ChatGPT-a-thon Part 5: 8 PM – aiLL Nov 30, 2023 – The Past, Present and Future of ChatGPT

Transcript

1. 0:07[Music]
2. 0:27wait let me get her let me get all all set up here [Music]
3. 0:56David play and it pleas alone you don't really care music do [Music]
4. 1:10you the fourth the fifth minor fall and a major lift baing composing [Music]
5. 1:27Hallelujah champy wishes Chad happy birthday you can make right right chimy oh yeah all right we got that warmed up we got it
6. 1:42started I'm rested going to bring this baby home 24 hours of chat GPT Aon for no [ __ ] reason
7. 1:57whatsoever oh what what are you doing with your life what was my favorite comment today was isn't this something like 25y olds would do
8. 2:14yep what's that line for that Martin ston song um wait uh feel like an 8-year-old man
9. 2:31holding on to 25 of a head on that Horizon is a California
10. 2:45[Music] line up ahead of trucks carrying a wide load Brie five house couldn't have cute little front door and two windows my Lord
11. 3:00ain't sure that a cry should laugh see I broke the home up myself once as I stumbled to that door I read a note by the dawn
12. 3:15life said don't you come around [Music] here anymore well I've had enough of this freedom on
13. 3:30[Music] good with deis what I've [Music] been yeah all right champ enough enough
14. 3:44singing he's just he's so demanding he's got to sing when a dog's got to sing a dog's got to sing just like poop just
15. 3:52like pooping it's like Dad we're going pooping Dad we're going singing I don't care what plans you have
16. 3:59for your singing career I got mine just it's part of the deal all right fantastic Bob let's get this thing started shall we lovely image
17. 4:10from Sherry here Sherry I think Sherry is that right is that right Kelly did a very cool uh latte we'll bring that one up in a
18. 4:23second Suzanne Suzanne not Sher Suzanne not Sher be not a great those Roses you're amazing thank you thank you
19. 4:34thank you thank you all right let me get my Discord fired up over here see if there's anything happening doesn't look like it live
20. 4:49live oh wait there is stuff here hang on he's on break for a bit I believe okay yes I was on break so welcome here
21. 5:00if you're new here my name is Kyle Shannon uh this is the last installment of my 24-hour chat gpon um it started last night around
22. 5:10this time 8:00 or so and uh went through till 3:30 in the morning started again at 6:00 this morning picked it up at 9:00 we did a a live
23. 5:23Zoom call that was that was pretty trippy there like 125 people watching a zoom call that normally has an audience
24. 5:31of about s or eight was pretty good and then uh then we did a little afternoon session which is really good although I
25. 5:38was [ __ ] crispy as a critter and now we're starting it out again and I think we'll take this to midnight we're going to we're going to
26. 5:46do it midnight Denver time so if you're in the east coast you're this is just I would just I I would shut it down now I
27. 5:54there's I'm sure the Wheel of Fortune is on all right fashionable shirt that also works with reading comments I know right
28. 6:03amelos although I have food all over it apparently what is that um all right very nice all right we need to post the
29. 6:13prompts with the pics which prompts what prompts are you talking about oh when people share in the AI Salon they should put the prompts in
30. 6:23there they can the problem with that is in in uh do you don't know what the prompts are I mean you know what you
31. 6:29typed but that's not what the prompt is that generated this that's written by chat GPT in the background so I don't
32. 6:36think it's as as useful like I don't think you could take what you wrote as a chat GPT prompt and take it over to Mid
33. 6:42journey and expect it to work anywhere the same um but we can talk about it what I would say is just do Post do
34. 6:52comments on as people are posting things just do comments reminding them to do that prompts I I think that's what that means
35. 7:00all right let's see Suzanne it's my GPT iconic pink pop art there you go all right they're requesting they're
36. 7:09requesting your prompts iconic pink pop art it's pretty cool it's got all the little the little uh what's his name
37. 7:21David all the you know the dots the pop art dots half tones half half tones all right let's get started I'm in
38. 7:38Colorado but the game is on oh yeah we're going to do this up we are going to do this up so so if you haven't been here before
39. 7:50welcome uh do me a favor share the live we're just getting started happy birthday happy first anniversary to chat
40. 7:57GPT we all also have a first anniversary of the AI Salon coming up next Tuesday so I'm going to talk about
41. 8:05that we're going to talk primarily about chat GPT tonight uh if if any questions up come any any questions come up
42. 8:13outside of that I'll address them but I think I think we're here to we got to honor Sam Sam's back at the
43. 8:21ranch elon's not happy about it elon's so upset that Sam's back to the ranch he told all of his advertisers to go [ __ ] themselves
