aiLL 7/9/23 - 1 - Unlocking the Power of Generative AI

Transcript

1. 0:05foreign here with the AI learning lab welcome
2. 0:20happy Sunday everybody we're getting started we gonna have a party we're going to talk about artificial intelligence
3. 0:35Is it artificial is it intelligent what is it really no the topic is generative AI hey J word what's happening
4. 0:47welcome everybody my name is Kyle Shannon I'm an entrepreneur and a big mouth [Laughter] blabbering on about generative AI
5. 1:00best way to get started learning AI all right so let's let's do this best way to learn an AI the generative AI stuff if
6. 1:10you want to learn the machine learning and the engineering side of it that's a whole different conversation but if
7. 1:16you're like I keep hearing about AI what the hell is it how do I use it in my work go to one of these three URLs the top
8. 1:22one is the official chat GPT website if you haven't been there if you haven't played with chat gbt go do it go do it
9. 1:29now go do it now go do it now um let's see let me do this so you can see all those URLs um the next one down bing.com that is
10. 1:42Microsoft's search engine but if you click on the chat button it's gpt4 connected to the internet and you can
11. 1:49also generate images there and it's all free so that's that one and then po.com there's six different models you can
12. 1:55play with there from three different companies some of them are subscription only you have to you have to have a post subscription
13. 2:03um but but you can play with the with three of them for free and then pro.com also has the all these Bots these these
14. 2:10GPT powered Bots that are that are actually really cool um so all three of those are worth playing with so that's where I'd start
15. 2:16and then basically go to that URL prompts.chat and um it will teach you how to use these things so it'll give you a primer on
16. 2:27what prompting is and and kind of you know what it means and then it gives you a whole long list of these personas that
17. 2:33you can tell chat GPT to act like act like an accountant act like a screenwriter act like a psychologist act
18. 2:39like a mathematician all that sort of stuff so powerful powerful stuff and then futurepedia.io that is just a giant list of tools
19. 2:49all right so that's how I would get started if I were you do I do you have access to code interpreter yes I do and
20. 2:57I can show that in a in a bit here let me check out some other questions looks like we have a bunch of people here hold for a second okay
21. 3:12uh welcome everybody my name is Kyle Shannon this is the AI learning lab and we are learning AI well I'm trying I'm
22. 3:19doing my best to teach what I know uh I'm I'm relatively new to it I've been kind of deep diving in this stuff for
23. 3:25about a year and you know obsessive compulsive for about six months since chat GPT came out I think it's going to
24. 3:34change everything and uh I've I've started 12 or 15 companies in my day but one of the companies I started that I
25. 3:40think has some relevance here is I started one of the early digital agencies in the mid 90s and so the early
26. 3:46days of the World Wide Web the parallels to what's going on with generative AI right now is just insane so um so anyway yeah let me
27. 3:56go look at some questions here pop some questions if you've got questions about AI pop them down below and I will do
28. 4:02what I can to answer them hey professor race car fever did you watch the F1 race today I mean verstappen verstappen but Lando P2
29. 4:17Hamilton P3 in England I love Silverstone what a track it's so cool that those two English guys won all right turn on the air conditioner
30. 4:32it's a it's a little toasty in here open the window get some things going how much is the best version of chat GPT well there there's
31. 4:44there's um you can do Standalone servers for Microsoft starting at twenty nine thousand dollars and I think they go up to like a million
32. 4:51so so I think that's the most expensive but the uh chat.openai.com just regular chat EBT if you want to get chat GPT
33. 5:00plus that includes code interpreter and it includes plugins and it will include connection to the internet again
34. 5:09once they fix whatever was wrong with it um that's 20 bucks a month and I'm not affiliated with it so if you go pay for
35. 5:17it I don't see any of that money I don't really give a [ __ ] if you do or not but with code interpreter I think it's worth
36. 5:24paying for and it's 20 bucks all right do you have access to code interpreter yes I do that blows my mind it can do that I
37. 5:35don't know what can do what can it do the vocal form as well I'm not sure what you're talking about hit that
38. 5:43like button yeah feel free to hit the like button share this live just so we can get some some folks in here we got a
39. 5:48good a good crop of folks in here I'm starting a little early tonight I'll probably just do like an hour or so and
40. 5:53then have dinner and then come back and do do the late night party we'll be partying I'm coming to you live from Denver Colorado
41. 6:01Saturday Saturday nope it's Sunday all right [Music] professional singer oh yeah
42. 6:18good Lord are you watching like all these tick tocks of the Taylor Swift concerts it's [ __ ] insanity
43. 6:25it's Insanity I mean she's amazing and and then like 80 000 people [ __ ] losing their mind simultaneously for four hours good Lord she is a
44. 6:41Juggernaut it's amazing it's cool all right let's see can chat GPT take over a vocal form as well AI helps Ai and us huh
45. 6:53not sure what that means but sorta yeah nope the Astros oh you watch the Astros today yeah sorry I was watching F1 this
46. 7:01morning and then I watched uh there was some good tennis on today what does generative mean okay so chat GPT
47. 7:12stands for generative pre-trained Transformer and then you hear this thing generative AI so here I'll show you what it so
48. 7:20generative AI so so GPT stands for generative pre-trained Transformer the pre-trained part means it's been
49. 7:31pre-trained you don't have to bring your own data time was you did with machine learning and chat GPT and these other large
50. 7:39language models have been trained on essentially the public internet so it kind of knows think think of like all of
51. 7:46the crap that's been uploaded to the kind of nose think think of like all of the crap that's been uploaded to the
52. 7:53world wide web for the past I don't know 30 40 maybe 50 years um that's what it's been trained on so it knows all the [ __ ] knows all the
53. 8:04languages knows all the [ __ ] knows all the programming knows all the information knows all the misinformation knows everything
54. 8:11generative means it's generating something from nothing so the way it actually works is if you've ever used
55. 8:16Google and you you're like typing a sentence and it kind of finishes like the last three words of a sentence for you like in Gray
56. 8:23that that sentence completion thing that's what chat GPT is but like on steroids okay so so it basically instead
57. 8:30of just knowing you know some grammar and some words it knows all of the internet and then instead of just
58. 8:35filling in two or three words it will write you you know massive documents so the generative part is it's generating
59. 8:42right so here we go so I'm going to say um right a uh short story about a guy who finds a dollar and passes it on and ends up
60. 8:58being a millionaire by the end of the day see what it's doing there all that writing that's the generative
61. 9:13part it's generating that story it's actually a pretty long one um Once Upon a Time in the bustling city of
62. 9:24metropolis there lived a young man named Alex he was an ordinary guy with Big Dreams but limited beans Alex works as a
63. 9:31delivery driver barely making ends meet however you're not good blah blah blah and goes on right so it writes it writes that whole story
64. 9:39um but it but it can do other things as well so I could say um make this an epic poem like The Iliad
65. 9:58and so now it's writing it as a poem big long one not that long let's see sing Muse of a man named Alex
66. 10:07whose fate was changed by a dollar found by chance so great in Metropolis where dreams did store a humble delivery driver he of lore
67. 10:17right then we could say um make a table comparing The Iliad with
68. 10:32Curious George which no one in their right mind would ever do why would you do that but it can do it right because it's been trained on
69. 10:41all the [ __ ] so genre epic poem genre children's literature author Homer author H.A Ray right I could say um write me uh a python
70. 10:56application that generates an epic poem about curious
71. 11:10primates uh uh after ask wait let's see after asking a user for a topic
72. 11:28certainly here's a python application right and so so this is the generative part and and this is Lang you know chat GPT does it
73. 11:43with language but we can we can go over to something like Discord and we can hop into a mid-journey channel in Discord
74. 11:56and it also does it with images so here are some someone asked me to do um a logo of a flower shop the other day
75. 12:06and I I just basically described it in a few words and it generated this stuff generated generative from nothing
76. 12:14and these are not copy and pasted these are generated out of randomly noised images that get denoised into this kind of magical thing so
77. 12:27hopefully that answers it um we can go much deeper that that's really what this channel is about is talking about all that stuff and uh yeah
78. 12:37so there you go yeah the Silverstone the F1 race was good today well AI robots robots become sentient sentient or evil
79. 12:53um I don't know I well I I think over time they're going to be sentient I there's a lot of people working right
80. 12:59now on what's called alignment um in fact um Sam Altman uh open AI just said that they're gonna they're gonna dedicate uh
81. 13:1220 of their compute resources to what they're calling super alignment and Alignment basically means are the robots
82. 13:19aligned with human values or human um rights might be a better way to put it and if they are then as these things get
