55:46um Jen what is the significance of the Wired Magazine behind you um I'm in it so this is from July 1995
759. 1:56:02and I've talked about I started a one of the first online magazines it was called Urban desires
760. 1:56:16I probably should have bookmarked this Nate did I was it just there no I'm in the right section I think
761. 1:56:30that was me 30 years ago um and that is urban desires and that's the little write-up so as in Wired Magazine
762. 1:56:44and it's Richard Dawkins which is cool too so that's that and then if you know the uh if you know the dude
763. 1:56:54uh Jason kalkanis the the investor dude he did this uh magazine in New York called silicon alley reporter and that's
764. 1:57:02my uh partner Chan Sue and I from agency.com was the agency we started and we were profitable so it was called men in the black get it
765. 1:57:13it's a profit joke so yeah that's that's the 19 devotees we're down to 18 now can I can can I suggest hardware specs such as
766. 1:57:27CPU side SSD memory for best results using AI it really depends on what you you want to do the the open source
767. 1:57:33models are getting really efficient if you're on the Mac I would say like with Max in general always get the latest one
768. 1:57:41and get the highest highest one you can the new the new M2 Ultra which I think that's like 40 Grand so you don't need to get that
769. 1:57:48um but but like anything Apple silicon there's a lot of developers developing stuff for Apple silicon if you're on PC
770. 1:57:55you basically need a good um GPU a good graphics card like a 30 or 30 90 or higher but that's if you want to
771. 1:58:02actually do the training if you just want to install large language models and use them you can probably get away with something
772. 1:58:09lesser as these open source models get smaller and smaller so I'd say give it six months and you won't you won't need
773. 1:58:15a really special machine you'll even be able to run it on phones thanks for hanging with it's been every day lately cool all right
774. 1:58:22um I'm out of here I gotta go do some mowing I got to get ready for dinner tonight it's been fun hanging with you I
775. 1:58:28appreciate the questions I appreciate I appreciate the hardcore 20 that are left all right we'll talk soon bye