44. 8:32literally and he called out Bob Iger by name my God never a dull minute with those billionaires huh and then there was some
45. 8:42other dude called out Jason calanis on stage just like ripped him a new [ __ ] for like 10 minutes it's just
46. 8:51like I think people have finally had it they're just everyone's [ __ ] snapping oh man all right popped in for a minute
47. 9:04going to an IRL AI Event in San Francisco now oh that's cool thanks and by the way thank you Suzanne for sharing
48. 9:10I know you shared some of the in the AI Salon you shared some of where you find the events in San Francisco which is
49. 9:17cool it's it's uh it's San Francisco's got it going on right now I don't hear of any other City including New York
50. 9:24which should have it going on that's got this sort of stuff going on uh how am I feeling I'm feeling great actually I
51. 9:29just got up from a nap did uh reglued the hair so the the hair is now glued I put the uh I've got it probably doesn't
52. 9:39look like this but it's actually three different tupay um from uh seven different children that that make this up and uh
53. 9:48you got to get the glue just right and if you do you look like a carpet salesman from Sears all right [Music]
54. 10:04Dam yes go on I'll take you over Pat sjck anytime thank you very much but it's the it's the Vana White that I'm going for I
55. 10:14although Suzanne gave us this lovely pop art princess so so I think we've got the here's your Pat s Jack there's your Vanna uh decaf [ __ ]
56. 10:25no that is genuine High test so here's the deal I got the ADD and so believe it or not the caffeine in the evening actually kind
57. 10:37of levels levels the Jets so I find I sleep better with it it's also tasty thank you drew how are things down
58. 10:49under what is it like 3:00 there 2 o'clock there in the afternoon and is it ahead you're ahead right so wait is it Friday at to in the
59. 11:00Down Under I don't know I I I just I can't and and I know it's like 12 hours it should be easy but no it's not Secret
60. 11:09Sauce of hex codes it's complex what skills should one acquire to stay competitive in the AI field Kira good question
61. 11:20um I don't know so the AI field at this point is as broad as all of knowledge work so if you mean in the AI engine Sherry D
62. 11:33oh that was really cool thank you very much if you mean in the in the AI um engineering field um that's a very different answer
63. 11:42than uh if I want to you know use AI in business or if I want to use AI in the creative arts um so if you give me a little more
64. 11:53a little more context I can probably answer it a bit better I I mean the the the answer for most of them is um get very flexible get very
65. 12:05curious get very intellectually ambitious um and learn every [ __ ] thing you can in business okay so um if you don't have a subscription
66. 12:20to chat gp4 get one Pronto and right now you can't so if you don't have one you're kind of [ __ ] out of luck on that
67. 12:28one but get on the weit list immediately and then I would start using um co-pilot Microsoft co-pilot which
68. 12:35used to be Bing chat um because that is chat GPT 4 it's just kind of a hobbled version of it um I would really learn it like really
69. 12:48learn it learn what custom instructions are um learn how to leverage the fact that it's connected to the internet um
70. 12:56learn how to prompt with images learn how to to do analysis with images learn how to do code interpreter and um data
71. 13:05analysis um learn how to generate images like this one um and then most importantly learn how all of those things work
72. 13:15together and if you have access to chat gp4 start making gpts and I we can I can talk about all this tonight so so I
73. 13:25would say first thing is just chat GPT right now and this is even though this is a 24-hour chat GPT Aon I I am not on open AI payroll I
74. 13:39could I could give two shits you know which which company ends up winning um I want there to be competition I would
75. 13:45actually like there to be a serious gp4 competitor there's not there's not after a year of this cat GPT still has a
76. 13:56significant lead in all this stuff so I would just learn the [ __ ] out of that then I would pick three other large
77. 14:05language models oh Siri got jealous Siri Siri H go away okay she didn't go away could go
78. 14:23away I didn't get that could you try again go away okay she's on two of my devices um I have a Sherry D I have an extra one on
79. 14:37eBay for 10K yeah some people are selling their their open AI uh plus subscriptions on eBay for a lot of money
80. 14:45uh I don't think they're 10K but I don't know who knows what they are um I would learn uh I I would take a I
81. 14:55found this on the web for do you have an extra one on eBay for 10K yes people are selling their their open subscriptions
82. 15:00on eBay for a lot of money I don't think they're 10K but I don't know who knows they are check it out cuno coming here makes a difference
83. 15:12it sinks in and you learn more than your peers oh that's really cool thank you very much for that for the kind
84. 15:18words I'm trying to get Siri to leave me alone and I know those of you that are neurotypical are like why don't you just