83. 13:32more powerful that's good for us and if they're not as these things get more powerful that's not as good for us
84. 13:40because if they decide that they're more important than we are then bad things could happen but if we get the alignment piece right
85. 13:48then we'll be okay now because there's a lot of Open Source large language models out in the world right now that are not being
86. 13:59um controlled in any way by companies like Google or open AI or Microsoft there's a there's a there's a there's a higher
87. 14:10risk of Mal alignment than there was if it was with big companies but if you have deep-seated distrust of big companies then
88. 14:21that might be a risk worth worth giving up it's it's a really complicated subject because these things are very very powerful
89. 14:28I signed up with chat AI on my phone and now I can't sign up at my desktop it does not authenticate did you go to chat.openai.com on your
90. 14:37phone or did you get the official chat GPT app it it should authenticate um if it could just be a browser cache
91. 14:47issue so try opening it with a uh incognito window or try it in a different browser um I I'm always a little skeptical if
92. 14:57you signed up with chat AI the way you wrote it there's a lot of scam apps out there so if you're not using the
93. 15:03official chat GPT app or if you didn't go if you didn't go to a browser and go to chat.openai.com if you want any
94. 15:11anywhere other than that or the official chat GPT app you might have gotten a shitty knockoff thing that's charging you money
95. 15:20that is not actually that which is why I have these here I don't make any money off these I just have these here so people to go where do I
96. 15:28start why do I do I heard about this AI thing I heard it's gonna kill my job and then kill my family is that true no
97. 15:38go there start playing with it the more you know the better it is I want to know if chat GPT can impersonate
98. 15:45others that would be not not good yes it can tell it to impersonate Shakespeare and it will tell it to impersonate you know
99. 15:55now now can it impersonate them like like a video that's a different thing but that all those generative AI stuff is being
100. 16:04applied to everything from images to audio to voices to music to animation to film to pictures to words and so yes it
101. 16:15can all mimic other things and we're gonna get a giant steaming lesson in the 2024 election about what generative AI can do some someone
102. 16:28earlier asked what is generative we're gonna get a real good lesson starting in probably um a month or two two months November
103. 16:41we'll have a good solid year of it and and you'll see lots and lots of uh impersonation stuff going on and then I think after the election
104. 16:49we'll realize uh we should probably get this under control and we'll start to build technologies that make it harder
105. 16:56to do that based on current landscape where do you see opportunities to generate Revenue using these tools um
106. 17:03short term just if you get AI literate and start talking about it you know one good way to generate revenue is um as more and more companies
107. 17:14start to get curious or desperate about needing to learn this stuff they're going to be looking for people that know it so
108. 17:23um so start learning these tools and start talking about it talk about it on LinkedIn talk about it within your
109. 17:28company talk about it amongst your friends even if they think you're [ __ ] weirdos everyone in my life thinks I'm a [ __ ] weirdo doesn't
110. 17:34matter this is profound it's going to change everything so start getting educated about it start understanding how to make
111. 17:41this stuff do what you want it to do and then tell other people about that and you'll start to discover places
112. 17:51where you can make Revenue it could be Consulting it could be whatever a YouTube channel it could be um doing Services starting an agency
113. 18:00doing services for other people that's probably the quickest way to do it hey you want you want me to build an
114. 18:06automation for you I build zapier automations you could build an entire business around zapier automations
115. 18:11because pointing zapier automations to large language model tools like chat GPT are insane like what you can build is [ __ ] nuts
116. 18:23um if you're within an organization right now if you have a job right now within an organization start learning
117. 18:29this stuff start talking about it and some people may be afraid some people may be like well it's against our policy
118. 18:35to use all that just keep talking about it because at some point someone's going to say um we need to figure this out who knows about this
119. 18:43oh Kemp knows about this yeah Kent's always talking about it he won't shut up about it well get Ken in here I gotta I gotta talk to Kim
120. 18:52what's going on to say I can't be the guy they ask that's a way and then and then the beyond that you know I would sort of put
121. 19:02it back on you to say what are you good at are you a project manager are you a creative are you a writer are you
122. 19:11someone who's really good at mechanical process what I would say is as you get up to speed on what these tools are and what
123. 19:20they can do start to figure out how you can use these tools to do what you already do better and then in that you'll probably find a
124. 19:29lane where you can go oh I have an idea for a business it's the thing I do but like times 10 times cheaper 10 times faster 10 times better
125. 19:40that's what these tools can do so I'd sort of put it back on you to say what are you good at because these are just
126. 19:45tools like making money with AI is no different than how do you make money with Photoshop I don't know get good at
127. 19:52photoshopping [ __ ] and make people pictures good enough that they're willing to pay you for them same thing
128. 19:58with this get good enough with this stuff that you can solve the problem for someone good enough for them to pay you for it
129. 20:04I know that's not a great answer but it's the answer because what's not the answer is that's not the answer
130. 20:13don't listen to those people they're full of [ __ ] you know how they're making money with chat GPT getting money from people like you
131. 20:20who want to make money and they're just making you know stupid little courses that they're not actually building anything
132. 20:30um got Lang chain and pine cone hang on Tick Tock just jumped me to the bottom again okay God dang it geez you can't
133. 20:39bring them freaking wreck them suck them yes at some point tick tock's gonna fix this problem and I'm not gonna know how to
134. 20:47handle myself got Lang chain and Pinecone working have it running in collab drop PDFs on my hard drive and it
135. 20:53Imports it in I'm jealous jugged gear Gear I haven't done that I haven't gotten it up and running yet I know it's
136. 21:00probably it's probably only a two-hour commitment of focus time but it's like it's just such an area that I'm like so
137. 21:06clueless in that and there's like all the apis that I gotta hook up the API keys so that's awesome that you're doing
138. 21:13that and and I and listen I think that I think that if you can if you're geeky enough to get Lang chain up and running
139. 21:19you should be doing it because I think that's where these things very quickly head is you know take your data and
140. 21:28point these these uh machines and robots and chat Bots at your own data in fact the code interpreter version of
141. 21:39chat GPT you can upload like 100 megabyte file to it you can take a big chunk of your data upload it to this thing and start
142. 21:47interacting with it it's pretty amazing awesome calculations that's fantastic I'm currently learning python do you
143. 21:58think it's worth it to pay for it wait is it worth to pay it um is it oh wait I'm current do I think it's worth it to pay for chat GPT
144. 22:11um Plus or to pay for a course in Python um GPT 3.5 the free version of chat apt will will help you do coding
145. 22:22um bing.com is gpt4 for free that'll help you and then I would say go to github's copilot and that's a pair programming
146. 22:31it's a GPT tools specifically designed for for coding so if you want to use these large language models to support you in learning python um
147. 22:47there's free options for you and and what I would say is get get good at learning to program in conjunction with these tools and and
148. 22:59quite frankly let me go show you code interpreter what it's doing because this if you haven't seen what code
149. 23:05interpreter is doing it'll blight [ __ ] mind Okay so if you've never been to chat gbt this is chat GPT if you pay for chat GPT plus
150. 23:21um you get GPT 3.5 and gpt4 so for gpt4 you get regular gpt4 you get this new thing called code interpreter
151. 23:30and then you also get plug-ins there's a fourth mode which is browse with Bing but they disabled that four days ago
152. 23:36because it was downloading content from behind paywalls and apparently some Publishers got pissed off so that's not
153. 23:42there for now it'll come back um okay what code interpreter allows you to do is upload files up to 100 megabytes in size
154. 23:54and then it starts to do things with them so I'm going to say plus here for upload and then let's see
155. 24:08oh this is good okay so this is a ZIP file and in this ZIP file is uh a CSV file from IMDb about television shows so I'm just going to say
156. 24:24um please tell me what is interesting about this data and visualize visualize
157. 24:40it in ways that help me under stand it okay so it's a zip file so it's a compressed file I write one little sentence about it
158. 24:57all right let's start by unzipping the file and examining its content so it understood that it was a zip file and
159. 25:03then it ran a little plug-in but now let's go look at what that plugin actually does or what it's doing it's writing python
160. 25:14it's writing its own python so it Imports two libraries zip file and Os and then it um then it opens let's see there's a bunch of CSV files
161. 25:30in there I guess one for each category it looks like the data consists of multiple CSV files each representing a
