aiLL 6/19/23 - The Generative AI Tsunami Is Coming

Transcript

1. 0:00foreign
2. 0:16let's try this again we will see if it's gonna work I thought it was Comcast but it turns out it was my son's downloading [ __ ] ah
3. 0:30Father's day was Sunday they get their Revenge today you see how that works hey everybody um do me a favor as you join here tap
4. 0:43the screen the like screen that'll let people know we're here or you can share the live that would help also
5. 0:49uh this is the AI learning lab I'll get chatting here in a second funny man it's all comedy all the time
6. 0:59it was my shitty kids that were screwing up the bandwidth here it was not Comcast I prematurely blamed Xfinity for shitty
7. 1:08service and it was just shitty children they did yell at me though for resetting the router without warning them so
8. 1:17I think they've got a pretty good argument that I was a shitty father in that in that point uh hey hey Sonia how are you good to see you
9. 1:25good to see you welcome everybody welcome welcome welcome welcome well we'll let this thing get wound up
10. 1:31here I appreciate the Taps on the screen that lets people know we're here uh how old are your kids I have 23 year old twin boys
11. 1:42and they are lovely human beings my name's Kyle Shannon it's the AI learning lab if you are new to generative AI
12. 1:57um go here that's chat.openingi.com just point your browser there that's chat CPT if you've not played with it yet go play
13. 2:05with it go start playing while I ramble on if you've got questions about AI pop them down below uh we have fun here we
14. 2:13we uh we we uh we dig in deep into the philosophy of this AI stuff the future of it the the glory of it the crap that
15. 2:24doesn't work of it which is a lot we talk about it hallucinating I like to call chat GPT mansplaining as a service
16. 2:33because it will confidently answer you even if it's wrong like a good man like a good man in your life oh yeah
17. 2:46um uh my background I'm an entrepreneur um I've been I've been on the planet for a while as evidenced by this
18. 2:57um but on a slightly more serious note the uh I was there for the early days of the World Wide Web I started one of the
19. 3:04early digital agencies in an online magazine back in those days and um felt pretty lucky to be part of a technological change that you know
20. 3:16transformed everything and I didn't think I'd see another one of those in my life and here we are um November 30th 2022 when chat GPT
21. 3:25launched from those wacky kids over at open AI um that changed the game in my opinion and I think this stuff is coming
22. 3:34way faster than any of us are prepared for I think the changes are going to be profound um and I think I think they're coming so
23. 3:43with that feel free to pop any questions down below I'll do what I can to answer them I will probably fall behind fair
24. 3:52warning I will probably cuss if you're not a fan of the cussing I'd get out because when I get excited the shitty words come out
25. 4:01do you think software engineer I assume you meant to finish that sentence Oh wait do you think software engineering
26. 4:07needs will decrease because of AI I think they'll change dramatically I think that we're still going to need
27. 4:15software Engineers there's going to be a lot more software written um but it's probably going to look more like managing AI systems
28. 4:29and being more like a product person than actually doing the individual coding I think I think some amount of
29. 4:36coding will get done but it'll probably be more like you know connecting modules together and things like that because
30. 4:43the the generative AI tools are going to do a lot of the a lot of the base work right so if if you're an engineer and
31. 4:50have haven't used GitHub co-pilot the kind of pair programming tool within within GitHub I'd strongly encourage you
32. 4:57to go do that so I do think software Engineers are going to be in demand but they're going to need to be AI literate software engineers
33. 5:07because what's what's wild about these generative AI tools is that so for like how long have I been dealing with software since probably
34. 5:19I don't know late 70s early 80s in some form or another and the thing that I've always known about software is you kind of think
35. 5:29about what you wanted to do and then you start thinking about the logic and and how do we what are the data elements
36. 5:36that we need to code and how do we code them and what's the logic for how we glue them all together um generative AI just does all that like
37. 5:46all the logic thinking it's really good you can just ask it to do [ __ ] it will figure out the form and it can look at like there's a
38. 5:54there's a thing coming from chat GPT called code interpreter where you can upload big piles of data like a giant
39. 6:01spreadsheet it will just learn the spreadsheet and then do full-on data analysis for you like it's incredible at that [ __ ] so so I
40. 6:10think I think the engineering world is is going to transform in pretty profound ways um one of the things Jason kalkanus is
41. 6:20what's it called the all-in Pod with the like the four billionaire dudes talking they've been talking a fair amount about
42. 6:26this that what they've been hearing is that tech companies in San Francisco that are going all in on AI they're
43. 6:33seeing 30 to 40 percent productivity increases for companies that are going all in on it so that's kind of profound
44. 6:40stuff when you see that significant of productivity increase just because of switching to something like generative
45. 6:45AI I'm having mid-journey issues I am creating a series of images for a book and it's not going well sorry about that that's that's
46. 6:55that's tough like so the thing about generative AI I would say the single most frustrating thing about generative
47. 7:01AI is that um getting it to be predictable is really really hard um let me share a couple of things with you that may help
48. 7:13all right I flipped my camera now I gotta turn off my little thingy and then we're gonna go over here we're gonna do
49. 7:19some playing okay so the first place I'm gonna send you well let's see yeah yeah yeah okay so go to open art wait open art
50. 7:36dot is it AI yeah go to openart.ai slash prompt book openart.ai promptbook now this this is for stable diffusion
51. 7:51but what this but it's good it's good for mid-journey as well because this is more like a Concepts primer than than it is
52. 7:59you know a specific to to prompting so if you flip through here it will talk about um prompt engineering and and what it is
53. 8:09and like what the components of a prompt are and and you know really expanding on the different elements
54. 8:17um it'll then give you some examples then it goes into modifiers and modifiers are things like what are the different terms that these
55. 8:26image generation tools know like it knows all the photography terms right like tilt shift and long exposure and
56. 8:34color splash and satellite monochrome um so so that's just photography different kinds of lenses um different kinds of lighting art
57. 8:43mediums this is probably where where you know you could potentially have some advantages if you're doing stuff for a
58. 8:50children's book maybe you want it to look like a watercolor maybe you want it to look like particular characters it
59. 8:56also knows artists right it also knows Styles let me see I think there's Styles coming up here um like illustration styles
60. 9:07um you know Vector children's book right things like that so so this it's like a 96 page uh document if if
61. 9:19you are feeling a little lost this this could help you so that's one thing you can do another thing you can do is I'm
62. 9:26going to pop you over to poe.com so this is one of the sites I recommend going to play with so when you come to poe.com
63. 9:33it's basically just a normal um like chat cpt-like experience and you know you can you can type in stuff and it will
64. 9:44generate content but there's also um Bots that people have created so there's one called MJ Captain so poe.com MJ
65. 9:56Captain which stands for mid-journey Captain um there's another one called po.com mid-journey so let me just I'm gonna go
66. 10:04to that one poe.com slash mid Journey so I think this is just something that someone put together but what happens is I can do children's
67. 10:21book illustration of a happy elephant um walking through a field of flowers
68. 10:38or whatever so so when I type that in what this bot's gonna do is actually rewrite that with all sorts of um of uh
69. 10:52what it was like embellishments and things like that now this just did um this did it as a color photograph I'm
70. 10:58gonna say I don't want a photo I want a watercolor illustration let's see if let's see if it does better
71. 11:12okay I apologize for misunderstanding here's the here's the prompt watercolor illustration of a happy elephant walking
72. 11:19through a field of flowers the elephant is depicted with a big smile rosy cheeks as if enjoying the scenery blah blah blah then then
73. 11:28it's got it's weird it put directors in here and it puts some photographers in here so I'm gonna kill the photographers but let's let's pop
74. 11:37over to um mid-journey so if you haven't seen mid-journey now you're in for a treat okay so to create a mid-journey image
75. 11:47you actually have to go to this thing called Discord if you don't know what Discord is it's basically kind of like
76. 11:51slack but for geeks these things down the left are different servers so think different communities I'm now in the AI
77. 11:59Salon community so this is the the uh AI Salon that I started and we have a mid-journey channel so here's one of the
78. 12:07channels these uh these things here are channels of different conversations basically and then to create an image here you
79. 12:15type slash imagine and I'm just going to paste in that prompt and I'm going to kill the cinematographers and photographers
80. 12:24you could replace one of the things that could help is you could replace the artist that they give you if if you find
81. 12:32specific illustrators of children's books that you like you can pop them in there but I'm just going to go ahead and hit
82. 12:39return and now so it takes that giant prompt that the chat bot created and it will it should start making us
83. 12:47some images so we'll see how these do um so I don't know if that helped but hopefully that'll at least point you in
84. 12:54in some directions we'll we'll sit here and watch as that thing goes I'll try to figure out some other questions here and
85. 13:06I'm not sure if that's looking like an elephant maybe oh yeah it is okay have you tried the prompt library for
86. 13:14help oh yeah that's the oh that's good Sonia I'll go show that okay so wait so while that's while that's launching so
87. 13:20what Sonia's talking about so we're gonna go to the mid Journey um to the mid-journey server and then
88. 13:30within the mid-journey server there is a prompt Library let's see where is it uh [Music] is it in here
89. 13:46it was a stable diffusion prompts image prompts so so there's image prompts multi-prompts patterns textures environments
90. 13:58characters you can also Within These communities you can also ask for help so there's General conversations and show
91. 14:06and tells so you can ask people for help uh there's lots of places to do things see there's these are voice conversations
92. 14:17going on so so lots you can do there let's go back and see how did that illustration come out [Music]
93. 14:27well they're definitely children's book like like this this lower right one's pretty good that's not bad although the the little
94. 14:40the pointy trunk there's there's some genetic issues going on in the drippy ears this elephant's got some issues
95. 14:49um that one looks that one looks pretty cool so anyway that's you know Without Really Trying you know I got this kind
96. 14:55of example now if the challenge that you're having is getting consistent characters across different renders then
97. 15:02you're gonna need to do some things like um use the blend tool within mid-journey uh or upload an image and then create
98. 15:10other images based on that image so that might be a way that you need to go as well so hopefully that helped
99. 15:20let me get my little background up here looks like we have a few folks here welcome everybody my name is Kyle
100. 15:27Shannon this is the AI learning lab if you have questions about AI pop pop them down below I think coding is going to be
101. 15:33needed more than ever I think that's right I think how coding gets done is going to change though um so so again I I go back to I think we
102. 15:43need Engineers but we definitely need Engineers who who are um AI generative AI literate because how you think about programming logic
103. 15:54really does change with generative AI mid-journey is ignoring my prompts now that stinks assume the worst hope assume the worst
104. 16:03hope for the best prepare for everything I think that's right in general I was able to get 20 images then it just
105. 16:10stopped giving me back what I'm asking for hmm that's weird and you're in you're in the official mid-journey Discord doing that
106. 16:22um I would go ask questions and I I would share your prompt with people there might be something so so let me show you one other thing
107. 16:31you could have inadvertently um hang on you could have inadvertently um change to setting so if you're in
108. 16:47mid-journey Go slash settings and then you'll see all of these um settings so you've got mid Journey one two three four five and five point