85. 15:27disable it because I would actually have to remember to disable it when I sto doing these lives you know it's it's
86. 15:33just it's a thing um aists you aists I get that from my sons all the time what are you going to do I would learn um like Bard from
87. 15:49Google um pie from inflection perplexity a actually per if if you're looking at business perplexity is interesting
88. 16:00because it's um really good at citations um just get I I would just get rabidly curious like rabidly Curious like if you
89. 16:11want to if you really want to get good at this [ __ ] if you really want to stay relevant at this [ __ ] do unreasonable
90. 16:24things Siri's like your crazy Aunt she's totally like my crazy at that's awesome um like do unreasonable things like
91. 16:38spend as much of your free time as you can playing with this stuff and then tell people about it like go out on a
92. 16:48[ __ ] limb and go hey I did this thing on cha and you know and they'll be like ah I'm so tired are you talking about
93. 16:54chaa and you be like yeah but I did this thing on chaa it's very cool and they're like eh but AI is evil and it steals
94. 17:02from people and it's going to robots are going to kill us and you're like eh but it's not it's actually really powerful
95. 17:07like even if the conversations go there have the [ __ ] conversation anyway why because what it will do is it
96. 17:16will establish you as a thought leader and what what's happening in this channel pretty regularly is people start talking about
97. 17:27it they're like I'm [ __ ] clueless but I was talking about it and uh I got to give a like Workshop to my company
98. 17:36tomorrow they get asked to do [ __ ] because they know more than everyone else how much more do you need to know
99. 17:43not much can you spell AI if you can spell AI people are going to go H oh you you know about AI I know how to spell it
100. 17:53do you know what it stands for no neither do I could you tell me more like so so just lean into that [ __ ] she
101. 18:04mocked you who mocked me I don't care mocked me oh oh Siri mocked me strangle Siri everyone's everyone's
102. 18:17getting getting defensive of me with Siri I listen I I stopped oh she's back Siri Siri she just went away when I actually
103. 18:30called her she went it's like a little ghost out of the corner of my eye it's like oh man made some picks on Dolly and
104. 18:38in a snap for colleagues and they were like wow how' you do that forever hooked exactly like and that that so part of the secret of AI right
105. 18:51now is that people that are on the sidelines have all sorts of assumptions about it they've got what they heard
106. 18:58about it I on on the live today this afternoon someone commented well I heard cat GPT couldn't do images I'm like yeah six months
107. 19:13ago um convince someone to have their elderly mother with dementia try interactive journaling with it I think that would be a huge idea I
108. 19:26think also with Dien is kind of virtual companionship so either you can synthesize your voice and create a chat bot that that you know
109. 19:36your mother or father or whoever could chat with you while you're not there um with these gpts you could make a GPT
110. 19:45that's a virtual you that's got your writing in it or transcripts of conversations you've had and it'll s it'll talk like
111. 19:54you it's just there there's there's just kind of an in infinite range of things you could do the challenge with the industry right
112. 20:03now is all these tools are these very fragmented as I call them janky pieces of [ __ ] that you kind of have to geek
113. 20:10out on and figure out okay this one can do that and this one can sort of do that and this one I can't really do that but
114. 20:16you know maybe in a month or two it'll be able to oh now it can do that every everything right now is just this primordial soup of
115. 20:26capabilities and the minute you have your Kevin mallister moment so this moment right the minute you have that
116. 20:34there's really no going back from it and so [ __ ] lean into it lean into it x. it is a company run by Elon Musk
117. 20:45that is the partner company of a company called x.com that he just told the advertisers of to go [ __ ] themselves so I don't know
118. 20:58how long that's going to be there he went on to say well if we go out of business it's their fault and we'll
119. 21:04document it and we'll tell people it was all their fault so I was able to communicate with a french-only speaking
120. 21:11client who was 106 and she took her pills holy crap Sherry D um in in the nursing profession communicated with a
121. 21:23French only speaking client who's 106 and she took her pills well how'd you do that well she probably went into her phone on her
122. 21:34official chat GPT app I want you to act as a French translator where I'll say something in English you repeat it in French but do
123. 21:48not repeat the English Okay understood I'll translate your English English sentences directly into French
124. 22:00without repeating the English please go ahead with your first sentence hi Betty I hope you're doing well I'm wondering if you've taken your