162. 25:38different genre of TV series such as mystery sport biography and so on and then it writes another um
163. 25:48python script and executes it to go look in those CSV files so each one of those has the title the ID the release year
164. 25:57the genre the cast The synopsis the rating the runtime the certificate the number of votes then let's see what's it do also let's
165. 26:05load all the other data files and concatenate them into a single data frame for the complete analysis so it's
166. 26:11going to string all that data together and then it's gonna it's now coming up with um things we could look at cleaning the
167. 26:21year the release year column converting the runtime into a numeric column seems to be there there's an error it's going on and look
168. 26:33now it wrote python to generate a script this is a distribution of ratings across all of the TV series so it looks like about
169. 26:457.4 is the sort of highest very few tens very few ones kind of you know leans into the sevens um what else what else did it do for us
170. 26:57this is number of TV series in each genre so romance is the most westerns are the least here's National number of TV series for
171. 27:08each content so what code interpreters doing no it's still going it gave itself instructions based on my
172. 27:16little thing about what's what is this data tell me what this is and visualize it for me and it did all this and it wrote The
173. 27:25Python for you so it's is it a is it a bad thing to Learn Python Programming no it's not but you need to understand that what's coming
174. 27:34are tools that can write it and execute it and so if you can get good at um instructing these tools what to do
175. 27:42where you're the pilot and it's your co-pilot and it's doing these things for you quickly that's great but anyway if
176. 27:50you haven't seen code interpreter that's what it does it's nuts it's nuts it's nuts Italia nuts um here we come to save the day
177. 28:06is chat GPT used useful for visual art no but mid-journey is and uh it so you can also if you go to bing.com
178. 28:18let's go there let's go there quick that's Poe I'm gonna go to bing.com
179. 28:35rubber ducky you're the one you make bath time lots of fun that was from Sesame Street I want an original poem
180. 28:49don't give me some stinking copyrighted work can you make an image to go with that
181. 29:03poem poem and see if it does it I'll try to create that
182. 29:17and so now it's off it it probably generated some prompts based on that poem and we should see some rubber duckies here boom there you go so
183. 29:32this is not chat GPT this is Dolly but there's Dolly there's stable diffusion there's mid Journey so mid journey is
184. 29:40the highest quality stable diffusion is the most flexible but it requires the most geekiness dolly is now incorporated into Bing so
185. 29:49you can just use it just like I did there you can you know make uh uh an image of a 70s muscle car on a mountain overlooking
186. 30:07a valley and it'll go do that and then there's also um Adobe Firefly or generative fill within the new Adobe Photoshop beta
187. 30:22all of those are image generation tools that can help with the art side of this and then if you go to so there's our 70s
188. 30:29muscle car overlooking a valley I mean come on like like what did that take 30 seconds like like if you needed to do something
189. 30:40for a PowerPoint presentation or a or whatever and it didn't need to be an accurate actual 70s muscle car because that's
190. 30:49some sort of weird ass hybrid but I mean it's pretty nuts it's party knots it's party nuts if you know what I mean
191. 31:02I mean that's Purdy that's party Mark here right come on come on this is why this channel exists
192. 31:19[Music] do me a favor hit the double tap the screen do do the likes so we can let people know we're here
193. 31:29we'll get as many people in here as we can have something stuck in my teeth drive me crazy oh got it wow I'm a little busy so I guess my
194. 31:39questions didn't make much sense sorry that's okay if you ask them more clearly I'll I'll ask him down with the public internet
195. 31:49you mean you are down with it or you don't like the public internet you cannot like it but it's sort of the the infrastructure that
196. 31:58we run everything on now debating on having it saved my past conversations in the vector database I judge I think that's actually
197. 32:09I think so so first of all I think that um I think Google's gonna do that I think Microsoft will probably do that I think
198. 32:19Apple's gonna do that I think Apple's gonna do it really well um I think if nothing else it would be a brilliant brilliant
199. 32:29um exercise to learn on because I think I think having like just take your entire social graph right like your history of Facebook your
200. 32:40history of LinkedIn your all that sort of stuff and have that in a vector database where you could now say
201. 32:48um I want you to write a blog post in my voice and it actually can understand what that actually means
202. 32:53I think that's and and understand like your point of view on on everything I I think it's I think it's a brilliant brilliant idea I think
203. 33:01it's a really really smart idea judged really good like yes like like I don't know if you're asking for my
204. 33:08encouragement but you're getting it um chat EBT also I have to say is I like it it's still excerpt from success stories
205. 33:26it's stole excerpt from success stories what did what's the it in that sentence and which success stories from because the the because it it
206. 33:37sounds like the um the straw that broke the camel's back for Chachi PT connected to the internet why they took it down was it it was pulling
207. 33:50some Fortune Magazine content and I would assume timing came to them and said um we're Tire man can you better knock this off and
208. 34:00Microsoft probably said yeah they're our biggest customer one of our biggest customers so you got to deal with this nobody reads anymore
209. 34:10much more consumables not quite sure what that means I write articles but nobody has time to read them anymore I
210. 34:17know 10 words people will read headlines only well and that's that's again that's a place where um where this generative AI stuff could be
211. 34:27really interesting I I could imagine doing a um so imagine doing an automation that takes your long-form content
212. 34:37and distills it down into increasingly smaller distillations right so so you know a five-word headline is
213. 34:48the smallest and then a 10 word headline and then a you know 140 character tweet and then a you know a short little you
214. 34:57know Punchy Facebook post that goes with an image and then a LinkedIn post and then a medium article and then a long form article
215. 35:05and then you could Auto publish those to all of the places and they all cross reference each other or point to the long form article
216. 35:15or you can say you want to dig a Little Deeper here's the next layer down you want to dig a Little Deeper here's the next layer down
217. 35:21that could be a product lately on this channel I've been like taking comments like that I'm like okay here's the product
218. 35:27so so generate multiple Links of the same article automatically publish them to appropriate channels Point them all
219. 35:34to the long form article I don't know if you can make any money on it but if you built a tool that did that
220. 35:41you could probably license that to the time inks of the world and the economists of the world and things like that where where they can just
221. 35:48automatically take their long-form content and Auto generate videos Auto generate animations Auto generate short form
222. 35:59content Auto publish it everywhere track all the engagement maybe you find out that people do read long form articles but they have to read
223. 36:08the short headline first I don't know maybe you confirm your suspicion that nobody reads anymore I don't know man there's just there's
224. 36:17[ __ ] to do man you know what I'm saying man you go learn this [ __ ] and your brain kind of goes yeah man and you [ __ ] you know you
225. 36:28[ __ ] eat a shroom man and then go there and then just you'll see the future man it's right there you know man
226. 36:41this generated stuff have a copyright um good question the copyright office isn't quite sure yet the copyright office is basically
227. 36:51saying if a machine did it [ __ ] no now they don't say that about cameras do they so they haven't quite figured it out yet and there's also
228. 37:01so so there's lots and lots of gray areas GPT three point in all these different things have all been trained on copywritten data
229. 37:16but one of the misconceptions is is that when you generate some writing or when you when you generate something in
230. 37:22mid-journey it's copying and pasting from that copywritten work it's not doing that it's generating new original work
231. 37:29so there's one argument that says well if I went to art school for four years and I learned about Andy Warhol and I got really inspired by
232. 37:37him and I left art school and I started my art career and started doing work that looked like Andy Warhols that's my
233. 37:43work inspired by that knowledge so there's one school of thought that says generative AI is just like that there's
234. 37:50another school of thought thought that says you're stealing my [ __ ] work stop it I'm not getting paid and and they're both valid and like
235. 37:59everything in this polarized shitty ass world right now the reality is somewhere in the middle there are there are things
236. 38:06that are generated from these models that absolutely are infringing on someone else's work there are other things that are
237. 38:15borderline and there are other things where it's not even close um it is unclear right now what's copyrightable it is unclear right now
238. 38:25what if anything that you generate could cause copyright issues um in the future so if you're talking about images I
239. 38:34would and you're in a in a in an organization that is concerned about these things I I would use Adobe Firefly
240. 38:42use the Adobe products because their models have been trained on public domain and Adobe stock images that they've got all the rights to