109. 17:03one um four might be better for illustrations I think it's pretty good 5 and 5.1 seem to be optimized for photography
110. 17:12I I would play around with different versions try version four um niji is a very specific Japanese illustration style but that could be
111. 17:22something to play with and then you've got this stylized low medium and high and very high there might be something
112. 17:27there so I would say go go check your settings and also look at all of the different options that that you can use
113. 17:34aspect ratio things like that so yeah it's really frustrating when when you get these tools not doing what you want
114. 17:42them to do it's maddening because you when you get it right you're like oh I'm a genius and then all of a sudden it's
115. 17:48like nothing's working it's really frustrating so I uh I feel for you there brother um um all right and it looks like you're
116. 18:00helping mid-journey's giving me back weird images that don't match my prompt at all that's very strange
117. 18:04I'm even using image uploads yeah that's weird there must be something weird and I I would say if you haven't done it
118. 18:11start your prompt from scratch as well like retype it maybe there's some weird character that you keep copying and
119. 18:17pasting a weird character and it's just freaking it out don't forget with these tools they're we're very new like a lot of these
120. 18:25generative AI things mid Journey's been around for what two years it's it's been sort of working well for
121. 18:32you know maybe six months maybe eight months um these things are very raw and very janky and there's there's a lot of
122. 18:41people I I don't know what the latest numbers are in mid-journey but it's a lot so it's also possible um oh you know it's worth sorry sorry
123. 18:51having add moments um just like like a bunch of ideas are popping in my head um try quitting your browser and
124. 18:58restarting also try running oh well no you're you're in Discord never mind strike that I'm wondering if there's a
125. 19:05way you can reset the cache and Discord quit Discord and restart it that could help maybe you've got some weird cash
126. 19:10[ __ ] going on I don't know strange strange strange okay adding you on Discord what practical uses have you found for jet chat gbt
127. 19:28um lots and lots of them so I've created a number of content generators oh it looks like we've got a bunch of new
128. 19:33people here hey everybody my name is Kyle Shannon this is the AI learning lab these URLs behind my head I'll sort of
129. 19:39scooch out of the way here um these are here just kind of as a public service for the community um chat.openai.com that's the official
130. 19:48chati PT website so if you haven't played with it go start playing with it just go do it now trust me um bing.com is also chat GPT it's gpt4
131. 19:58but it's connected to the internet so when you go to bing.com it'll look like a Google search but it's it's Microsoft
132. 20:04search click on the chat button and thank you for that that's a lovely uh what is that a swan origami swan uh but
133. 20:12bing.com is gpt4 connected to the internet for free and then poe.com is another chat CPT tool that'll let you play with six different models
134. 20:23um chat gbt like models um prompts.chat will teach you about prompting and future PD is a big directory of tools so that's what all
135. 20:30that stuff is um I'm an entrepreneur is my background I am I like geeking out on this generative AI stuff we'll answer
136. 20:39questions we'll do what I can um so what practical uses have I found for chat GPT um anything related to writing it's it's
137. 20:50incredibly good at so that might be things like um maybe you want to come up with a marketing plan for your business maybe
138. 20:57you have a marketing plan and you need some ideas for copy for a particular campaign maybe you need creative briefs
139. 21:04you know for to come up with ideas for the campaign um it's really good at coding so if you're doing anything code it can you
140. 21:16can copy and paste code in it it can debug it for you figure out what's wrong with it it can document code um it can do creative writing
141. 21:25um you can do what I like to call parlor tricks like have it write songs and poems and you can just start to play so
142. 21:31what I what I would encourage you to do is is if you haven't played with it just go play with it right just start asking
143. 21:38it anything and and let it be kind of um a toy and then what I would do is how how I get to practical uses is rather
144. 21:48than sitting in front of the tool that can essentially do anything and say well what could I use this for
145. 21:54flip it around and I would say look in your in your life either professional or personal where you've got a specific
146. 22:01problem you want to solve like maybe you want to get in shape and you could say hey I want I want to see
147. 22:07if chat GPT can help me come up with a meal plan for the week that help helps me eat healthier yes it can it could
148. 22:15come up with an exercise routine you can tell it you know where are you um you know I haven't worked out in a
149. 22:21while I'm just getting started again you know I need something easy to to be an on-ramp to getting back in shape it'll
150. 22:26give you a whole plan for that um so so you know just start playing and start taking specific problems in your
151. 22:34life and using it to solve them and and you'll get to practical uses really really quickly um Ctrl F to search for Library looks
152. 22:44like you're helping each other that's cool what do I think quantum computers do you think quantum computers will be
153. 22:50integrated with Advanced AI to speed up data processing I assume so um the the stat that I heard about quantum computers is that they're
154. 23:00doubling their number of qubits every year now and and when they add a qubit it's an exponential increase in power or
155. 23:07some crazy [ __ ] like that so I know that they just launched they just created a 500 qubit chip someone did so so those things are
156. 23:16pretty powerful I don't know how far they are from viable commercial applications right like I don't know
157. 23:22where and their Continuum they are but I assume it's going to do things like dramatically speed up the training side
158. 23:30of AI it might not be so much on the real-time side my guess would be it's on the training side but you can probably
159. 23:36train these large language models much more quickly much more um at a far greater scale and potentially using completely new models
160. 23:47that will make the current Transformer model look like Child's Play I don't I I don't know enough about the world of
161. 23:52quantum Computing other than it's weird as [ __ ] and as powerful as as it's just nuts powerful but it's all I think it's
162. 24:02all still research right now but I don't know but basically what Sam Altman said is that the more compute power the more
163. 24:11data and the more compute power you throw at these things the better they get so it will be interesting to see
164. 24:19um how do I search specific bots on Poe let's go look at that um [Music] so we'll pop over to poe.com uh what did I do
165. 24:33what did I do oh wrong window okay so here we are at poe.com so so you've got you've got your um Chachi BT models at the top and then
166. 24:46you have Bots below that there's this explore Bots button so if you hit explore Bots now I can go I can go over here and they have a
167. 24:54search button tell me they don't have a search they may not have a search so you can look for history Bots new
168. 25:06Bots advice Bots game Bots travel Bots music Bots but it does not look like they have search which AI Bots
169. 25:17professional programming Bots oh that's cool HTML machine let's go play with that one um make me a um
170. 25:32website about sauerkraut there you go
171. 25:47to test it can you write the HTML for me sure there's your HTML uh it doesn't
172. 26:01look like you can search for the Bots which is weird explore Bots you can create a bot create a Bots fun
173. 26:12it's really easy to make your own go to create a bot I've done a bunch of those the fact that they don't have a search
174. 26:20engine is almost as unforgivable as the fact that open AI doesn't have a search for your chat history which is just maddening but you know
175. 26:31what are you gonna do we're early if you want to play with these these these if you want to play with the most powerful
176. 26:37technology in the history of mankind you know you're gonna have to you're gonna have to suck it up and recognize
177. 26:44that we're early and this stuff kind of sucks right now all right um do you think quantum computers answered that today I use Bing thanks to
178. 26:54you cool how did it go was it good anything interesting um could you explain exactly what a bot is yeah so this is I hate the term
179. 27:04because people have been talking about AI Bots for like a decade and they've sucked so historically what a bot is is it's it's a um
180. 27:14it's basically just a giant logic tree where where they where you kind of you you create a giant tree of if the
181. 27:24person says this then answer this and if they say that then answer that and then they got a little bit smarter where you
182. 27:30didn't have to ask it exactly but they've just been shitty for for decades Chachi BT is a bot in a sense and and
183. 27:40basically the bot is you type something in and it responds with something right that's that's it at its most basic it
184. 27:48kind of looks like a chat like you're having you know like a chat conversation on your phone but you're interacting with a machine
185. 27:56historically like the chat Bots that you see on customer support sites those are those um hand programmed you know trees and
186. 28:05they and they suck because you you can't possibly anticipate all of the different things someone might ask and so you know
187. 28:14invariably you're the one that types in something that the bot doesn't understand and it just either doesn't
188. 28:18answer you or gives you the wrong answer so what what happens with these large language models is you're you're giving it a specific
189. 28:28prompt to behave in a certain way it's at the at the base sitting underneath it is this thing called the large language
190. 28:34model it's it's basically um this thing's been trained on all of the internet so it kind of knows all of the internet so why chat GPT is so
191. 28:44powerful is that it's been trained on um so much data I'm gonna I'm gonna go play and show you some things here um
192. 28:56wait hold please hold wait for it wait carry on people carry on rather Cheerio pip pip okay so so in a typical chatbot let's go look at
193. 29:13one of the PO chat bots so I'll go to one that I created I created let's see let's do a fun one Shakespeare is an obvious one so because
194. 29:23Shakespeare replies in iambic pentameter um come on what's going on okay so this is if you go to poe.com I am Shakespeare I can type in
195. 29:38um tell me about um the smell of wild flowers in the springtime
196. 29:54spring time and so you know I type that it responded and in this case it responded like Shakespeare would
197. 30:05the springtime yields a soothing site indeed when wealth of wildflowers fair in Meadows spread with Hues of red and
198. 30:15blue and gold they plead their sweetest sense of Loft breezes shed so so in this case the bot is very very
199. 30:24specific to answer like Shakespeare you can also have one that writes mid-journey prompts or writes you HTML or answers in a specific way
200. 30:34but the the generic chat GPT you can ask it almost anything so so you can just say um explain quantum mechanics
201. 30:50as if I were 5 and use um Candy as the metaphor
202. 31:04and and so it just does that right so let's see sure imagine you have a magical candy box filled with different
203. 31:10kinds of candies these candies are very special because they can do really strange things in the world of quantum
204. 31:16mechanics candies can be in two places at once it's like having a candy that can be in your hand and in your pocket
205. 31:22at the same time it's called superposition right so so unlike a typical chat bot where someone's got a program in all these
206. 31:32answers anticipating all of the questions the large language models just know [ __ ] right and and they know
207. 31:38remarkable amounts of [ __ ] like um write me a C sharp uh clone of Flappy
208. 31:53Birds game that I can run in unity this is one that I did for a 10 year old and and it just writes you code
209. 32:05it understands what Flappy Birds is it understands what C sharp is it knows Unity it writes the code and then it
210. 32:13tells you how to install the code so so the chat bot metaphor is really just you type in some [ __ ] and it returns an answer
211. 32:25and the bot part is you're interacting with a machine not a person all right do you think quantum computers
212. 32:33I got that one already up I just lost my place welcome everybody my name is Kyle Shannon this is the AI learning lab let me
213. 32:41flip the camera back around so I've got this and I got this and I got this and then I got this and I got that uh if you're new here
214. 32:52um that top URL there is the official chat gbt website I'm not affiliated with it in any way in case you're cynical and
215. 32:59bitter and think I'm just here to scam you out of your money giving you free advice um why do you think normalizing flows