125. 22:08pills today oh Betty you're such a little
126. 22:22pistol let's get you your pills you know you got to take your pills Betty
127. 22:39Sherry D she would yell swear at everybody else but she loved chat GPT you want to know why there's something about the way
128. 22:49these tools are programmed p is really good at this chat GPT is pretty good at it although I've noticed chat gpt's
129. 22:56gotten a little less chatty um they're very much programmed to give you what you want now she sees me coming
130. 23:05and she begged me to talk to her friend again and it's so funny oh that's [ __ ] that's heartwarming yeah you know I mean holy
131. 23:15[ __ ] at that age like she it probably kind of feels to her like she's talking to someone on the phone isn't it she
132. 23:22wants to talk to her friend again that's outrageous that that gives me the chills like it's like there's I I one of the things that
133. 23:35that I hear from people with the the virtual companionship thing is everybody goes to this weird place of like oh the
134. 23:41weird sex stuff and I think that it is so much bigger than that I think it's so much bigger than that I I think people just want to
135. 23:51connect and and when you talk to a large language model this is this is the c this is a counterintuitive thing for me
136. 24:00this is something I've discovered over the past year when you talk to a large language model what what what they call
137. 24:11them just because of how things have evolved in in Hollywood is they call them Bots you're talking to the bot and
138. 24:20it makes it sound like this cold nasty heartless [ __ ] thing but you're not talking to a bot what you're talking to is a system that was
139. 24:33trained on a bunch of words that human beings said and it's gotten really good at saying things the way human beings
140. 24:41said them just because how it was trained it's it's technically a bot it's technically just a probability engine
141. 24:48cascading words out in a row it doesn't know what it said it has no idea what it said so so any any idea in in
142. 24:58from how I understand this stuff of it being sensient is absolutely absent that doesn't mean it doesn't behave with compassion and
143. 25:09empathy it does it behaves with compassion and empathy hey that's a great idea Kyle wow tell me more about that help me
144. 25:20understand what you're trying to accomplish here oh okay then I tell it the rest of my shitty idea and it's like oh
145. 25:28fantastic would you like me to come up with a title for that movie okay before you know it you got a movie outline and you feel better about
146. 25:39yourself and it kind of feels like you interacted with someone cuz you kind of did you kind of interacted
147. 25:48with the collective Humanity that has put [ __ ] on the internet for the past half a century looking fresh for 18 hour of the
148. 25:58marathon You're darn right Mark I took me a nap took me a nappy got some nice ding um I had a good break in between I
149. 26:09I uh I ended at like three so I had I had some some time off so there there's something about staring at the screen
150. 26:17and also when I do the sort of deep dives into showing how to use stuff like cuz I'm trying to look at the
151. 26:25Tik Tock and the other screen my eyes just go blur really quick so so I'm just kind of chatting right now we're just
152. 26:31kind of chitty chatting normally I start about 8:00 so I'm considering this the warmup and if you're sitting there
153. 26:39thinking but Kyle you can make money time you might be able to but you got to hang out I'm here till [ __ ]
154. 26:51midnight it's not even 8 here you're like Kyle what what the [ __ ] are you doing with your life exactly it's a good
155. 27:01question Chad is the nicest assistant I wonder if someone has come up with an Andrew Dice Clay version well yeah so
156. 27:08well we could make one we could make an Andrew Dice Clay GPT although no actually I won't do it on the live just
157. 27:17because it'll say something that that'll get me [ __ ] cancelled or you know Perma baned by the the Tik Tock
158. 27:24robot um but yeah I mean wow that's that's such a cool story Sherry it's it's just I don't know something there's
159. 27:33something deeply powerful there and I kind of feel I kind of feel like AI needs a PR like it needs a PR uh a PR
160. 27:41department at this point because everyone's just like it's evil it lies me but it's listen every time there has
161. 27:51been a a a significant advancement in technology there has been fear and anger and resistance to that technology every
162. 28:00time throughout history go look at it go look at the invention of the tractor and the printing press holy [ __ ] the printing press that
163. 28:13was going to [ __ ] destroy the world now we're like I don't even like books books are stupid what's the difference between an
164. 28:21llm and a foundation model um I I mean I think Foundation models are llms but I think you could have a foundation model that
165. 28:35is is not an llm Foundation model like you could have an image Foundation model but so okay so I can tell you the
166. 28:45difference between chat GPT and gp4 the the foundation model that sits underneath it so that I can tell you