241. 38:51so Adobe is taking the most conservative approach to this on the image side on the word side I like like I
242. 39:03it was trained up on all the stuff on the internet some of that is copywritten work but when I have a generate an original
243. 39:12poem for me the fact that it sort of read you know all of I don't know some Modern authors Tom Clancy's books
244. 39:20and might have included something that kind of sounded like him in my in my little paragraph here I don't know how
245. 39:27you make that connection necessarily so so I don't know it's a very gray area um My Philosophy right now is if I'm doing
246. 39:37anything for clients I'm taking a very conservative approach if I'm doing anything for myself I don't really give a [ __ ]
247. 39:43but I'm certainly not going to a place we're moving into a world of infinite content generation infinite everyone can make beautiful images
248. 39:52everyone can write great infinite content generation so if you generate something or if you collaborate with AI and write something you're
249. 40:01really proud of and you want to copyright that you'll likely be able to find a way to do it if you can prove that you
250. 40:12materially contributed to the generation of that but we're also moving into a world where there's kind of an infinite amount of
251. 40:19that stuff out there so does the model completely change over the next 10 years my instinct is that it
252. 40:28does what does it look like I don't have a [ __ ] clue um alignment is only as good as the humans aligning it and open AI isn't the
253. 40:37only player in town and the open source models aren't aligning [ __ ] well some are but some are like you know Screw the
254. 40:45safety guidelines let's put this out in the world no they're not the only ones Google's doing this although one of the reasons
255. 40:53that Elon Musk started open AI was he didn't like Sergey brin's Cavalier attitude about AI and so he said let's start something
256. 41:02where we pay attention to safety and now he doesn't like how Sam Altman's doing it so he's going and starting his own AI company called xai
257. 41:11and then you got meta who's open sourcing all their stuff and then you've got AWS Amazon has started their own thing Apple's gonna
258. 41:21have their own thing so you can have you're gonna have all sorts of players in the game but yeah I agree good evening Troy
259. 41:28hi Danny hi Ai guy will robot dolls be spouses um Mr Shannon you're early I know I came on early I I so we we uh we're having
260. 41:41ribs tonight because you know why not but we put them in late so they're not going to be ready to put
261. 41:47on the grill till seven so I figured I'd do like an hour or so here and then I'll come back later [Music] um
262. 41:54the the will robot dolls be spouses I think that that um that virtual companionship is going to be a very real thing and I I think if you
263. 42:06if you there's there's sort of a like the immediate reaction to that is like ah it's creepy um there's a lot of lonely people out
264. 42:13there and and these tools are really good at conversation and they're going to be increasingly good at conversation
265. 42:19um and so yeah so I think virtual companionship is going to be a thing and it and it might not just be companionship like love companionship
266. 42:25might be like friend colleague advisor um I think we will have multiple agents that we interact with and some of those
267. 42:34may be more conversational and I want to deal with some personal [ __ ] and this happens to be something that's
268. 42:40incredibly good at Psychotherapy and it's incredibly good at empathy and listening and as these as these tools get
269. 42:49increasingly multimodal where it's not just typing in or that weird thing where you talk into it and there's this really
270. 42:56long pause and then it responds back with this stilted thing probably within two years you'll be able to talk to a something
271. 43:05maybe it's got a face maybe it doesn't where the response will be immediate and it will be indistinguishable from you
272. 43:13talking to another person so I think I think and because people are attention [ __ ] someone will go first someone's gonna
273. 43:23marry an AI bot or an AI bot connected to a physical robot someone's gonna do that again probably within two years
274. 43:31just to get the headline and then I think it'll be it'll become a whole industry just like you know online
275. 43:37dating where online dating was weird and creepy for a while I think this will be similar oh and I think it'll help a lot of people
276. 43:46believe it or not and some people it will do creepy [ __ ] with it but that's sure I think of these tools as Humanity
277. 43:56amplifiers and it amplifies the good and the weird so you can be a world-class weirdo with AI [Laughter]
278. 44:07Young video game nephew kid asked me what's the reason to read a book listen to him
279. 44:22um I mean if if you have an answer you talk to him about it right um but I would say learn from him what's his experience with books maybe
280. 44:33he hasn't read a good one maybe he's got a way of consuming what used to be books in a way that you can't even imagine
281. 44:41maybe he's just clueless I don't know learn learn from him I I love the idea of reverse mentorship right where older people are mentoring
282. 44:50younger people and vice versa like you know I'm a 58 year old with a tick tock Channel like I could have a young person teach me to twerk
283. 45:01no all right can a I write better scripts and is this why they are striking um it can't write better scripts I so I've
284. 45:11I've written seven feature-length screenplays in my day um what AI can do remarkably well is get you to 80 percent like that
285. 45:25so it could it could write the outline of a screenplay incredibly quickly and just by you know giving it feedback like
286. 45:37going back and forth probably 10 iterations with an outline of a screenplay you could probably have a pretty world-class outline immediately
287. 45:46and then you could have the AI you know start to block in like like how I would approach it is this I would say
288. 45:53I I would have it write me an outline I'd say okay I've got this idea for for a screenplay and it's kind of like these
289. 45:59movies and I want to use the save the cat story framework or I want to use Sid field or I want to use whatever
290. 46:06framework you want Joseph Campbell Heroes Journey it doesn't matter I want this to I want this to mimic um the structure of Star Wars
291. 46:16identically why not isn't that what JK Rowling did with Harry Potter um or one of them where I don't know anyway like like
292. 46:29you know story structure story structure and a hero's journey as a hero's journey right so uh so so I'd start with that I
293. 46:37would then tell it you know my characters I would I would have it flesh out my character personas um
294. 46:46and then I would have it right scene one maybe give it some ideas for scene one um but I've already got it in the
295. 46:55outline I already know what the major beats are for scene one I would have it right scene one I might give it
296. 47:04um example dialogue from another screenplay that I like the tone of maybe one that I wrote maybe one that I
297. 47:09didn't and say kind of match this tone and then I would sit there and I would I would take whatever it generated from
298. 47:16scene one and I I would go through and edit it so I would sort of flip into an editor's mode then I would say Okay
299. 47:22based on this outline and this scene one write me scene two have it do that wait based on based on this outline these character personas
300. 47:32and this scene right we've seen two and I would just sort of knock it out like that so so it can't write it wholesale
301. 47:39but I think I could probably knock out a feature length screenplay if I did it you know four to six hours a day
302. 47:49like even after work and then two days on a weekend I bet you could knock out a solid feature-length screenplay in a week
303. 47:59is the app available on Android Okay so no well chat gbt's not Chachi PT has an official IOS app but there are
304. 48:11lots and lots of fake apps so be really careful with it I would go to openai.com and go to their products section and go
305. 48:19to chat gbt and you'll see a link to the App Store so if you're going to download the app for for chat GPT make sure it's
306. 48:27the official one what I do is I just point my um my my mobile browser Chrome Chrome or Safari whichever one you use to
307. 48:37chat.openai.com and then make a shortcut on my desktop for it because the website's incredibly simple it's a
308. 48:44single input box and you type into it and [ __ ] comes out and then on the phone you can hit the little microphone button
309. 48:50and talk into it and it'll it'll respond so um so that's how I do it I I I downloaded the official chat GPT app but
310. 48:59I don't think it's as good it was it was good when it was connected to the internet but they killed that because of
311. 49:05the Fortune Magazine thing um what has brought what has AI brought to the table in terms of innovation such as new scientific methods well
312. 49:28you know it's funny it's it it's like what the what the large language models do technically we've seen before like we see it in our current
313. 49:38you know Google sentence completion all it's really doing is sentence completion on steroids so so from that piece of the technical
314. 49:48perspective it's like what it's what it's doing is pretty simple now what's different about it technically
315. 49:56Google invented this thing called the Transformer that allowed these machine learning models to be trained on massive
316. 50:02data sets and so they trained these large line these machine learning models on all of the public internet and that's why they know so much
317. 50:12um so the scale of knowledge that it has if you think about what Google is Google's an indexing platform where
318. 50:21it learns all of the internet but it really just learns it so that it knows which of the 40 websites it might send
319. 50:29you to are the most relevant to what you're looking for so it's indexing you know a list of links what large language models are doing is