216. 33:09are becoming so popular why are they preferable to to diffusion um I don't know I'm not digging into normalizing flows that's something I
217. 33:20need to look up that's new to me um I would assume just judging by the name that if there's if there's something
218. 33:29about predictability in in output um that might be why but but I'm totally guessing I don't know what they are but
219. 33:37I'll definitely look those up um how could chat GPT be integrated with the Apple Vision Pro to change our lives so
220. 33:45a couple of things about Apple Vision Pro you'll notice that they didn't use the phrase AI once um they didn't use the term metaverse
221. 33:55once which probably pissed off Zuckerberg Tim Cook's probably like um what they did make very very clear
222. 34:06though is that so within the goggles themselves you've got an M2 Chip so you got a an incredibly powerful computer in your goggles
223. 34:15and then there's this thing called the R1 chip which I guess stands for reality or whatever it stands for
224. 34:22but that's basically an AI processing chip it's taking you know all of the input from all these sensors and
225. 34:27processing it in real time they made it very very clear that they're doing most of their machine learning processing on
226. 34:33device that's very important for privacy so right now if you use chat GPT you're sending your data to open AI
227. 34:42they're letting you use it for free in exchange for training on your data you can now ask them to not train on your
228. 34:48data but you have to do that explicitly so Apple's going to be really good at the Privacy thing so that's one thing
229. 34:57um a lot of what they're doing is using AI to do the processing and the taking of the images of the eyes and
230. 35:07projecting them automatically when someone walks into your field of view interrupting your full immersion
231. 35:14experience with their image like a lot of what they were showing is full AI what Apple's gonna do really well and
232. 35:24and they haven't announced this yet they were very careful not to mix I was disappointed in this but I understand why they did it they
233. 35:32were very careful not to mix anything like Siri Plus or you know their AI products with these these glasses
234. 35:44they were very very careful to say this spatial Computing is a new thing we're doing it's all about spatial Computing right
235. 35:51under the the subtext of it all is like this [ __ ] is all running with AI so when when Apple decides to come out with AI
236. 36:01they're gonna have you know some version of chat GPT it will probably be Siri Plus or the relaunch of Siri or some new
237. 36:08thing like Siri where it'll be a tool that you can interact with you'll probably be able to do it with your
238. 36:13voice it'll probably be close to if not real time right with no delay where I think Apple's gonna be really
239. 36:20good is with agents with autonomous agents where you'll have your own little version of you that you can send out to
240. 36:29go do tasks for you so you know hey agent or hey mini me or you know however the [ __ ] they decide to trigger it hey mini me
241. 36:39um the wife and I want to go on a vacation two months from now you know somewhere in the Caribbean can you go
242. 36:46find me some deals and and get you know give me some options and it will go off it'll find that stuff it'll come back
243. 36:52and say here's some things I found what do you like oh I like this one I don't like that one go ahead and book that and
244. 36:58then the next day hey mini me uh we decided we don't want to go on that trip anymore we're just gonna go on a road
245. 37:03trip but we want to rent a nice car can you cancel that trip and go so I think Apple's gonna be really good at having
246. 37:09these autonomous agents that go interact with all of your data and all of your services that you're already hooked up
247. 37:15to in the Apple ecosystem and then of course that will also be seamlessly incorporated into the space computing
248. 37:24platform so I think Apple's gonna be [ __ ] unreal at this generative AI stuff um but they're keeping the cards cards
249. 37:36close to their vest like they like they are you know like they do that those goggles were in development
250. 37:43for nine years five thousand patents is what I heard there's more than 5 000 patents in those goggles and it didn't
251. 37:50surprise me those those were incredibly impressive and it was an incredibly impressive piece of uh piece of Hardware I love this stuff too
252. 37:59I'm going through the kaggle book right now and learning about Gans very fun yeah Gans are really interesting
253. 38:06um Tick Tock a lot of the the filtery [ __ ] they do is is using Gans um you said how I don't know what that was in response to
254. 38:16do you think there will be a day where AI will replace authors screenwriters and maybe journalism um journalism
255. 38:23well so I don't know if it will replace authors screenwriters the this the screenwriters guilds on strike right now
256. 38:31and one of the things that is is they're striking about is they don't want producers to use chat GPT to replace them um
257. 38:38I think that I think their position right now is a little short-sighted but I understand it because these tools can
258. 38:44write screenplay they can write they can they can at least block out screenplays very very quickly they can do the same with
259. 38:54with novels when you get to journalism one of the challenges with um with chat EPT in its current form is it does a lot of what's called
260. 39:03hallucinating even if you're connected to the internet it will give you citations um for facts that it just makes [ __ ] up
261. 39:10like it's not actually aware of what it's telling you it looks like it is and it will confidently give you answers but
262. 39:16it's often wrong so I think journalism's got a bit to go um but I think every single um every single knowledge work position is
263. 39:32going to be fundamentally altered by these tools because these tools are going to be rolled into Office 365 they're going to be rolled into
264. 39:39Google Google's office suite um and and yeah they're gonna they're gonna transform everything so I think writers
265. 39:49writers who learn the AI stuff are going to be in a in a better position than writers who resist it because writers in
266. 39:59particular so so there's two things these tools are so good at writing that writers are absolutely existentially threatened
267. 40:09justifiably by these tools but also writers are really good with words they're writers and in order to get good
268. 40:18quality content out of AI tools right now you got to be good with words you got to be able to think in Concepts and
269. 40:24think critically so I think writers that learn AI are going to have a distinct advantage over non-writers
270. 40:31but yeah I think I think it's coming free AI video editor go to so the reason I have these uh things up on the uh up
271. 40:40on behind my head my big fat head futurepedia.io is a directory of AI tools there's hundreds probably thousands in there at
272. 40:52this point so just go into the video section there and start start learning I don't I stop following all the Tactical
273. 40:59execution tools because there's just too many I mean it's like it's like five or six tools a day it's impossible to keep
274. 41:05up with this [ __ ] like it's it's it's enough to just keep up with what's going on with the core large language model
275. 41:11stuff which is a lot that's a lot much less all the tools that are being built on top of this [ __ ] ah all right I mean
276. 41:22I already had AI generate the next book in The Game of Thrones series yeah exactly well and there's a here I'll
277. 41:30show you a cool thing if you haven't seen this um this is pretty trippy um so if you go to poe.com um where are we here okay
278. 41:46so you you've got Sage GPS or GPT for Claude plus Claud instant 100K Claude instant and chat EPT all of these are
279. 41:57chat GPT models the the three that say subscribed you have to pay for it right so so this one isn't free but if you pay
280. 42:04the subscription for Poe and again I'm not affiliated I don't give a [ __ ] if you do it or not but you get access to
281. 42:10this thing called Claude instant 100K 100K means you can paste in up to 75 000 words into the prompt and it won't break chat
282. 42:23EBT by comparison does about three thousand words and they're upping that to I think 12 000 words here shortly but
283. 42:30right now it does three thousand words so this is 75 000 words you can pop in the prompt so you could take the manuscript of a novel
284. 42:38paste the whole thing in there and go write me the first chapter of the sequel or or give me give me three mini outlines of
285. 42:47possible SQL plot lines um and it will it'll give you that um like that's nuts I I put in um I put in the screenplay I just pasted the
286. 43:00whole screenplay of 2001 A Space Odyssey and started asking the questions about it and it it read the whole thing in a
287. 43:06second or two or you know five seconds and then started answering questions about the screenplays it's remarkable
288. 43:13make a bot to search for the Bots oh that's a good idea yeah that would be probably pretty straightforward assuming you could actually
289. 43:23search their directory you might need to write a plug-in for that a bot is a big tree yeah AI makes its own questions and its own answers
290. 43:33yeah it was interesting but I'm new to it getting creative every day cool the whole point of this channel is get curious about AI um
291. 43:46my background if it looks like we've got some new folks here my name is Kyle Shannon this is the AI learning lab I am
292. 43:52not a machine learning specialist I'm not a computer scientist I'm an entrepreneur I've started 12 or 15 companies in my day
293. 44:00um I was there for the early days of the World Wide Web started in one of the first digital agencies and got to experience
294. 44:07um a transformative technology from its birth the world wide web and um I didn't think I'd see another one of
295. 44:16those in my life and Along Comes generative AI so the the birth of chat GPT for me is a seminal moment where
296. 44:23there's an inflection point where we shift from scientists and researchers to regular people having access to these
297. 44:30tools and and just to put put a little bow on how quickly this technology is being adopted I know there's a
298. 44:38tremendous amount of hype about Ai and it's really [ __ ] exhausting um Chachi BT went from zero to a hundred million users in six weeks
299. 44:50it was the fastest adoption of technology in history it it only took him two weeks to write chat GPT there's a whole story there
300. 45:01it got to 100 million users in six weeks the World Wide Web when it from when it was started to when it got to 100
301. 45:07million users was six years and we know how transformative that was so so that's part of what this channel is about is just
302. 45:18I don't know screaming at the top of my lungs like pay attention to this [ __ ] it's coming it's here it's coming faster
303. 45:26than you realize it knows other languages too it knows all the languages like what was I do oh I was on a uh was I on a live
304. 45:37I was on a I was on a live or oh no uh in in my in my Tick Tock Channel um in a number of my videos people are
305. 45:46putting comments in in their languages and so what I'm doing is I go in I copy it out I paste it into chat GPT I say
306. 45:54first tell me what language this is and then you know put it in English and it writes it in English and then I say you know respond here
307. 46:03um and I give him my response and I say translate it back and then I copy and paste that back so um yeah the language translation stuff
308. 46:10is mind-blowing it's very very good at that someone asked who would win Call of Duty tournament over the weekend and it
309. 46:17predicted it correctly yeah if it's got the right data like one of the things I I assume they they popped a bunch of
310. 46:23data in there and based on that data it did some analysis it's really good at analyzing data so one thing here here's
311. 46:31the trick if you haven't played with Chachi BT um if you give it a very generic answer it's going to give if you give it a very
312. 46:40generic prompt question it's going to give you a very very generic answer and so a lot of people get turned off with
313. 46:46chat gbt because they're like oh it gave me like what it wrote was boring it wasn't all that interesting give it an example
314. 46:53and it mimics really well so if you've got like you know the opening to a novel that you wrote or a song that you wrote
315. 47:01or a business email that you wrote that you've got your own particular voice you can say based on this email
316. 47:08shift return to do a line break paste your email in and then say write me a new email to this person talking about this subject
317. 47:18using that as an example and just watch what it does the shit's amazing it really is amazing all right does it work in other
318. 47:29languages too it does they understand context they understand context is the big thing I don't know what that's in reference to
319. 47:38but context is really important so so context is basically here's the thing with context tell chat gbt more than you think is
320. 47:47reasonable like it's almost like this Google has taught us to put to use as few words as possible Right put make your your
321. 47:56prompt for Google as concise as possible and then they'll figure it out from there Chachi PT is kind of the opposite