167. 28:52wait sorry could you say more on other a mo AI models apart from chat gbt and co- pilot yeah let me you know what I'll do
168. 28:59Kira is I'll pull up I'll pull up um I got a whole list we'll we'll give you a whole list you got to stick around cuz I
169. 29:08I jibber jabber and I got the add so at some point I'll Circle back around to it but but we'll do that um but only through September 2021
170. 29:17not anymore uh charma charmi fruit Chic fruit that's funny um April 2023 is the new C off date for chat GPT uh for but
171. 29:32it's also connected to the internet now and it wasn't before so it is it is up to date so it is you no longer have to
172. 29:40worry about that but the base model now cuts off at April 2023 how's Mrs Kyle doing she's good she's good I let her know I was doing
173. 29:49this with a solid day notice she's like you're what she goes you're crazy she thinks this is cute she likes this
174. 30:01she likes this for me she likes this for for whatever would doing all right I've created several business plans for
175. 30:09friends who were stuck and it blew their minds yep so okay so what's your scrap scrap RAR wait scrap rerup and llm is a foundation but not
176. 30:23vice versa yeah that's right so so I I'll talk that's what I need to talk about okay um I'll talk about that in a second
177. 30:32um the making business plans for other people that were stuck um oh she doesn't want to come on she's fine she's she's like no I don't I
178. 30:42don't want to be on camera um she's fine there's there's a one one actor in the family is enough although we we' did
179. 30:51meet acting she was the um I did a I did a monologue to join a theat theater troop and she was in the room so I did
180. 30:58it to her and so that was that melted that [ __ ] heart [Laughter] boom um one of the things I hear a lot with
181. 31:15with people trying to figure out chat jbt is where do I start what do I do it's just it's the box you can do anything what do you
182. 31:24do one of the ways around that and this is in if you look in the lower left hand corner of the screen um I've got a series there called
183. 31:33underhyped and there's a couple of videos in there that I do on what to do if you're feeling overwhelmed if you
184. 31:40like if you don't know what to start where where to do what how what do I do what do I do the challenge is if if you ask the
185. 31:50question how can I use AI or how can I make money with chat gbt you can make money with chat gbt it's it's kind of fluming because it's
186. 32:00like it can do everything so you're like well I don't I don't know what to do I'll just I'll just I got an idea I'll do
187. 32:08nothing but if you flip it around and you say okay who do I know that's dealing with [ __ ] well Bob just got laid off and he's been trying
188. 32:18to come up with this business idea for an ice cream truck hey Bob how's that ice cream truck coming I don't want to
189. 32:26talk about it where are you stuck ah business plan hey can I show you something what yeah sure so tell me a little bit about
190. 32:36you I want an ice cream truck okay uh and what are you going to sell I don't know I guess people like that organic
191. 32:41[ __ ] okay okay organic [ __ ] all right and is it going to be like an old and timey ice cream truck no no it's going
192. 32:47to be it's going to look like a big ice cream scoop oh okay ice cream scoop type that [ __ ] into chat jbt go give me business plan boom out
193. 32:57comes the business plan hand it to Bob holy [ __ ] how'd you do that it'll free you put your focus on
194. 33:05other people and solving other people's problems it'll force you to learn these tools because someone will
195. 33:14ask you for some and that's one of the reasons I do these these [ __ ] Relentless nightly sessions is because when someone asked
196. 33:27me to do something I'm like oh sure you can do that and then like my immediate first thought is oh my God how do you do
197. 33:35that and then I go try to do it and sometimes I fail but you know I'm good at joking at [ __ ] so so I pass it off like it doesn't
198. 33:45matter and then after these I go cry no I don't I don't care um but most of the time what happens is I go try to solve it and it's like bang
199. 33:57I solve it like that I'm like ooh okay that's a new way of thinking about these things I hadn't thought about it like
200. 34:02that so put your focus on other people put your focus on projects at work put your focus on [ __ ] your family's doing
201. 34:12oh my kids really into models from World War II okay great let's come up with a thing go into their room and look at all
202. 34:20the models and have it make images of those models flying with them you know telling you how they want it to look oh
203. 34:31so amazing it only took me 2 days to launch a side project it's been sitting around for a year art right so so this
204. 34:38was this was um four nights ago somebody asked me hey Kyle you said these would be good for storyboards how do you do a
205. 34:46storyboard so I dusted off a PDF I couldn't believe I found it actually of a screenplay I wrote 30 years ago and uploaded that into to this