320. 50:39they're learning all the stuff and they're learning all the stuff so when I ask it um to explain to me who Plato was and
321. 50:47what his philosophical theories were it can just answer that it doesn't take me to 40 different sites about Plato where
322. 50:54I have to go figure it out um and it's so good at stuff like coding and things like that where it's actually coding right so so one of the
323. 51:04advancements here is that and this was my big Epiphany about chat gbt was when I first started playing with it I'm
324. 51:14like oh this is kind of like Google you type in a thing and it gives you a response and then I had it write me a
325. 51:20poem and I'm like oh but and when I use chat GPT for when it came out I'm like oh that's much better like it was good
326. 51:26and writing songs and so like it's it's a little bit different the thing that sort of put me over the edge where I
327. 51:32thought holy [ __ ] the world's just changed is I thought out I should I hear python something good to learn I should I heard
328. 51:42it can do code this is back in December and I said you know write me some python code that takes an input and it did that
329. 51:50and then I realized I don't know how to run python code so I said hey chat epd how do you run python code and said oh
330. 51:56you you go to the the terminal program on on your Mac and you type in this command line that I can install this
331. 52:03library and I'm like no I I don't want to do any of that isn't there a website I can just go to and paste this code in
332. 52:08it and see if it works and it said sure and it gave me a list and the first one was replit and I went to replit
333. 52:14and so I ca I copied and pasted my code into reflet I hit the Run button and it gave me an error and then I said hatch
334. 52:20EBT here's the error game he said well you don't have a library installed and I said how do you install the library and
335. 52:25said you click on this button and then you search for this Library name and then you hit the install button and then
336. 52:30you hit play and it should work and it did within 90 minutes I had written a sophisticated python application that
337. 52:38asked someone for a topic and then asked them three questions about it and generated five pieces of social media
338. 52:43content in 90 minutes I've never written a line of python in my life I still haven't and I had a functional python application so
339. 52:56the the way I think about what what these large language models do is they shift us from what Google did for us and what the and
340. 53:07what the world wide web and the internet did ostensibly is we could have information on demand so we didn't have
341. 53:13to go to the physical library and dig through the Dewey Decimal System with the cards and go find the book and check
342. 53:19out the book and get yelled at by the library lady because we didn't return our last book we could just go on this internet thing
343. 53:26and find the information and Google got really good at helping us find the information they turned it into a you
344. 53:31know trillion dollar business or however big they are what these large language models do is they give us knowledge on demand right I
345. 53:41don't have to go learn how to do Python Programming now if I go to Learn Python Programming will that help me use these tools better it will
346. 53:51but me never having written a line of python I wrote an application without doing that so I had the knowledge of a programmer in this tool
347. 54:02so so for me that's that's the big difference is that we're shifting from you know the information age to the
348. 54:09knowledge age or information on demand to knowledge on demand um and and there's lots and lots of other things that it does it's
349. 54:15accelerating things like scientific discovery um there there's a group within Google called deepmind that created this thing
350. 54:25called Alpha fold that predicted the um the the the shapes of of proteins that historically would take a long time to
351. 54:37to figure out just one it did like 400 000 of them in I don't know a few days or a few weeks some ridiculous amount of
352. 54:44thing so the amount of drug Discovery that's going to come out of just that one project and it's it's kind of like
353. 54:50anywhere you point these things it it's solving things much more quickly and and those discoveries and Innovations are
354. 54:59stacking one another and they're accelerating exponentially so so we're entering what I'm calling the great
355. 55:05Renaissance where we're going to see increased creativity increase scientific discovery increased productivity
356. 55:14um you know in almost every sector imaginable uh uh me too they think I'm crazy for being interested in AI I know whatever let
357. 55:27them be great because what will happen is they'll make fun of you for a while or in my case like they just ignore me
358. 55:31you know I just noticed people responded to me initially I I had one when I started doing the the the project that
359. 55:38that weird eyeball there I did this project called Kyle Shannon dreams which was all these AI generated self-portrait
360. 55:45kind of things I had two different people reach out to me and ask if I was mentally okay foreign no I'm not but you know
361. 55:56but I'm doing that [ __ ] on purpose would an AI president be better than Trump or Biden maybe this is a good thing um
362. 56:09the thing that I notice about asking chat gbt about I will ask Chachi PT about topics I wouldn't touch with a 10-foot pole with
363. 56:20human beings right because there's so many trigger words like if you ask about trans or this there's guns or whatever
364. 56:26like there's just there's like like the English language is a [ __ ] Minefield anymore right like the past 10 years of
365. 56:33American politics has just destroyed the ability to talk with any Nuance about anything right it's just you know are you on my
366. 56:45team or on the other team because you're on the other team I'm gonna beat you up and sometimes I like I would like to
367. 56:54actually be able to like talk about a topic and not have it immediately go into that emotional place Chachi BT's really good at that
368. 57:04because it doesn't understand trigger words you just ask it a sentence about whatever the topic is you can't talk to people about
369. 57:13and it will give you incredibly nuanced answers so you know would it would it be a good president not right now because
370. 57:20Chachi PT doesn't reason it it really is just a prediction calculator right it's it's predicting the next word and the
371. 57:28next word and the next word it's not actually conscious in any way of what it said where we're headed and where Sam Walton
372. 57:38wants to take open AI is to a thing called AGI artificial general intelligence or super intelligence where
373. 57:45these things are smarter than human beings and reason better than once you start to get into that neighborhood then you can start to have
374. 57:52a conversation like could it be a president president but here's here's the challenge with that if the they call it alignment if the
375. 58:04alignment of the of the model is not in line with human rights you could end up having something with
376. 58:14tremendous power like let's say you give some version of this political power and then it decides it doesn't like the
377. 58:20direction we want to take the country or the world and it says no we're going to take it in another Direction
378. 58:26and that direction is not in line with our rights or our values or what we want well then we're in [ __ ] trouble right
379. 58:33so before you can get anywhere near having an AI president you have to solve the alignment problem and that's that
380. 58:40that's the thing that Sam Altman just said they're committing 20 of their compute resources to what they're calling super
381. 58:47alignment which is basically make sure these [ __ ] don't kill us so that's good it's a good thing but
382. 58:54we're not anywhere close to having you know turning over any of that over now what what you could potentially have is
383. 59:01you could have ai generated tools that accelerate um legislation crafting and the ability so you know how sometimes you know
384. 59:10Congress will play those stupid ass tricks where they'll create a 800 page document and then they say you've got
385. 59:17eight hours to read it and then we're voting on it and they like Ram [ __ ] through the thing and no one can read it
386. 59:22well you could take that 800 page legislative document run it through chat gbt and have it summarize it and have it
387. 59:29summarize the the core pieces of the thing so people could actually understand what they're voting on so I
388. 59:35could see there being tools created that help with the legislative process but you still got human beings involved to
389. 59:41make sure that everything you know everything aligns with what what we're up to AI versus climate change how is this
390. 59:48possible yeah so I think that I think that when it comes to energy um generation when it comes to solving
391. 59:55climate change when it solves to coming things like neurological diseases um you know spinal cord injuries cancer
392. 1:00:02you know complicated diseases I think these AI tools are gonna um dramatically accelerate Solutions in all the spaces
393. 1:00:11um how's it going to do it with climate change I I don't know enough about climate change to know where it would do it but
394. 1:00:18but one of the things these tools are remarkable at is ingesting lots of data processing it and analyzing it so we're
395. 1:00:26going to be able to accelerate um a lot of the the possible solution sets um derived from data the other thing to to realize about all
396. 1:00:37this AI stuff is as new as it seems the the ai's been around for decades right so they're already applying you know AI to
397. 1:00:48climate change into energy and all the different things what what these tools do the generative AI tools let's say sort of chat GPD
398. 1:00:57forward is that it makes this the these tools accessible to more and more people who don't need to be data scientists
399. 1:01:05it's just regular people can start using these tools so I think what you're going to see is many many more people applying