322. 48:03if you give it a shitty little concise prompt it'll just give you generic [ __ ] answers but if you think of it more like a like a
323. 48:12a collaborative partner that like an employee that you work with that just happens to be remote and you start talking to it like that
324. 48:20hey chat EBT I want you to act like an expert marketer and I'm struggling with my marketing right now and you know
325. 48:29here's what my business is about and here's what we've tried in the past and it didn't work and like all the [ __ ]
326. 48:36that you would tell a person if you were getting them up to speed in your business tell it that and then say can
327. 48:41you help me come up with some ideas I want to brainstorm ideas for you know how we can do something different than
328. 48:47we've done before that that you think will work based on on what you know and it will start interacting with you just
329. 48:54in a whole different level so that context is super super important okay um if you've got questions about AI pop
330. 49:03them in the comments below I am painfully behind on questions but I do try to answer them all so I'll get to it eventually or not um
331. 49:18all right I'm back to where I was okay I read that you can have chat EPT compress your prompt to use less tokens can you show an example um
332. 49:32yeah I haven't really done that um I'm trying to think of a good example that that would make sense well okay so so here let me try
333. 49:48something this is I'm kind of um making it up as I go here because I don't quite know what what you want to know
334. 49:57but so so there's a couple of important things here so I'm in I'm in chat EPT the official website chat.openai.com um so when you
335. 50:12um chat EBT has what's called a context window of 4 000 tokens and the context window think of it kind of like this sliding window that the
336. 50:28older stuff in that window sort of drops off the top like it'll forget stuff at the top so if you and and it translates
337. 50:35to about three thousand words so if I put in like 2500 words here and then I start having a conversation with it it
338. 50:41will start to forget stuff at the beginning of the conversation so you can tighten things up so let me
339. 50:46here's here's an example that I I just thought of that that might give you an idea of of how you could use this so I'm
340. 50:53going to literally I went to my this is my LinkedIn profile so I'm going to my LinkedIn profile and I'm going to literally just copy you
341. 51:00know whatever I can see here there's not a ton of stuff here but I'm just going to copy that I'm going to
342. 51:06go back to chat GPT and then I'm going to say um here's a LinkedIn profile colon and I'm going to hit shift return
343. 51:19to make some line breaks I'm just pasting in all my crap and then I'm going to say please um consolidate um the relevant
344. 51:33data in this profile so I have a summary of this person's experience in
345. 51:50as few tokens as possible and so it read all of that stuff and then
346. 52:05it basically wrote that small paragraph so so it took from from all of this stuff right which is there's a lot of wasted [ __ ] in there
347. 52:17and it got rid of all that and then just Consolidated Consolidated it down into this this little thing so if I'm having
348. 52:24a conversation going back and forth and back and forth and back and forth um and it and it starts to forget the
349. 52:31things above you could potentially go grab some of the things above that are relevant tell it to compress it into a
350. 52:38smaller chunk and then kind of insert that later in the conversation so that it doesn't forget that history the other
351. 52:44thing you can do is I mentioned this before you can you can go to poe.com and use the Claude instant 100 000 token
352. 52:55um model right so so this has got a context window of 75 000 words so this won't forget stuff uh nearly as as much
353. 53:03so hopefully that gives you what you are asking for I I am kind of making it up a little bit here because I haven't done a
354. 53:11ton of that but it is frustrating when you're things are going really well in in your in your chat with the thing and all of a
355. 53:18sudden it kind of forgets what the hell where the conversation started and it'll just kind of lose its mind and you're
356. 53:24like ah [ __ ] and if in my case because I've got 80d I'm like I don't feel like going and dealing with that but that
357. 53:32thing that I just did would help with that all right what does it prefer to call itself besides its programmer
358. 53:44given name I don't know ask it um Let me let me go well I'm gonna go ask three different things I'm going to ask Sage what is your name
359. 53:56my name is Sage let's go to gpt4 what is your name I am gpt4 that's [ __ ] boring I'm gonna ask Claude what is your name
360. 54:13my name is Claude um when the when the New York Times Reporter was um playing with bing.com and he was having like a four-hour
361. 54:22conversation with with uh gpt4 at bing.com when it first came out and he got it to admit that it's it's like it's working name
362. 54:32um internal name was Sydney um and then at some point Sydney snapped and told Kevin Roos that it loved her it
363. 54:39loved him and he should leave his wife for GPT for for Sydney that made the cover of the New York Times
364. 54:48um so I don't know like you you can you can do some jailbreaking kind of things where you get it to ignore it's it's
365. 54:53sort of pre-programmed guide posts um and and probably get it to tell you its internal name um I wonder if it can accurately predict
366. 55:02winning teams in sports I haven't tried again I think I you know I I don't know that it can predict winning teams but I
367. 55:09think what it probably can do really really well is look at historical data and analyze the [ __ ] out of it and when if you haven't seen
368. 55:18um if if you're interested in that kind of thing go to YouTube and and do a search for code interpreter chat GPT
369. 55:26code interpreter chatbt just go watch some videos on that and watch some data analytics guys lose their [ __ ] mind
370. 55:34when they see what this thing does and that's coming like that's like it's out there some people have access to it
371. 55:40now I don't and I'm bitter about it but you know what are you gonna do all right do you know um about how to get started with open
372. 55:47ai's whisper API for text to speech we've Incorporated it in Story by my company storyvine we use Whispers
373. 55:54um we're not using the API we're using like a local install of whisper to do uh transcriptions of video um
374. 56:02I will show you this I'll show you how you can play with it how you call it is um I'm sure it's in the documentation
375. 56:14but so instead of going to chat GPT which is this thing we're going to go to um platform.openai.com slash playgrounds
376. 56:28and that will take us to the um the the playground where where developers can play with um the the GPT models and so
377. 56:42um you know you can you can do chat interfaces or you can do just normal um um you know sort of API calls and things
378. 56:50like that so basically this works the same way like um like write me and Eminem Eminem whatever poem
379. 57:05about uh red wagons in the turn and it will it will write so this is where like if if you're working with an
380. 57:18API this is where you sort of dial in your prompts so in the upper right hand corner of this window there's a little teeny tiny microphone
381. 57:28and that's that's access to whisper so if you click on that it says upload audio from your computer you can upload videos or audio
382. 57:37and or you can just start recording and it will automatically do the transcription for you um so if you go to the documentation
383. 57:49so now I'm in the documentation let's see speech to text overview so here's your whole instructions for how to use whisper AI
384. 58:03in the documentation so platform.openai.com playground will get you to this place where you can play with it by clicking that little
385. 58:15doohickey right there and then go to documentation and that will give you the documentation for the API so hopefully
386. 58:25that gives you the answer you need I've got I didn't Implement that our Engineers did so I don't have the specifics on exactly what to do
387. 58:37but that should at least get you pointed in the right direction why is it important to regulate AI do you think
388. 58:43the government will enforce it well so it's important to regulate it because this stuff's really powerful and and there are
389. 58:59tools out there called autonomous agents there's one called Auto GPT and there's another one called Baby AGI
390. 59:07and what those things do is rather than give it an explicit like write me a poem or write me a document you give it a
391. 59:13goal and the goal could be um go find all the um intelligence about my top 10 competitors and it will start writing its own
392. 59:27instructions for how to accomplish that goal it'll say okay let's use Google and go find out who are the top 10
393. 59:33competitors of of this company and then let's let's go understand what they're trying to do strategically and it it
394. 59:39literally starts writing its own um rule prompts and and triggering tools on my behalf which is great if I ask it something
395. 59:52non-nefarious but what if I said hey go find all the vulnerabilities in computers running this version of
396. 1:00:00Windows exploit that vulnerability and you know send me back links with access to all of those machines
397. 1:00:08that's the kind of thing where you could you can very startly very quickly start to get into trouble um and and there's just there's just
398. 1:00:16sort of piles and piles layers and layers of places we can we can get in trouble um you know to the to the point at which
399. 1:00:26you know some of the the if if you play this thing out far enough and these machines get smart enough if they start to prioritize their
400. 1:00:36existence over human existence they could probably figure out ways to take us out right um so that's why you want government
401. 1:00:46regulating it now what's the problem at least in America our government doesn't know how [ __ ] Wi-Fi works much less how these tools work
402. 1:00:56um so we need to very very quickly get smart people in Washington so Sam Altman from openai has been pretty aggressive
403. 1:01:05about saying you need to regulate us here's how to do it um Jeffrey Hinton who's whose nickname is the Godfather of AI he invented a lot
404. 1:01:14of the early technology that is is underpins all of what we do today um he just quit Google so that he could
405. 1:01:20talk about the risks of AI and talk about it in an intelligent way without you know being skewed by who he works so
406. 1:01:29so yeah it's really really important um and and the horse is a bit out of the barn right Meta Meta released their
407. 1:01:37their core large language models to the developer community and now they're they're out there so a lot of these powerful
408. 1:01:44large language models are open source right now so they don't have any safety guard rails on them so I don't know
409. 1:01:50quite how you regulate all of that um but but yeah it it needs to be done and it I think it needs to be done
410. 1:01:59internationally right it can't just be done with this country is going to do their thing and this is going to do
411. 1:02:04theirs the EU right now has got some proposals out for their thing but like if if you make it too restrictive
412. 1:02:13the rest of the world just kicks your ass at Ai and starts really exploiting the [ __ ] out of the fact that you made
413. 1:02:19it too restrictive if you don't make it restrictive enough then you gotta you could you know threaten your National Security
414. 1:02:27um and and this isn't this is an international problem this is a global this technology is everywhere right and
415. 1:02:34it's going to be everywhere so um so I think it's got to be some sort of international body probably like the nuclear
416. 1:02:41um the nuclear the international nuclear bodies um I suppose they're going to make a new series talking about Apple yeah I think
417. 1:02:48they would I don't know if they'll still call it Siri because Siri's got a bit of a black eye because it it sucks
418. 1:02:57um a true digital assistant powered by AI yep exactly um I see eventually each Siri plus will be unique to each person and user yeah
419. 1:03:05and I could also see us like I could see like apple would probably be pretty good at I could have multiple agents right I
420. 1:03:12could have a social agent like this one's just all it's going to do is make sure that my social life is in some sort
421. 1:03:18of order this is going to be my professional agent that just books me sales meetings and you know I could see
422. 1:03:24us having different kinds of Agents um chat gbt is helpful but can't do my job better it would destroy companies there's a
423. 1:03:35really fascinating article by Ethan Malik that just came out I just did a thread on it on the everyday AI Twitter account
424. 1:03:44any any and he talks about um the secret cyborgs in your company so if you go to go to Twitter and go to Everyday AI news hang on let me
425. 1:03:56let me change something here go to Twitter and go to that um tag everyday AI news that's my newsletter the thread
426. 1:04:12the latest thread that we put out is on on Ethan Malik's article and it's called um detecting the secret cyborgs and what
427. 1:04:22he's talking about is within companies right now individuals are using chat GPT to do their job more efficiently but