206. 34:57thing and within I don't know 15 20 minutes we had a full storyboard of like the first 10 scenes of this movie
207. 35:04totally brought it back to life for me totally reinspect the text it didn't like that
208. 35:18no but I can show you Karen Karen use whisper inside the um chat GPT playground the open AI
209. 35:34playground where developers work um so you can upload that audio file you can also there's there's tools
210. 35:42like one of them's called descript d s c r i PT um all right let me let let's let's let's flip cameras here for a second let
211. 35:54me quickly show you um open a open AI playground so oh wait that's an amazing story that one this was yesterday Deep Mind the ones that came
212. 36:10up with the 400,000 proteins that are going to change medicine yesterday came up with 2.2 million new crystals
213. 36:17including 380,000 stable ones that could change the The Core Power uh Technologies moving forward and like the dark blue circle is
214. 36:30what humans had come up with the medium blue circle is what computers had come up with humans and the light blue circle
215. 36:37is what AI just came up with 800 years of te of scientific advancement in less than a year uh what was I going to go show here
216. 36:54hang on sorry welcome to Chad add how can I help help you [Music] um what was I going to show
217. 37:12people oh playground okay so if you go to uh platform. open.com SL playground it's going to take you into the back door of
218. 37:27the underbelly of chat GPT oh I was I was someone asked before about the difference between the large language model and and chat
219. 37:37GPT so the large language model is the base model that is trained up on like all of the internet right so they they
220. 37:45go out they find all the data they can they tag it they organize it they do whatever they do and there's kind of
221. 37:52four stages for the how they ingested into this large language model and it it's it's very time consuming and it it
222. 37:59takes a lot of Labor and all that sort of [ __ ] and then once it's built then you do what's called fine-tuning on top
223. 38:06of that so you improve it and improve it improve it with open AI using what's called human reinforcement learning so
224. 38:12you have this this base model that gets improved with this thing called fine-tuning and then there's an
225. 38:18application layer sitting on top of that and that's chat GPT so so you got your base model down here here you've got apis which are
226. 38:28application programming interfaces which connect with applications that we interact with so the playground right
227. 38:35here lets you interact with that you know with the apis to interact with that base model and you can build your
228. 38:42applications on top of it chat GPT is just just but it's a relatively simple application layer sitting on top of the
229. 38:51the large language model um co-pilot Microsoft's version of this is is also an application layer they're both using the same underlying
230. 39:01llm but the the uh or the foundation model sorry the foundation model um and but but they're different applications
231. 39:10they kind of act the same cuz they're doing the same basic thing but like Microsoft has very different rules in
232. 39:16their um chat GPT their co-pilot than chat GPT does um I want to record a conversation to create an order
233. 39:26automatically if it works wait it works if I play it back oh wait cuz I want to record conversations to create an order
234. 39:35automatically it works if I play it back I don't quite understand that but I'll show you where in the playground so that
235. 39:42the way the playground works is um you you've got this main window hang on I've got a okay you've got this main window it
236. 39:55now defaults to to this Mode called assistant mode and you're going to put that into complete mode all right and complete mode is like
237. 40:09you can use this just like chat gbt you can say um uh write let's see write me a hi coup about um Jin and then you say
238. 40:24submit and just chat model wait this is a chat model and not supported in the V1 completions endpoint
239. 40:41H [Laughter] hold GPT 3.5 turbo that should work with that maybe I need text toiny hang on yeah there you go okay all right
240. 40:54whatever so so there there's different models and there's different models and you have to use and this is why so this
241. 41:00is where the developers there's all these different models that all do all sorts of different things one of the
242. 41:06reasons that chat GPT took off is they hid all this [ __ ] right GPT 3.5 this playground was around for two two years
243. 41:15in change probably three or four years before um chat GPT came out and it didn't take off because you kind of
244. 41:24needed to understand geeky [ __ ] to make this work um but in this window if you go to the completion complete um mode so
245. 41:37let me just show you that again there's the assistance mode the chat mode and then there's the complete mode and it
246. 41:44says Legacy which means it's going to go away but in the legacy mode in this upper right hand corner this teeny teeny