400. 1:01:13machine learning and AI towards solving problems that wouldn't have done it otherwise where where it would have just
401. 1:01:18been often a research corner somewhere you might have organizations or non-profits or even individuals coming
402. 1:01:25up with solutions that they couldn't have done otherwise so I think that's how it will help how specifically within
403. 1:01:30climate change I don't have a flipping clue does AI stand for authentic authentic ice cream stands for actual intelligence I don't know no
404. 1:01:43uh AI is actually not the right term it's machine learning is is more accurate for the stuff that we're
405. 1:01:49playing with but ai's easier to remember and that's what it is in all the movies when I put my text in a prompt it is now
406. 1:02:00owned by chat GPT yeah but you can turn that off let me show you how new chat okay so if you're at chatgpt um
407. 1:02:13you've got your chat history here your prompt history if you go to your settings within settings there's this thing called Data controls
408. 1:02:24so go to data controls and then you can this top one is clear your history and training so this will get rid of all of
409. 1:02:32your prompt history so if you want your prompt history make sure you export it first you can export your data so make
410. 1:02:38sure you do that if you want to save your prompts this is one of the reasons I haven't turned this off yet is because I don't feel like
411. 1:02:47like I want my prompts there I like my prompt history but you can turn off your chat history and you can basically say I
412. 1:02:54don't want you using my data for training so so that's there right now and then open AI said they're coming out
413. 1:02:59with a business version of chat GPT that's going to have robust privacy controls so that that should help
414. 1:03:11love it for back cover copy that's cool the power of all of the stuff that these folks have been working on for the past
415. 1:03:18three four five decades available to all of us and from the day they launched chat GPT until it hit 100 million people was six weeks
416. 1:03:29it went to a hundred million people in six weeks it was the fastest adoption of consumer technology in history
417. 1:03:37um because this [ __ ] goes deep it runs very deep what it makes possible is is insanely profound um so what you're asking is you know did I
418. 1:03:49have any part pre-inflation Point not really I I messed around with it a little bit just like I messed around
419. 1:03:54with the internet in the early days with command line stuff but then when the World Wide Web came out I'm like oh holy
420. 1:04:00[ __ ] this changes everything and I started one of the first digital agencies company called agency.com and I
421. 1:04:05ended up building a lot of the websites for the fortune 500. um so I didn't invent any of the World Wide
422. 1:04:12Web or any of the underlying technology but I used it I applied it to business and so that's what this channel is about
423. 1:04:18and that's what I'm doing with all this stuff that's the inflection point that we're on the other side of now
424. 1:04:24um AI versus climate change we did that one oh Tick Tock just jumped me to the bottom um how are we doing time wise a little bit
425. 1:04:34I just watched 60 minutes talking about AI very cool stuff CBS has stuff all week about AI yeah this is you know there's there's uh
426. 1:04:43there's there's um if you go look at go look at the early days of coverage of the World Wide Web um there's a famous one with Bryant
427. 1:04:55Gumbel and Katie Couric where she's like that little at sign and he goes I think it means about and she goes no I don't
428. 1:05:02think it does and and they're like what is this internet thing anyway it's it's quite a Charming little piece and I like
429. 1:05:10I feel like we're starting to see those same kind of uh things with AI but I think what's clear about AI to to everyone is that this is
430. 1:05:21significantly more powerful and more profound than the World Wide Web was and the World Wide Web changed the [ __ ]
431. 1:05:28world it changed everything changed How We Do Commerce it changed how we communicate it changed how we date changed how we bought things
432. 1:05:36like malls you know are like dinosaurs now because we don't go to malls anymore um it changed everything and so this is going to change
433. 1:05:48everything more profoundly and uh I think the transformation is going to happen way faster I think it's I think
434. 1:05:55it's going to happen faster than we're prepared for which is why I have this channel thanks I just tried it out awesome Esteban that's great
435. 1:06:10I'm gonna fix my credit card oh my credit with this oh this would be great for fixing your credit yeah take have it
436. 1:06:17explained to you how credit uh things work get your reports pop it in there tell it to you know guide you through
437. 1:06:25fixing your credit it's brilliant use of This brilliant use what about Cosmos 2 I don't know it uh I've been able to search user IDs
438. 1:06:35using Google I'm thinking AI has a way too um probably I don't user IDs where like user IDs oh like connecting someone's
439. 1:06:46user ID to who they really are probably like it may have that stuff in it I don't that might be one of those things
440. 1:06:53where they've got a safety guard rail in there where where it's like yeah it looks like you're trying to find a human
441. 1:06:58being we're not going to allow you to do that but you could it'd be something to experiment with for sure and then if
442. 1:07:05um chat gbt won't do it go play with one of the open source models that doesn't have any safety guidelines on it
443. 1:07:11not that I'm encouraging that for research purposes only of course [Laughter] thanks uh I misused it I want to fix my
444. 1:07:25credit with this I think it's a great use case can you give an example of what zapier and chat TPT can do yeah if you go to go to um content
445. 1:07:38contentevolution.net it's an organization I work for I'm actually the chief generative officer there I'm pretty sure I'm the first
446. 1:07:48Chief generative officer anywhere in the world if you go to contentevolution.net on the on the home page is this thing called
447. 1:07:58The Challenge engine and it has you put in your email and like what business sector you're in and then you put in a problem like a
448. 1:08:07challenge that you have for your business and you answer a few questions about that challenge like what does it
449. 1:08:12cost you how would things improve if you fixed it and then when you hit submit 30 seconds later you get back a list of
450. 1:08:20questions that Consultants would would ask you here's the questions you need to answer to solve your challenge and it
451. 1:08:26does it in six different categories like branding and strategy and customer experience and there's six different
452. 1:08:32strategies and it asks you three questions in each of those categories specific to your problem so that was done in zapier the input
453. 1:08:41form that's at contentevolution.net that's just a Google form that's embedded in the website and then when you submit the form it
454. 1:08:49triggers a series of calls to open AI to the GPT API that generates the content and then I send the content back to open
455. 1:09:01Ai and I say for me format this with HTML so I basically have a a call that formats it in HTML and then I stick that
456. 1:09:09HTML formatted content into a Gmail template and it gets sent back to you and it all happens within 30 seconds so
457. 1:09:16what I've been using it for is kind of content generation tools um I use one for my company storyvine
458. 1:09:24where we'll ask clients you know what's the what's the kind of video you want to create and and the way our platform
459. 1:09:30works we ask a series of prompted questions in our in the storyvine app for someone to tell their story so we
460. 1:09:38say to the client what's the kind of story you want to tell and who's the audience and in the ideal video what would the person be saying
461. 1:09:45and normally what we would do in in the creative process is we would have a whole dialogue about writing the
462. 1:09:51prompting questions and it got really complicated and not good and so now we have chat EBT to it where we ask him all
463. 1:09:59these questions we send that to open AI with a very specific prompt design to generate our kind of prompts it sends
464. 1:10:08them to our slack Channel and then we have a starting point so we say Okay based on our conversation here's five
465. 1:10:14questions that I think capture what you're trying to accomplish and then we just start editing questions rather than having a generator
466. 1:10:21from from Whole cloth so yeah it's it's amazing that way and and that's zapier it you can kind of think of it like this
467. 1:10:30the way those automations work there's what they call a trigger which is kind of the input right so either a new line
468. 1:10:37gets added to an Excel spreadsheet like a Google Sheets you know row or a document gets uploaded to a folder or
469. 1:10:46you know someone fills out a Google form so that kicks off the Automation and then you can take the data from that
470. 1:10:53initial trigger and use that in another prompt to in the case of open AI send it to open AI do an interesting prompt with
471. 1:11:00it then take that content and send it send it to someone so you have input processing output and then you can have as many steps as
472. 1:11:10you want in the middle of that but that's basically what it is what's the coolest thing you've seen or done with code interpreter
473. 1:11:20um the thing where I I don't think I can see it anymore because one of the weird things about code interpreter is if you go away from it um
474. 1:11:30it kind of erases all of the outputs oh here it is it's still there good so I up check this out this is just wild
475. 1:11:41and and this is like this is this is just scratching the surface like I haven't really done [ __ ] with code
476. 1:11:45interpreter yet I just haven't quite figured out what I want to do with it I uploaded this image so this was an image that I
477. 1:11:59got off the the AI salons Discord server that one of the artists there Ali created and I said to to code interpreter