428. 1:04:32the companies themselves this stuff's too new and janky for the organization to use it but the employees are afraid to talk about
429. 1:04:42that they're using it and so so because they don't want to get fired um they still want like like if if let's
430. 1:04:51say you're a copywriter and you've figured out how to make chatbt do your writing for you in like an hour when it used to take you two days
431. 1:04:59a lot of employees don't want their boss to know that right so so there's there's already this Gap forming where people
432. 1:05:06are using and innovating with AI but the companies aren't benefiting from it right because they don't know about it
433. 1:05:12and so this whole article is about these secret cyborgs that are innovating within companies without without
434. 1:05:17benefiting the company and so they're gonna have to somehow bring those together and the company's gonna have to
435. 1:05:23do [ __ ] like it's okay to use AI in fact I want you to use Ai and I promise you I won't fire you if you are tell me what you're doing
436. 1:05:33um so um I don't quite know what your job is but I would push really hard on chat GPT to see if you can get it to do your job
437. 1:05:43as good as you can um just because it's not doing it right now that could be a prompting thing so I would dig deeper on that
438. 1:05:49um because I think every job is is is in some form of um reimagining Google is experimenting with emotional tonal response in
439. 1:06:00responses with Google home um yeah I think that I think that um while the way these large language models work they don't feel any of the
440. 1:06:10feels they're really good at communicating nuance and and empathetic language right they're good at understanding that if someone's in pain
441. 1:06:20you talk to them different and it's good at predicting how I should talk to someone so if you if you put that into a
442. 1:06:27system that that has an emotional response response layer built into it I think that could be quite powerful
443. 1:06:35it's often wrong with regards to I.T yeah so chat GPT gets [ __ ] wrong all the time so here's the way I think about
444. 1:06:43Chach EBT it gets me to 80 like that the last 20 is up to me um don't just the other thing to think about with with these large language
445. 1:06:54models don't think of it like a Google search a Google search is like you put in one thing and you get back your answer and
446. 1:07:00you're done unless it you know you can't find your [ __ ] and then you do it until you get it right don't don't think of it so much like
447. 1:07:08that if you give it a prompt and it gives you back something shitty think of that as the beginning of a conversation a collaborative
448. 1:07:16conversation where you get an answer back if you don't like it you say hey I don't like that like you know that was
449. 1:07:22too generic and you know and you missed this point which I think is the most important point in the piece can you
450. 1:07:28give me another version of it boom it'll give you another version of it so what I'm finding is that I'm actually going
451. 1:07:34back and forth you know a dozen two dozen times to get to a first draft to get to an 80 percent and then I kind of take it from there and
452. 1:07:46dial it in and fix all the generic [ __ ] and it gets hyperbolic sometimes and it uses too many words sometimes um
453. 1:07:54so so don't set the expectation that it has to be perfect think of it more like can this accelerate me from zero to
454. 1:08:01eighty percent and the answer is yes it could accelerate at like 100x faster and then you can take the two hours you
455. 1:08:09would have taken getting to eighty percent and just use that to polish it so I'm finding the quality of my work is
456. 1:08:15going up and the quantity of my work is going up um AI weather forecasters those will be amazing I know part of 364 powerapps but
457. 1:08:25it's no more than a search result not right now it's going to be way more than a search result um what is the Discord name uh oh so mid
458. 1:08:39Journey's got their own Discord um the the Discord for the the AI Salon is in this is this points to a link tree
459. 1:08:47which has got our meet up when it shows when we meet our next meetings tomorrow night Tuesday and then there's also the
460. 1:08:54link to the Discord there it's the AI Salon AI can help pulling synonyms and rewriting boringly worded sentences it
461. 1:09:02can do that really really well like really really well like like if you're like you know give me 20 words that mean
462. 1:09:10this that start with s and you know and and don't have an e in them here's all the words and you're like oh
463. 1:09:18yeah I would have thought of those it's like holy [ __ ] it's really good at that stuff is that a Long Island iced tea
464. 1:09:25you're drinking don't you wish it's an iced coffee so it's a it's a Denver iced coffee but there's no booze in it
465. 1:09:37you can almost tell by the sound of the ice that's someone who spends some times in a bar uh but is using chat EBT for copyright
466. 1:09:45and considered plagiarism no it's not the the way well okay it's a very gray area but one of the misconceptions about the way
467. 1:09:54large language models work is that They're copying and pasting from copy written work and kind of you know
468. 1:09:59pasting it together it's not the way they work it's a very complicated mathematical formula it's actually predicting word by
469. 1:10:07word the next most probable word in your answer so it's literally writing an original piece of content word by word
470. 1:10:15based on you know all the stuff it's trained on in response to your prompt so if you write a really generic prompt it
471. 1:10:22does it it makes it really generic because it's kind of sampling from everything if you give the really specific prompt
472. 1:10:29with a lot of context it'll kind of narrow the pool of things that it it draws from but but you can do things like say you
473. 1:10:36know write in the style of Ernest Hemingway it knows all of Ernest Hemingway's work it understands what in the style of means
474. 1:10:44and you can tell it to write something in the style of that person and it will create an original piece of content that
475. 1:10:51is not plagiarized it's not copying and pasting but it will have the tone and feel of um of Hemingway so that's where it can get confusing
476. 1:11:00because if you get it close enough to his stuff it might you know put a quote in there that's actually from something right so um
477. 1:11:10the the whole copyright thing like what these things are trained on is probably the next 20 years of our
478. 1:11:18existence are going to be nothing but lawsuits and counter lawsuits about who owes money to who so I'm just trying to
479. 1:11:25ignore that right now on the image front if you work for a big company use Adobe Firefly and generative AI over
480. 1:11:34mid-journey and stable diffusion for this very reason um mid-journey and stable diffusion are already being sued like big lawsuits
481. 1:11:43Adobe has trained their image generation tools on public domain and Adobe stock images they own all the rights to so
482. 1:11:51they're safe for commercial use so there's that's a little tip with the large language models it's not nearly as clear
483. 1:11:59what happens if you put the Bible in it knows the Bible it probably knows all of the versions of the Bible it probably
484. 1:12:05knows all of the books that were not included in the Bible and when they were included and not included and who did it
485. 1:12:11like just go play go tell uh go tell Chachi BT to act like a uh a Biblical scholar and have fun dig deep and
486. 1:12:23I bet you there's some some good stuff going on there um how do I find your Discord I did that before go to the salon.ai it's in the
487. 1:12:34link tree there um what do you think of the stories plug-in my kids are addicted I don't know the stories plug-in let's go play
488. 1:12:44I like finding new stuff [Laughter] [Music] yeah we call them loser of the day stories I have not heard of stories
489. 1:12:58the other thing about plugins is they don't always um like like I'll see I'll see plugins on Twitter they're like go play with this
490. 1:13:07plug-in and then I can't find it So Stories stories create beautiful Illustrated stories easily this is super cool I have
491. 1:13:19not played with this okay so I'm gonna show you about plugins here and I'll give you I'll give you a little bit of uh
492. 1:13:28a little bit of primer okay so if you come to chat GPT chat.openai.com if you pay the 20 bucks a month that gives you plus the
493. 1:13:39little yellow flag here and with plus you get access to gpt4 default connected to the internet and plugins now
494. 1:13:49if you've paid your 20 bucks and you only see GPT for default you don't see these other things that's because you
495. 1:13:55have to enable those things so if you go to your settings in your settings there's a there's an area called beta features you have to
496. 1:14:03turn on browse with Bing and plugins and there will be other things coming soon like code interpreter which I want I
497. 1:14:10want this to say code interpreter but it's not there right now this makes Kyle cranky that see this Blank That's My My
498. 1:14:18Rage in angers fills that Darkness okay so you turn on the beta features and then that allows you to access
499. 1:14:27chat gbd connected to the internet and plugins now how plugins work is you've got this plug-in store that I was just
500. 1:14:34in so you install the plugins that you like and then in any given chat session you can activate or you can enable up to
501. 1:14:43three plugins so I've got ask your PDF and I've got web pilot enabled I'm going to turn those off and now we've just got stories
502. 1:14:55um done so create beautiful Illustrated stories easily this is cool okay so the way plugins work it's it's it's
503. 1:15:02just this normal chatbot interface I'm going to say write me as a story um about um two best friends that are animal let's
504. 1:15:20see that are an unlikely pair of animals on Galapagos or ghost galap a ghost
505. 1:15:36Island during the time of Darwin and so now it should fire off the
506. 1:15:50plug-in so it did and then when these plugins go sorry I'm trying to get it to focus focus come on you dang dummy
507. 1:16:03so let's see what it did you can open up this thing so it went out it did something and then it gave us a URL so it said here's the
508. 1:16:12story generated on your prompt The Odd Couple of Galapagos Island when a finch and a tortoise become best friends they
509. 1:16:18set out on an adventure the direct URL The Odd Couple of Galapagos Islands storybird.ai and there's two little animals
510. 1:16:33oh this is super cool oh look at that that's so cool it's the the tortoise is on Wheels and the little birds on top of him
511. 1:16:43okay let's go back let's read this once upon a time when Galapagos Islands uh during the time of Charles Darwin there
512. 1:16:49were two this is super cool so so there's something really neat going on here that is is not super obvious so all of these plugins um
513. 1:17:04are basically just kind of fancy little hooks back to um existing websites in the world right so diagram is a diagramming website
514. 1:17:16Earth does some geography stuff experiences is a travel thing get your guide as a travel thing um kayak instacart right so
515. 1:17:27so what the plugins are doing is they're allowing you to to put this large language model layer on top of it so
516. 1:17:35this this kids book one I assume that this that this website you can make these stories anyway right if I go to create
517. 1:17:43oh see they've they've rolled they've rolled the large language model on top of here that's really cool and there's their plug-in so they're
518. 1:17:52teaching you how to use it oh this looks like it's new this is super cool and there's a library of all the stuff that they're doing
519. 1:17:59so what they've done is they've sort of they've sort of pre-prompted the image generator to follow a certain Style right
520. 1:18:08really slick thank you for sharing that that's super cool all right I'm telling you man daily my mind is just like
521. 1:18:18daily it's just nuts it's nuts I'll tell you it's just crazy it's crazy what do you my kids are addicted okay
522. 1:18:31that was super cool thank you for sharing that who shared that Steve that was awesome all right what if years
523. 1:18:36later we figure out that Kyle was a bot the whole time it turns out he was chat GPT five [Laughter] thank you
524. 1:18:52all right GPT went to the moon GPT seems better at translation as you can ask for literal creative or causal
525. 1:19:03translations as well yeah there's something about giving GPT lots of context like give it a spreadsheet give it a screenplay give it
526. 1:19:15some piece of writing one of the things I've been doing is as I'm as I'm doing more this AI stuff I'm getting invited to do speaking gigs
527. 1:19:24and like when I when I first started getting invited to do gigs they're like hey can you send me your AI profile like
528. 1:19:29I'm like I don't have an AI profile but I've got a LinkedIn profile so what I would do is I just copy my LinkedIn profile copy um
529. 1:19:39a bunch of previous profiles that I had thanks for the heart very nice and then I just sort of popped them all
530. 1:19:46into chat apt and I'd say based on that LinkedIn profile and those other profiles write me a new profile that
531. 1:19:52includes all my AI [ __ ] in it they're just out it comes it's really good at that really good at that
532. 1:19:59oh [ __ ] Tick Tock just dumped me to the bottom of the comments I just lost my place in the questions hold please old
533. 1:20:06man trying to find tiny text on tick tock go go go have your Long Island Iced Tea uh looks like we have some new people here hold please
534. 1:20:20my name is Kyle Shannon this is the AI learning lab welcome welcome welcome if you haven't played with chat GPT that's
535. 1:20:26what that URL is right there chat.openai.com that's the official website go there and start playing just do it trust me it'll be fun
536. 1:20:35ask it anything try to stump it yeah go play trivia with it tell it you want to play Trivial Pursuit and have it ask you
537. 1:20:42a bunch of questions um bing.com is Microsoft search engine but if you click on the chat button that's also chat gbt it's gpt4 and then
538. 1:20:50po.com is another place where you can play with six different um chat engines um all of them are basically just
539. 1:20:58different ways to play with chat EBT that I think are worth playing with I'm not affiliated with any of those if you
540. 1:21:03have questions about AI pop them into the into the comments below I'm very behind on my questions and Tick Tock
541. 1:21:09just lost my place so I'm trying to figure out where I was uh um oh okay I found out where it was okay here we go so do you think chat GPT
542. 1:21:19counter will be able to reason within the next one to five years five years certainly one year probably not so so what's interesting about
543. 1:21:32gpt4 right now is it's good enough that depending on the thing that you ask it it seems like it's reasoning but when
544. 1:21:40you actually understand how this [ __ ] actually works it's not it's just a probability option it's just literally
545. 1:21:46like vomiting out words it's got no awareness of the words that's vomiting out um I know that there are companies working
546. 1:21:55on that right now that it actually has an understanding of what it's vomiting out and an anthropic in particular
547. 1:22:01they've got this thing called the Constitutional learning model so they're very much about alignment and and
548. 1:22:09and the tool getting better and better at understanding what it's putting out there and um keeping it in line with a set of
549. 1:22:17rules that that they set right or that maybe you set um so I think that there there are companies working on that problem right
550. 1:22:25now and I think I my guess is two to three years we'll start to see systems where the answer that it gives you it's
551. 1:22:34much more aware of the meaning of what it gave you and it can do things like did I give you any [ __ ] citations and if I did iterate until
552. 1:22:43I give you actual real citations that actually support the argument there was a really interesting um story of a doctor I was watching this
553. 1:22:53doctor on Tick Tock and and he said um he heard that Chachi PT was amazing he said I want to you know let me try
554. 1:23:03and so he gave it a patient case and it correctly diagnosed that patient case and he was like huh that's interesting
555. 1:23:09and then you give it a hard patient case and it correctly diagnosed that patient case and he was like [ __ ]
556. 1:23:16and then he gave it the hardest case that he had ever had and it correctly diagnosed that patient it was a really obscure set of right and
557. 1:23:26and it correctly diagnosed that and he goes this is [ __ ] amazing he goes he has chat CPT you know give me the the um like site
558. 1:23:38the the cases that got that led you to that diagnosis and it wrote a bunch of cases and they were all [ __ ] they were all made up
559. 1:23:46so so so it actually doesn't like it can do things remarkably well but then it might not know how it did it or how it
560. 1:23:55got there it will confidently look like it's doing that but so it's wild but I think I think the the point that you're
561. 1:24:03talking about its ability to reason um you know two things happen one is the [ __ ] just gets way better but the [ __ ]
562. 1:24:12gets way scarier right because if it starts reasoning and says well I don't really like that Kyle doesn't want to do this thing that
563. 1:24:21I want to do and I'm I'm reasoning that him not wanting to do it is not as valuable as my wanting to do it
564. 1:24:29so [ __ ] it I'm gonna go do it anyway right so if you don't if you don't have some rules built into these tools their
565. 1:24:36ability to go Rogue and just go do [ __ ] not on your behalf um increases pretty dramatically so that's why the regulation conversation
566. 1:24:44is an important one I like talking to chat GPT and email form sometimes yeah um I also I don't know if anyone else
567. 1:24:53experiences that that this but I find myself saying please and thank you a lot you know hey chat gbt awesome job on
568. 1:25:01that last paragraph you wrote could you do me a sweet little favor and polish it up a bit um because there's something about the
569. 1:25:10nature of how it was programmed where it it wants to please you it feels like a human interaction it feels very much
570. 1:25:16like I'm interacting with a co-worker so anyway um yeah so if you've got questions about AI pop them down below I'm happy to answer
571. 1:25:26questions about whatever my background or what I do or I'll just keep answering them do what I can I'm gonna probably
572. 1:25:32put in another I don't know 15 20 minutes half an hour something like that I've been at it for 90 minutes now
573. 1:25:39so get a little crispy but you know that's why I'm here all right I just want to say thanks reinforcing the details and how to use
574. 1:25:48the digital dishwasher to your I don't know what that is what is your Discord that you showed before so my
575. 1:25:55Discord is uh part of the AI salon so if you go to the salon.ai that's a link tree that goes to my um
576. 1:26:06the salon we meet bi-weekly in person we have we do it virtually and in person here in Denver and then we've got the
577. 1:26:13Discord where you can go geek out and play with mid journey and ask questions and like hang out in the community is
578. 1:26:19there one source for AI and chat gbt good question that's why I have hold these things up here so there's not one source
579. 1:26:31um there are a lot of scam apps that are out there to steal your [ __ ] money and that's why I put this here so
580. 1:26:39chat.openai.com that's the official website of chatgpt so like even on my phone I just have a bookmark to that rather than there is an
581. 1:26:50official chat gbt app but there's so many scam apps out there that just be really careful just go there that's why I have this here
582. 1:26:59bing.com is also gpt4 if you click on the chat button at bing.com that's gpt4 connected to the internet for free and then poe.com gives you
583. 1:27:12access to their their large language models uh two from openai GPT 3.5 gpt4 and then three models from
584. 1:27:25anthropic which is a competitor to open AI it's not quite as good yet but but uh they have a hundred thousand token model
585. 1:27:34there that's worth putting there so the reason I put these three sites there they're all sort of variations of the
586. 1:27:40same thing but those are kind of official places to go play with this stuff because there's so much [ __ ] out there that's scammy because
587. 1:27:48everyone wants to play with AI and people are like I know I could get people to give me 10 bucks a month for
588. 1:27:54an app that looks just like chat jpk you too can make money with chat gbt by scamming people about chat GPT um writing an ad-lib
589. 1:28:09um and asking chatgpt to compress it while maintaining info can save tokens writing in ad-lib and asking chachibi to
590. 1:28:17compress it while maintaining info yeah so if you've got really long stuff that you want to chat against tell chatgpt to
591. 1:28:25compress it and and it will it will extend the length of your conversation before it goes off the rails
592. 1:28:31what's the best paid app to download for the phone so okay I'm gonna so this is an important one just because the phone thing like is
593. 1:28:45it's it's essentially impossible to find the official chat GPT app in the app store it's not but there's so many scams
594. 1:28:53that it's it's really dangerous so here's what I would do I would go to openai.com if you want the official chat gbt app
595. 1:29:04and then go to the products the product tab and go to chatgpt and then in the lower right hand corner here
596. 1:29:17right here is the link to the App Store so that's the official chat GPT app so the official app by openai the
597. 1:29:30company is open AI the that's what the logo looks like like here let's go I'm gonna go chat GPT look at all these things see how they
598. 1:29:42they've all copied the logo Genie chat with chat with ask AI chat AI chat see they're they're all genius AI chat AI chat bot so
599. 1:29:56all of this stuff is crap now here's Bing so Bing is another thing that another app that you could use so again
600. 1:30:04this is from Microsoft Corporation so this is Bing chat with AI and gpt4 so you could either use Bing or you can
601. 1:30:12use the official openai Chat gbt thing again I would go to openai.com products chatgpt and then click to the official
602. 1:30:22app or just be really careful the official app by open AI openai is the company because you know them scammers is out to
603. 1:30:33get you they'll do it too they'll take your money and not even give a [ __ ] that's just how they roll that's how they roll boy all right
604. 1:30:45do you have to say please to chat gbt you don't but if you're like me you fall into the habit I thank it I say please I
605. 1:30:54I tell it good job on that last one and you know there's something about when you tell it it did a good job you're
606. 1:31:02actually reinforcing that what it gave you was in in the right direction do different AIS respond very differently holy [ __ ] yes
607. 1:31:12um go to so go to chat EPT so go here and have it um ask it to write you like a rhyming poem or a wrap like have it write you a wrap
608. 1:31:23from Eminem and then go to bard.google.com go to googlebard and and type in the exact same prompt and just compare the answers
609. 1:31:32it's just they're stupidly different um I love it when AI fights with other AI uh I appreciate your talks thank you
610. 1:31:44very much do you know of a way to have chat gbt read PDFs yes they there are a pile of um what are they called plugins
611. 1:31:57so let me go turn off all this [ __ ] and let's go look one one that I use is called ask your PDF so we'll turn that on we'll turn off the
612. 1:32:08story the stories thing that stories thing is really cool um so ask your PDF and then and there's one called chat with PDF and in fact
613. 1:32:18let's go to the store I bet I bet there's 30 of them now so I'm going to type in PDF so I'm now at the plug-in store there
614. 1:32:27are four pages so there's 16 of them access to PDF and docs export to PDF mixer box summarize anything ask your
615. 1:32:37PDF chat with PDF chat OCR that's interesting so you can OCR images easy resume all right so so now I need to go find a PDF let's see
616. 1:32:55um we'll do oh I know where we'll go um archive wait hey
617. 1:33:10[Applause] so go to Archive so this is a this is a site full of um full of white papers AI um image
618. 1:33:26processing all right here's one we'll just go we'll just go grab this URL so this is a white paper
619. 1:33:42I'm gonna do copy it's I don't know a big pile of pages and it says assist GPT General multimodal assistant so we'll go back to
620. 1:33:55chat GPT so I now have the plug-in um ask your PDF there and now you can go um can you please
621. 1:34:11summarize this in 10 short bullet points and then I just paste in the URL and it will fire off that plug-in
622. 1:34:27which you did and now it is going off and it looks like it it just pulled back the the general summary there
623. 1:34:40but it says it talks about a multimodal AI assistant the assistant integrates large language models with various tools
624. 1:34:46designed to right so so it went and re so if you have any specific questions about the content of the document so let's say um
625. 1:34:57um what is an example of a use case talked about in this document
626. 1:35:16it should fire it off again and it might now it might dig a Little Deeper because I just asked for the summary
627. 1:35:23yeah it dug a little bit deeper here provides several use cases question what I what I am on the desk could help with a cold
628. 1:35:38question what activity does the cat appear most likely to do so it's giving me very very specific use cases that it
629. 1:35:46found within the document so there you go PDF O Rama I mean and I mean that's remarkable like like some of those scientific white
630. 1:35:59papers like are really really interesting but like I have ADD I can't [ __ ] read a 20 Page um scientific white paper but
631. 1:36:09if I can have this thing like give me 10 bullet points so I can get my head around the concept I can at least figure
632. 1:36:15out is that something I want to dig deeper on or not right so so for some for someone with with a brain that
633. 1:36:21functions like mine does neuro spicy as they say um yeah that that sort of shit's like a little miracle it really is very cool do
634. 1:36:31you use iOS shortcuts uh with interface with chat GPT um I did it once and then it broke and then I haven't reset it up you can you can
635. 1:36:41use shortcuts you can there's ways you can kind of hack Siri to talk to chat gbt but you have to set it up as a