478. 1:12:13let's see I said find the 15 average colors in this image plot their relative percentage in a pie chart make each segment
479. 1:12:25the color represented and label each color with a creative crayon color that is memorable and brings that color to life
480. 1:12:33and then so what code interpreter did was it it wrote a set of instructions for itself so I'm going to flatten the
481. 1:12:40image I'm going to apply clustering we'll be using the k-means algorithm for instance we're going to calculate the
482. 1:12:47relative percentages of the colors we're going to create a generative and memorable name for the color and then
483. 1:12:51we're going to plot a pie chart and so then it starts doing that it starts writing um python to pull the colors out of the
484. 1:13:02image and then it says it took too long because the image is Big so we're going to down sample it so then it wrote code
485. 1:13:08to down sample the image now I'm not telling it to do any of this this is it self-instructing itself self-coding and self-executing the code
486. 1:13:19so then it got the colors and then I said oh now we got to make good names for the colors and get their
487. 1:13:25percentages so it came up with a bunch of names for the colors with the relative percentages and then it it got stuck it did it it
488. 1:13:31told me it generated the image but it didn't so I had to two times I had to tell it to make the pie chart and then
489. 1:13:37and I said oh I apologize for the inconvenience there was a memory error and then let me try that again and so it
490. 1:13:45created this pie chart so these are the the main colors in that image and the names are like Raven's Wing Mossy Stone
491. 1:13:52Sky reflection ocean depths burning Ember and and like these are good color names so in I don't know a minute and a half
492. 1:14:04without having to do anything it did all this now is that actually useful not for me
493. 1:14:19right now but it's like like right now where I am with code interpreters where I was with chat gbt for the first month
494. 1:14:25or so parlor tricks so so right now for me code interpreters parlor tricks but particularly watching videos of people
495. 1:14:35on YouTube who are data analysts using Code interpreter and watching their faces go I mean the fact that it
496. 1:14:49it you kind of give it a go you say here's a blob of content a blob of data tell me what's interesting about that
497. 1:15:04you can't wait that's not possible to do right if you have a big data set you need someone like Gartner or an
498. 1:15:13analyst or a research scientist to look at all the data think about all the data understand all the data and then come up with
499. 1:15:22interesting ways to analyze it and then come up with interesting ways to visualize their analysis it does that now in like 30 seconds
500. 1:15:33so you could take you know we live in this age of [ __ ] up misinformation and and just [ __ ] right and that's going to get worse
501. 1:15:42it's going to be relatively straightforward for just normal human beings to take really complicated data
502. 1:15:47and say tell me the truth based on the data tell me what's going on based on the data and it will visualize it and use
503. 1:15:56metaphors and whatever you want it to do however you want to look at the data it'll look at that for you um
504. 1:16:03and that's just like like the Teeny tiniest tip of the iceberg so I I don't know I don't know I I am I'm I'm
505. 1:16:12relatively speechless about code interpreter because it's it's so much deeper than my capacity to even begin to understand what it
506. 1:16:22makes possible that it's a it's a little humbling I feel I feel I feel a little bit like a dum-dum I'm like a little bit
507. 1:16:28of a dum-dum I take all this data I put it in there make pretty pictures but I don't know what to do with it I just a dumb dumb so
508. 1:16:37I don't know over the next month I'll get my head around it because that like my mission right now in life is to
509. 1:16:49just learn as much as I can as quickly as I can about what this makes possible I don't really care if I'm an expert at
510. 1:16:57any piece of it I just want to understand what it is and what it makes possible because I feel like if I have
511. 1:17:03that knowledge then I can start to do things like oh you want to do that here's how you would do that oh you want
512. 1:17:10to do that here's how you do that oh someone's stuck in solving that problem I know how to solve that problem like I
513. 1:17:15want to I want to be able to understand how I can even think about this stuff so that I can solve problems with it
514. 1:17:23because it's the most powerful tool like in the history of humanity and we get to play with it if you don't get how lucky we are to be
515. 1:17:32alive right now this is printing press steam engine fire written language level technological advancement
516. 1:17:47this makes the world wide web look like [ __ ] Child's Play like it's big it's profound and we get to play with it we're in at
517. 1:17:59the ground level of it that's lucky that's lucky do you think it will benefit sysid minerals or eliminate them all together
518. 1:18:09I think that AI operations I I think I think I think AI tools are going to here's the way I think about these generative AI
519. 1:18:20tools with any complicated system I think that the AI tools are going to put a a simplification layer on top of
520. 1:18:27complex systems in fact go look at the video of Microsoft made an announcement that they're going to roll co-pilot into
521. 1:18:35Windows 11. go watch that video because basically you can just say in human terms change these settings in Windows
522. 1:18:42and it just goes and change all those settings in Windows you don't need to you don't need to learn how to change
523. 1:18:48your Windows settings when this stuff is incorporated into windows so imagine that for the entire Tech stack of a big
524. 1:18:56Corporation right where there's going to be this simplification layer sitting on top of these really complicated systems
525. 1:19:03once that layer is there it's going to be very very easy for people to do their own system administration stuff given
526. 1:19:10the right permissions right but figuring out what AI tools to use and how to set those permissions the AI
527. 1:19:16operations of setting all that up is an entirely new category of work so I think sysadmins will become AI admins and and
528. 1:19:26then they'll maintain that simplification layer and and they'll also I think there's still going to be room for assisted men's to you know
529. 1:19:35install all those systems but I think the AI will do the installation and we'll do the patching between the
530. 1:19:41Frankenstein [ __ ] that's there and make it much more simple to manage but someone's gonna have to set up the AI so
531. 1:19:48I don't think sysadmends go away but I think their jobs change dramatically if that makes sense so any sysadmin that goes I think AI
532. 1:19:58stupid they're going to be out but any assisted man that says holy [ __ ] this could change everything for the better
533. 1:20:05they're going to be fine um we can go Google troll accounts oh yeah that's I that's one of the cool things
534. 1:20:15about this AI stuff is like part of the fear is well the bad guys are going to have these AI tools they're going to be
535. 1:20:21able to [ __ ] with us but we're going to have these AI tools and we're going to be able to [ __ ] with the trolls in the
536. 1:20:25in the thieves it's going to be good it's gonna be fun foreign the scammers they get in and they hack their machines and delete their hard
537. 1:20:39drives and [ __ ] like that I love those things that's gonna get much more entertaining what do you think of using llm and
538. 1:20:48placing human Concepts into the data sets like bravery I think Concepts like bravery go have an interaction about um
539. 1:20:58uh jungian archetypes and Joseph Campbell and the the you know the heroes journey and things like that like it's
540. 1:21:05all in there especially things like that human Concepts like bravery and things like that because it's been in
541. 1:21:12literature for so long and it's all public domain there's so much written about all of that kind of stuff it'll be brilliant at that
542. 1:21:21um go to if if you're if you're kind of new to these things go to prompts.chat and understand how you can tell chat gbt
543. 1:21:28to act like a philosopher or act like a Humanity Humanities Specialist or or a sociologist or you know social Anthropologist
544. 1:21:38um sort of get it focused into the your area of interest and then start interacting with it I it it is probably
545. 1:21:45way smarter than you realize on on those issues particularly I just entered the room when you were speaking about PowerPoint slides what AI
546. 1:21:57app did you use I don't what was I talking about I don't know what I was talking about Becky
547. 1:22:11um I haven't used any AI apps there's the the one that I know is called begins with a T I can't remember what it is now
548. 1:22:21I've been I haven't used any of the AI PowerPoint making things I have used chat gbt to give me PowerPoint outlines
549. 1:22:29so like presentation outlines I have done that but that's all those all those all those tools that are making PowerPoint slides
550. 1:22:37are they're combining two different Technologies they're basically it's a chat GPT prompt to say write an outline
551. 1:22:42and then write the copy of the of the slides and then they're taking the copy of the slides and sending sending that
552. 1:22:49to an image generation tool probably Bing or dolly um and they're generating the the image based on the content of the slide from Dali
553. 1:22:59so it's probably two open AI calls it's probably an open Ai call for the outline trying to think how I would do it yeah
554. 1:23:14I'd probably do the outline first as one call then I would do the text of each slide and then I would send the text of each
555. 1:23:22slide to generate the image of each slide and then they're just assembling that so um so you can accomplish that just using
556. 1:23:30chat EPT if but if you want to go find that go to that that bottom URL futurepedia.io I bet you there's 200 different PowerPoint