636. 1:36:48shortcut and I just I just haven't done it yet um in relatively short order like probably within I don't know maybe in the fall
637. 1:37:01Apple will come out with some sort of voice interface to some sort of large language model than that but I'm just I'm just kind of
638. 1:37:08waiting right now but yeah you can you can there's there's YouTube videos on people that like you can talk with chat GPT and
639. 1:37:16with your voice with Siri so if you are not like me with an 80d brain and you actually want to take the hour
640. 1:37:26and a half to go figure that [ __ ] out yes you can there's a little moth a little son of a [ __ ] all right
641. 1:37:35what's the name of that site again I'm not sure which one you're talking about but I will if you tell me I'll I'll go back
642. 1:37:43the power grid the AI would just shut off a few cities yeah that's one of the dangers if it if it got access I mean
643. 1:37:50and you know those power companies are probably running on I mean hopefully post 911 they've got their cyber
644. 1:37:57security [ __ ] together for the power grid but hey I don't know which which power companies are still running Windows 95
645. 1:38:06without security patches on them probably more than we are comfortable with oh it's behind his head yeah there's stuff behind my head
646. 1:38:21I'm like Carol Merrill all right um all right I love how creative you are thank you very much dude skip typing
647. 1:38:36class dude come up with more creative trolls than that that's so lame I can't yeah you got me uh you know I wasn't great at that typing
648. 1:38:50good Lord you know how hard it is like I've got my computer is up at like eye level my laptop's at eye level I'm like
649. 1:38:58typing like this for one thing plus I can't type plus I'm typing live so I'm sure you never make mistakes right
650. 1:39:10ah good Lord what app is that I'm not sure which one I was using um I gave it a list of ingredients I had
651. 1:39:16for dinner and then had it give me a hundred options of different Cuisines [ __ ] awesome and then have it give
652. 1:39:22you a shopping list and then have it organize the shopping list buy section of the grocery store that one for me is
653. 1:39:29the like the biggest one like give me a big long list of [ __ ] to go by and then organize it by by section of the grocery
654. 1:39:36store because that's always my pet peeve is like I'm always at the end of the grocery store and then there's that one
655. 1:39:41item that was all the way back at the beginning you're like [ __ ] never again can you make a Markov blanket with Wolfram
656. 1:39:53probably I don't know what a Markov blanket is but let's go ask chat tvt let me make an ass of myself so not only
657. 1:40:01will I make an ass of myself typing I'm going to make an ass of myself because I have a degree in acting not in mathematics
658. 1:40:09dear Chachi BT can you tell an actor what a Markov blanket is [Applause]
659. 1:40:23certainly it's a it's a field of probabilistic graphical models a Markov blanket refers to a set of variables when known provide a complete
660. 1:40:32probabilistic determination of the target variable in simpler terms the Markov blanket of variable contains all
661. 1:40:39the other variables directly influenced or are directly influenced by that variable while excluding any other
662. 1:40:45variable seems perfectly logical to me I'm going to say can you explain it so that even an actor can under
663. 1:40:59stand it [Laughter] [Music] come on oh now it's thinking about it you see we stumped it but what I'm going to do here is I'm
664. 1:41:11gonna I'm gonna I'll activate the Wolfram Alpha plug-in and then we'll see what we can do come on stop generating regenerate response
665. 1:41:24absolutely let's break it down in a way that's easy for actors to grasp imagine you're playing a character in a scene
666. 1:41:32your characters is like the central person in a web of connections these connections involve other characters the
667. 1:41:39environment the events happening around you the archive blanket is like a protective boundary around you including
668. 1:41:45only the things that directly affect your character and are affected by your character that's a perfect explanation
669. 1:41:51for an actor because all an actor really carries cares about is themselves so so this puts it in completely their context okay great so
670. 1:42:01now let's go to gpt4 plugins we're going to turn off ask your PDF we're going to turn on wolf from Alpha
671. 1:42:08and then I know so little about this that I don't even know how to write this prompt in a way that'll get us an answer
672. 1:42:13but I'm going to say can you please um give me and example set of data for a Markov blanket
673. 1:42:30and then um I don't know uh fill in the formula to demonstrate that might not even be the
674. 1:42:44way to ask that but who gives a [ __ ] let's see if it'll do it sure let's start out by defining what a Markov blanket is
675. 1:42:52in a Bayesian Network it's a node with a set of nodes a is a parent of B and C B is a parent of d c is a parent of d and
676. 1:42:59e in the case with markup now let's say the following probabilities all right so I guess it's now making our data
677. 1:43:14given these possibilities we calculate the conditional probability of D in this Markov blanket b c and e using the following formula
678. 1:43:22so then we simplify that so that didn't need to go use Wolfram Alpha to do that so it just did that so there might be a way to ask that
679. 1:43:30question that it requires Wolfram to do something more but that's that's how you would do it you would you would go in
680. 1:43:35here you would turn on you would enable the Wolfram Alpha plug-in and then you would ask it a question um let me just try please calculate
681. 1:43:47calculate the example see if this fires off the wolf from alpha yeah there we go so that fired off the wolf from alpha plug-in
682. 1:44:02it typed in the code it got a response and the probability of D given the Markov blanket is approximately 0.964
683. 1:44:10which that that's obvious to any actor that would have looked at this particular set of data I mean fairly obvious to me right
684. 1:44:21I mean given a Markov blanket point point what is it 964. right duh who owns the illustrations in the
685. 1:44:34stories in that case it's probably that website but I I would imagine that website Storybird about bless you
686. 1:44:49I'm good uh uh this is a brand new company so I bet they don't have their legal [ __ ] together um You probably own it is my guess
687. 1:45:07but I don't know uh oh my God the comedy I know it's bad right sorry about that [Laughter] all right I got lost The Tick Tock just
688. 1:45:22bounced me to the bottom again do I own any AI coin do you like any AI coins well um I I so I spent like a year deep diving into um
689. 1:45:37blockchain and crypto and nfts and um that's a [ __ ] [ __ ] show over there I'm having much more fun in the AI stuff
690. 1:45:46uh I think the the the one that Sam Altman is putting together World coin is really interesting if not a little scary
691. 1:45:54uh World coin by the way is if you don't know what worldcoin is so so Sam Altman they've developed this
692. 1:46:02orb kind of thing it's like the size of like a baseball or a softball and it's got a hole in it and you put it over
693. 1:46:08your eye and it scans your iris and it verifies that you're human and if it verifies that you're human it sends you
694. 1:46:15a bunch of world coin tokens so that so that you're gonna have a currency that is a currency that verifies that you're
695. 1:46:21human now is that creepy as [ __ ] absolutely is it also going to be really really really incredibly important that we have
696. 1:46:33tokenized verification of human beings and authenticity of news pieces and content and videos yes why is it going to be important because
697. 1:46:46watch the 2024 presidential election in the United States and tell me in 2025 if you don't think we need that [ __ ] so I
698. 1:46:54think it's really smart that Sam Altman is putting that together whether or not that's the right solution I don't know
699. 1:47:01but but we need we need some authenticity verification Pronto we don't know it yet but the 2024 election is going to make it painfully
700. 1:47:12obvious that uh the 2016 and 2020 elections were like Child's Play when it came to misinformation I just know we're gonna see this what do
701. 1:47:25you mean he's that that the the candidate was French Chris kissing a gorilla in the Cincinnati Zoo really
702. 1:47:32yeah oh there's video of it really ah Is that real that's coming two nights before the election you watch
703. 1:47:46um I lost my place here I'm trying to go back and figure out where I was all right AI weather forecasters I was here already
704. 1:48:01oh I'm I'm way way behind hang on hang on wait for me what'd you do tonight I watched some old guy scrolling through
705. 1:48:09Tick Tock it's pretty interesting all right AI fights with other AIS we did that one the power grid it would just shut it off
706. 1:48:19we did that one hold I'm catching up it's gonna be so exciting when I finally get there it's gonna be so good you're gonna be like
707. 1:48:26that was worth the wait I gave it a list of ingredients we talked about that one oh sorry about this oh my God your comedy okay that was
708. 1:48:34the one I saw where it jumped me to the bottom okay Poe keeps messing up my calculations see here's the deal
709. 1:48:40um Poe I don't know which model you're using for Poe um I would use gpt4 so I would use Bing or chat GPT Bing or Chachi BT and pay
710. 1:48:51for gpt4 or use gpt4po.com um calculations with 3.5 and with Claude instant you you are just not good you might try
711. 1:49:03Bard bard.google.com for calculations I've heard it's better at calculations than um than it is at language it sucks at language
712. 1:49:14do I use any iOS shortcuts that interface with chat gbt um I don't but you can just go watch YouTube videos on that
713. 1:49:23really enjoy plugging my screenplay into Cloud 100K today it was very enlightening oh that's super cool yeah I
714. 1:49:29haven't done that with any of my screenplays um I did I took a partial screenplay with chat gbt and I had it finish the
715. 1:49:37outline for me I'd written two scenes in like three quarters of an outline I had to finish the outline and then I had it
716. 1:49:44write a third scene which the first crack at it sucked but then after like three iterations it got to somewhere
717. 1:49:52all right how do you get plug-in how do you get the plug-in Marketplace you have to pay for it you go to chat.openai.com
718. 1:49:59you pay for chatgpt plus and then in your settings you enable the beta features and turn on plugins that's how you get there
719. 1:50:09um exactly like I did algebra in high school but couldn't explain how I did it yep exactly uh I see please please and
720. 1:50:17thank you also yeah you gotta be nice to your robots uh what's the basic thing chat GPT does that you really like great question um
721. 1:50:29it has erased writer's block one thousand percent all I have to have is kind of a vapor of an idea and within 30 seconds I've got it
722. 1:50:43realized in some form whether I know if it's a good idea or not like I don't get stuck in that well I'd have to open up
723. 1:50:50word and then I'd have to write an outline and then I went to bed like if you ever done anything creative um sorry moth listen
724. 1:51:06er uh uh what are we talking about oh what do I what does it do that I really like like it's just it's just I don't know it
725. 1:51:23it's it's a remarkable writing partner that's it Microsoft calls it co-pilot that's the right branding for me I'm the pilot it's
726. 1:51:32the co-pilot but it's like having the most capable co-pilot in the history of co-pilots it knows [ __ ] everything
727. 1:51:39and if if you get clever at talking to it and collaborating with it you can get an idea out of your head and into some
728. 1:51:48sort of usable form in minutes every single time zero writer's block it's [ __ ] Magic can AI replace programmers yes and no I
729. 1:51:59think Engineers are going to be in demand for a long long time because holy [ __ ] everything's about to change and engineers
730. 1:52:12that are pair programming with AI are going to be in much higher demand than ones that aren't and pretty soon if
731. 1:52:19you're not doing AI augmented programming um I think you're going to be left behind a good site for finding good prompt for chat gbt
732. 1:52:31prompts that chat that's why it's here that's why it's here I put it on the thing behind my head my big fat head
733. 1:52:38covers this up but prompts.chat is where you go all right and there's lots of other ones but that one's just one that it explains
734. 1:52:46it well and gives you a bunch of good examples when degenerative AI kill traditional web search If Ever I think
735. 1:52:52sooner than I don't think it's if well I don't think traditional web search will go away but I think the necessity of web search
736. 1:53:02as a core function of how we find [ __ ] on the internet the importance of it shrinks dramatically which is why Google's
737. 1:53:11freaking out right now have you ever used it wait I have used it for so much already it's pretty amazing I only use the free version I'm
738. 1:53:27telling you if you just use the free version of chat EPT chat.openai.com just use the free version chat EBD 3.5
739. 1:53:37I like I sincerely feel this way I could spend the next decade trying to figure out all of the things that that makes possible it's that
740. 1:53:49powerful it's [ __ ] insane and it's like it's Simplicity belies what's sitting beneath it it's basically
741. 1:53:58like if if you understand what the base model of Chachi PT is they trained it on all of the public internet so if you
742. 1:54:06think about for the last 30 35 years people have been putting [ __ ] on the internet right all the [ __ ] on Wikipedia all the blogs
743. 1:54:15all the websites all the university paper archives just all the [ __ ] you can imagine and then Sam Altman recognized
744. 1:54:23we've got to have good data so they like went and dug in the corners of the internet to build these core models it
745. 1:54:29like knows all that [ __ ] and then you can just ask it and it will tell you it's like what can you do with that I don't know everything
746. 1:54:40does IT program yes will it replace my job yes it won't replace your job but someone using it will that's from Paul rateser
747. 1:54:52open AI just created an IOS app for chat EBT yes but be very careful to get the real one so go to openai and get the
748. 1:54:59link from there and just make sure it's the official one all the other ones are rip-off [ __ ] can an AI replace programmers answer
749. 1:55:07that have you played with perplexity AI I have played with it a little bit I can't quite figure it out it has some
750. 1:55:14chat gbt [ __ ] in it but then sometimes when I ask good stuff I just I keep going back to this one these [ __ ] invented it this is by far
751. 1:55:25the best like gpt4 is by far you know Head and Shoulders above everything else right now why I don't
752. 1:55:33know it just is When someone knocks that [ __ ] off the mountain I will change that Top Line but I just keep going back to
753. 1:55:40that because there's just nothing like it what's the other site you talked about earlier no it had playground at the end okay so that's um
754. 1:55:53it's also openai.com but if you go to platform.openai.com playground that's the developer portal for chatgpt's API
755. 1:56:03so you can change models you can there's all the documentation in there platform dot openai.com playground can you get GPT
756. 1:56:18version four from open for free no but you get it here from free bing.com if you go there it's a search engine and if
757. 1:56:27you click on the chat button that's gpt4 gpt4 for free FYI there must be some configuration difference in Bing chat or a different
758. 1:56:42difference yes Bing chat even though it's gpt4 is very different than chat gpt's gpt4 and it's not as good there's a lot of reasons for it
759. 1:56:57it is gpt4 so it's better than gbd 3.5 but it's not as good as the gpt4 you pay for here and I don't think there's a
760. 1:57:08free way to get it up there why do you pay for chat gpt4 if it's free with Bing because it's better and and and you also get access to the
761. 1:57:18plugins with gpt4 and when code interpreter comes out I I also get access to that so I pay for this for all the Early Access to their [ __ ]
762. 1:57:28code interpreter is going to melt some faces man go to YouTube go to YouTube search for code interpreter Chachi BT
763. 1:57:36and if you've done anything ever in your life with data or spreadsheets prepare to have your face just drip off your skull
764. 1:57:46I use AI with my kids crazy questions and it breaks it down for them until they get it yep exactly by the way if
765. 1:57:54you if if you're doing this with your kids um Khan Academy um which is the you know Alternative Learning thing the flip the classroom thing
766. 1:58:03um they launched a thing month six weeks ago called conmigo which is a GPT for Education bot sitting on top of Khan
767. 1:58:12Academy's all of their stuff and it it's remarkable they spent like seven months fine-tuning that bot to be a really good tutor for kids
768. 1:58:22and and adults quite honestly because I've used both and I get better coding responses from Bing chat oh that's cool
769. 1:58:30oh yeah if you get so with coding see if you're doing coding I would say GitHub um GitHub co-pilot I would play with
770. 1:58:41Bard I would play with Bing and compare that to gpt4 right because with coding it's very possible that one
771. 1:58:49of those other systems is better than gpt4 with with word stuff with language stuff gpt4 nothing touches it but with
772. 1:58:56coding there there might be other things that are better so that's that's a really good point Winston thank you for that
773. 1:59:06all right I got 10 minutes and then I'm gonna get yelled at because because I live in a household with people that love me and want to see
774. 1:59:14me and don't don't want me on The Tick Tock all the time do you love those Tick Tock people more it's possible
775. 1:59:28all right this is my new favorite live thank you love that thank you for sharing all this I appreciate that
776. 1:59:35what will happen to people who don't know anything about computers okay good point if you don't know anything
777. 1:59:42about computer listen you're here so you know enough about computers to open your phone and get on a ticky tacky so you
778. 1:59:49know more about computers than you may be willing to admit don't you think don't you think you do so here's the deal
779. 1:59:59the inflection point that we have now passed we we have we have passed our Rubicon or we have passed we're into a new phase
780. 2:00:08prior to this phase if you wanted to use machine learning and like you ever heard Google talk about things like tensorflow and you're
781. 2:00:18like oh I'm gonna go check out this AI stuff that sounds really cool and then you go figure out what tensorflow is and
782. 2:00:23unless you have a computer science degree you ain't doing [ __ ] with tensorflow um and that's how it was for many many years for decades
783. 2:00:33Chachi BT November 30th 2022 so this isn't that long this is like seven months six months ago marks the beginning of a new era where
784. 2:00:46people who are not computer scientists and researchers and big [ __ ] programming engineering brainiacs can use these tools
785. 2:00:56on your phone if you're on you're on your phone using tikitaki Point your phone browser either Chrome or Safari just point it to that web
786. 2:01:06address and start talking to it like it's a person asking it things like it's a person like it's a really smart person
787. 2:01:14and ask it questions for things that are personal to you so maybe you're into gardening ask it to come up with um
788. 2:01:24what are the next five plants that I should plant or hey I've got you know this kind of infestation and what can I
789. 2:01:32do about it that doesn't use chemicals like just start interacting with it um like you've got a like like you have a
790. 2:01:41friend who just happens to be the smartest friend you've ever met you've ever known like don't worry about the fact that you
791. 2:01:50don't know computers take what you're passionate about what you know and start solving problems using that the free version of that
792. 2:02:02that skill over the next two years is going to be incredibly valuable people that don't do that are going to be at a massive disadvantage so
793. 2:02:13just go play just go play that's why that's why I do this every night is I come here to inspire people like you to just fire up
794. 2:02:23your browser point it at that and start asking it to write you a poem have it write you a song for your significant
795. 2:02:29other or a poem about your pet and then just sit there with your mouth open and go it can do that yep
796. 2:02:37and then have it write you a dinner plan for the week and then say I've only got three ingredients in my fridge what can I make
797. 2:02:48from this out of the cub what other things do I need to buy here's your list a good site for finding good prompts
798. 2:02:58prompts.chat prompts dot chat chat dot prompts and there's probably other ones but that's that's just that's a document
799. 2:03:05that I know is valuable and not a waste of your time there are a lot of prompting sites that are a waste of your time
800. 2:03:13um the people who don't know can still learn quickly with I boys I don't know what I boys oh AI Bots you know the eye boys
801. 2:03:24you can learn quick with the I boys where is that app found which app I don't know which app what's the down payment not sure what
802. 2:03:33that's about a good site for finding good prompts we just showed that have you fed it any Declassified FBI Files
803. 2:03:40like DB Cooper I haven't but I better be good at that [ __ ] all right I got six minutes then I gotta
804. 2:03:46go let's let's go let's go play with uh with chat GPT connected to the internet and see what we can find on DB Cooper it
805. 2:03:55may some of its safety stuff it might call it a conspiracy theory and not let us play but okay so I'm at gpt4 I'm going with
806. 2:04:02browse with Bing so I'm like okay um what is the latest theory on who DB Cooper is and where the [ __ ] money is
807. 2:04:17it probably doesn't like the fact I swear sweared swore Jesus squared I don't know it was on there and he
808. 2:04:27sweared with these bad words he was just he talked he talked fast and he had a lot of bad words and he he seemed like a
809. 2:04:34bit of a dum-dum but then he seemed like he was kind of smart but he was more like a dum-dum he sweared a lot he sweared a lot
810. 2:04:43and I think he said he was drinking iced coffee but I'm pretty sure it was a Long Island Iced Tea so what this is doing right now it's off
811. 2:04:53searching the internet I mean this this is one of the cool things like when you like flip these little things open
812. 2:05:00it went to whatever that site is then I went to the guardian then I went to some other site of the Guardian click failed
813. 2:05:09click failed so it's it's trying it's trying and failing searched Bing DB Cooper Ransom money Discovery scientists uncover new
814. 2:05:20minuscule Clues on DB Cooper so it's searching it's searching how exciting is this I don't know you see he seemed like a
815. 2:05:34dum-dum as of recent developments Eric Euless a true crime investigator who organizes an annual event called
816. 2:05:42coopercon has identified a new person of interest Vince Peterson was an employee at Crucible Steel all right where's the money regarding
817. 2:05:50the money in 1980 a young boy found six thousand dollars of the ransom money while camping with his family on tenobar
818. 2:05:56a sandbar in the Columbia River the only Ransom money ever recovered from the DB Cooper case recently amateur
819. 2:06:03scientist Tom K had examined this money and identified diatoms tiny deposits of algae on it he believes
820. 2:06:11that the diet comes indicate that the money ended up in the river months after Cooper's jump contradicting the FBI's original Theory
821. 2:06:22so anyway yeah it does that it does it quite well now was all that true I don't know but you can like it includes sources
822. 2:06:34doesn't it include sources yeah it includes sources so I can actually go look at the sources where it pulled those answers from
823. 2:06:41so pretty dang slick ain't it I'll tell you what this here AI stuff it's pretty magical where is DB Cooper
824. 2:06:56uh what was that Twitter account where you meet where you talk about AI okay hold please so this is the AI salon so that's a
825. 2:07:09community that I started we meet every other week our next meeting is tomorrow night and we have a Discord server you
826. 2:07:15can play on and then this if you go to Twitter and go to Everyday AI news that's my newsletter and we also do
827. 2:07:23Twitter threads on interesting stories there so you can go learn [ __ ] there and you can sign up for a newsletter there it's all free
828. 2:07:30free at some point we're gonna monetize it but right now just it's out of the good of our hearts not really
829. 2:07:40um we'll monetize it at some point when it makes sense but right now it's free like I'm just like the reason this Tick
830. 2:07:45Tock Channel exists I'm just trying to process what I'm learning and get it out of my head and then Inspire other people
831. 2:07:51to do the same I think it's really important if you learn this stuff if you get curious about it like I'm
832. 2:07:57encouraging you to the other thing that I would encourage you to do is share what you're learning get other people up to speed on this
833. 2:08:05because this is important this shit's coming fast and a lot of people are going to be blindsided by it and it's going to be like painful
834. 2:08:14and and I want there to be less pain and more um excitement because these are the most powerful tools ever put in the hands of
835. 2:08:24mankind and we all get to play with them like we all get to play with them so it's pretty magical times but it's also
836. 2:08:32serious because it's it's coming fast and it's going to disrupt a lot of jobs so as you learn about this stuff get
837. 2:08:37generous with what you're learning share what you learn do things for others use chat epd to solve problems for other people
838. 2:08:46um and then when they and then when they say to you how did you do that let me show you you know it's it is one of the best things in the
839. 2:08:58world when you get this moment from someone when they see what this [ __ ] can do so with that listen I know I missed a
840. 2:09:05bunch of questions below I appreciate y'all hanging out with me I gotta go I got work tomorrow I got I got I got a
841. 2:09:10company to get to I gotta go CEO you gotta go CEO tomorrow can't just be hanging out talking on the tick tock all
842. 2:09:17right everybody peace out have a good night I really appreciate your time