557. 1:23:39you know type in a sentence get a PowerPoint out of it I bet there's 200 of them in there umiza is a great business idea yeah man
558. 1:23:50love Kyle you are so funny hey all hey Joy thank you you all see the future man you all see the future man yeah I did so I was
559. 1:23:59playing with the AI and I saw the fugitive yeah it's [ __ ] gnarly dude yeah it's all right I am clearly sugar deprived I
560. 1:24:13gotta go I gotta go make dinner soon Beatles Led Zeppelin wait Beatles Led Zeppelin Steel music Led Zeppelin stole from Beatles maybe I
561. 1:24:24don't know they were more bitching about it when you got a singer like Robert Plant I don't care if you stole the music and that whole band
562. 1:24:36I I hope you're not busting on Led Zeppelin they're [ __ ] miraculous I mean the Beatles are too but
563. 1:24:43if AI Vector analyzed all my social plus they think I'm a socialist Dom footballer well okay whatever floats your boat I mean
564. 1:24:53that's I think one of the things that's sort of the creepy scariest the most to me is that I have normalized in my mind probably 20
565. 1:25:03years ago ah [ __ ] it I'm just gonna give all my [ __ ] to Google let them monetize it but I kind of like
566. 1:25:11put it into the machine thinking like yeah like no one's ever gonna really look at that and now these AI tools tools come along
567. 1:25:19it's going to be able to look at like our past four Decades of just nonsense and Ridiculousness and just go hey want
568. 1:25:29to see what an idiot you were in 1987. here you go it's like oh my God that's coming so oh well Mission Impossible Barbie or Oppenheimer
569. 1:25:41uh oh Mission Impossible Barbie or Oppenheimer I so it sounds like for me Oppenheimer is the one I'm excited to
570. 1:25:48see just because there's there's an IMAX version of it that he did in like it's not quite four by three but it's it's
571. 1:25:55just beyond that one point oh maybe it is four but I don't know four four by three um but but it's uh he shot it in IMAX
572. 1:26:05but to make it look like it you know it was on the TV um because that's that's what everyone learned about all that [ __ ] from so
573. 1:26:12that's the one I wanna if there of those three that's the one her is a great movie already in China you said that before Psychotherapy is
574. 1:26:24going to get better yep I think it is Willie's loneliness in our society especially with the elderly totally agree
575. 1:26:31imagine the magic robot pets will be yep totally agree yeah robot pets are going to be amazing and it's I mean we're we're gonna allow
576. 1:26:41there's a there's a company out of Boulder right now that's got this it's it's kind of weird but it's kind of cool so so it's a
577. 1:26:49it's like a mannequin head but with no facial expression but it's got a projector on the inside of it so it
578. 1:26:56projects the face onto like a blank mask so this goes back to like Greeks and the masks in theater right so so what the
579. 1:27:03what the object is is like it's like a bust right where it's got a platform that's got a speaker in it so it's like
580. 1:27:10an Alexa speaker it's got a camera that's looking at you and then it's got this face and the face can react and
581. 1:27:17blink and do all that but it's just a projection behind this mask and so I could see something like that being like I could see it having
582. 1:27:24different modes like it's got pet mode where it's like a cute puppy and it's just happy to see you all the time or it
583. 1:27:31could be therapist mode where it's like yeah go tell me more uh-huh and right and and so I could I could see us having
584. 1:27:39these agents that kind of shape shift and have different personas so it doesn't have to be everything for
585. 1:27:46everyone it can it can I could have 20 different things that depending on what I want to do it'll act like my travel
586. 1:27:52agent or it'll act like my therapist or it'll act like my friend or it'll act like my pet that could be kind of cool like I mean right
587. 1:28:03drink I don't know how can the average guy utilize chat GPT I don't code you don't need to code so go here I'll flip these coding is not
588. 1:28:16necessary in fact what what you're doing with these tools is your your kind of programming with the English language is
589. 1:28:23the best way to think about it so if you haven't played with chat gbt that top URL chat dot open at chat.openai.com is chatgpt
590. 1:28:34these other two things you can go to them as well but that's the real deal and then the thing down below here
591. 1:28:41prompts.chat that's just a it's a Word document that teaches you how to talk to this thing and just go there go to prompts.chat and
592. 1:28:52it will sort of teach you what these things are and what prompting is and prompting is basically just a fancy it's
593. 1:29:00a geek's way of saying you're going to have a conversation with this thing and think of it less like a Google Search and more like
594. 1:29:11like someone just gave you an intern who needs a lot of Direction but is a savant and knows everything that's kind of what this thing is
595. 1:29:20and so for for you as an average guy that's not a coder depending on what you do like I'll give you a good example there's the
596. 1:29:31maintenance guy in the building that I work at um guy named John Roberts really nice guy um he's using chat GPT
597. 1:29:41to create a bot that teaches other maintenance guys how to troubleshoot maintenance problems in the building so
598. 1:29:49he's got a bot for for HVAC he's got a bot for um for electricity for plumbing for AV systems things like that and and so he's
599. 1:30:01using it to like troubleshoot things within the building and then from that that gave him the idea to do this for other people
600. 1:30:08um so any take it take anything that you know and start asking that thing about the [ __ ] that you know so if you know cars or if you know
601. 1:30:21lawn care or if you know woodworking or it doesn't matter or if you know project management or how to deliver [ __ ] anything
602. 1:30:30start interacting with that and ask asking it questions like like you know how could I do this better how could I
603. 1:30:36make it more efficient could you give me tips on you know uh a a more organized delivery system that would make my routes more efficient
604. 1:30:46like it it literally doesn't matter but but go to prompts.chat what you'll see there is it's a whole long list of
605. 1:30:53personas where you can tell chat GPT to act like that kind of specialist so the guy John the maintenance guy in my
606. 1:31:00building said I want you to act like you know a master plumber that can help me troubleshoot a plumbing problem and it
607. 1:31:07did and it was really good um so so you don't need to know programming and so just start with that and it'll it'll help [Music] um
608. 1:31:22all right AI learning lab I need a business consultant to help with my small business um so you can just go to chat GPT and start
609. 1:31:32interacting with it like it's a business consultant there's also there's an app out there there's another tick tocker
610. 1:31:38named McKenzie Bose he's got an app called GPT boss and the basic idea with GPD boss is he's done all the pre-prompting already so
611. 1:31:47he's got a business consultant in there and he's got a marketing specialist in there and he's got an accountant in
612. 1:31:52there he's got all these different um he calls them employees that you can hire and it's it's uh I think you can use it
613. 1:32:00for free for a certain amount of credits and then you just buy more credits basically um but but go play with GPT boss I think
614. 1:32:08it's gptbossbot.com but he's a really nice guy and uh and yeah it's good it may help next steps for storyline well um
615. 1:32:22so that so there's there's two things um one is just figuring out um looking at what our current business is
616. 1:32:35and projecting out a year and saying what of our current value proposition goes up in value what stays the same and what drops and
617. 1:32:46so I think I've got that figured out so right now one of the key components of storyline that's our logo by the way see yeah um
618. 1:32:59one of our key value propositions is automation which a year from now is gonna be a big fat zero in terms of
619. 1:33:09value everything's going to be automated nobody's going to give a [ __ ] so right now we talk about we we have Auto Magic
620. 1:33:15editing and people are like oh that's magic that's fantastic a year from now they'll be like yeah what else
621. 1:33:23so I think that goes down in value we're very good at getting human beings to authentically share their stories powerfully
622. 1:33:32I think storytelling and authenticity go up in value so one thing I need to do is look at how I'm talking about the business and
623. 1:33:42shift that narrative so that it's future proof so that's one one thing and then the second thing is how do I take these
624. 1:33:49generative AI tools and expand what we're offering so right now when you use our system you get back a video well why
625. 1:33:57not get back a video and a whole bunch of related content with it like here's a description of the video and here's
626. 1:34:04hashtags for that video and here's tweets about that video well that's relatively straightforward to do and so and then how do I roll
627. 1:34:13those two two things together the messaging and the new features to actually create something new so that's that's what's next for storyline
628. 1:34:21all right I hear trans clanking I gotta run I will that's wow about the screenplay cool I'm watching your Tick Tock and ignoring
629. 1:34:34you thanks for everything that's a story of my life um I knew you were I knew words were coming out of your face I have no idea
630. 1:34:44what you said do you pay for chat gbt I do Kyle sent you a DM okay great I'll check it out I do respond to those so feel free to do that
631. 1:34:54all right listen I'm gonna run I know I uh I didn't get to a bunch of questions here but it's been great hanging with
632. 1:35:00you I'll probably come back later I gotta go do dinner ribs on the grill and you know stuff I got to do dad stuff I'm
633. 1:35:08gonna go do dad stuff peace [Music]