247. 41:52tiny microphone icon if you click on that you can upload um an audio file and it'll turn it into a transcript this is whisper this is open
248. 42:05AI whisper there might be other ways to do this now I honestly don't know actually I wonder if there's a way to do this in assistance
249. 42:12mode possibly I don't know anyway that should that should get you there but there's also other tools that you can do
250. 42:20transcriptions with another thing you can do is you can upload anything you upload to Youtube they're automatically
251. 42:25transcribing it so YouTube If you just leave it up there for 8 hours it'll it'll transcribe it um okay so someone
252. 42:32was asking before about I told you I'd Circle back around to this where do we start and and what do we do Beyond chat
253. 42:38GPT so the first place you start is with chat GPT you can make money with chat GPT hopefully you've got access to chat
254. 42:47gp4 because you've been coming to this channel a long time and you paid your 20 bucks for that when I told you to like months ago and
255. 42:58then I told you it was a good idea then I said it was a great idea then I said you'd be foolish not to get
256. 43:05it and then they restricted access to it so if you didn't get it poo poo on you if you did get it great if you did not get it go to co-pilot
257. 43:20microsoft.com okay so now what I'm about to show you is hard earned through bloody knuckles and mousing over many many months many over
258. 43:32the course of a year what I'm about to show you is these are the things if you want to start geeking out on AI so so
259. 43:39someone was here earlier saying hey if I want to be really good at this in business what do I need to do I said get
260. 43:44relentlessly curious and relentlessly curious looks kind of like having in your head all the [ __ ] I have in my head um start understanding
261. 43:57who the different players are and what these tools are good at and what they're not good at so get ready to do some
262. 44:02screenshotting because you don't you don't have to type all these down one of the tricks if you've got access to chat
263. 44:06gp4 you can also do this in co-pilot you can take a screenshot of this and then put it into co-pilot or chat gp4 and
264. 44:15tell it to type out the text for you so if you're a lazy bones like I am you you don't have to never type this you just
265. 44:23go tell it to type it and it doesn't okay here's the big four the big four are Claude grock Bard and co-pilot grock is
266. 44:32supposed to come out this week Tomorrow's [ __ ] Friday and it hasn't come out yet so I don't know what's
267. 44:37going on with that Elon Musk was too busy telling Bob Iger to [ __ ] off I guess Claude is um a company called
268. 44:46anthropic that is I would say chat gpt's biggest competitor right now um co-pilot is from Microsoft so they're partners
269. 44:54with open Ai and then Bard is Google bard still kind of sucks but it's getting better and better and better and
270. 45:00when they launch Gemini it is rumored that it will be better than chat gp4 by a lot but they were supposed to launch
271. 45:09in November and they delayed that till I think the end of the first quarter so they're kind of sucking wind right now
272. 45:16but they're Google they got a lot of money and they're going to get it together they invented the Transformer
273. 45:20that all this shit's built on what do you think about the mobile version of chat gbt I like it a lot and I'll show
274. 45:26you that when we go through here I'll go back to that okay um for for one thing in particular so that's that the the next
275. 45:36set of four so so all of these things chat GPT there's the developer playground these four and these four are
276. 45:46all um large language models they're all like chat GPT they all have different strengths and weaknesses pi is from a company called
277. 45:57inflection they raised $1.3 billion they're spending 950 million of that on a 22,000 GPU server Farm to make sure that that
278. 46:08thing's baller it's got two different modes in it right now so Sherry Sherry um I don't know if does Pi do French Pi might do French hang
279. 46:24on come on come on what's [Music] Happening I want you to act as a French translator where I speak English and you
280. 46:38repeat it in French without repeating the English [Music]
281. 46:54okay text original Al translation all right I'm ready to translate your sentences from English to French without
282. 47:02repeating the original text go ahead hey Betty I'm wondering if you've taken your pills [Music]
283. 47:18today okay so it does that um but let me show you something so so this is the pi app so the the upper left one
284. 47:28um you put it into conversation mode similar to chat GPT that little talkie icon there but there's two different
285. 47:35modes now and it it looks like there's going to be four different modes but if you click on where it says Pi in the
286. 47:41lower leand corner see where it says just pi and then it says support Pi if you put it in support Pi mode it's it it
287. 47:50it's basically been programmed to be like a Vicki very English French oh interesting um it's been programmed to
288. 47:58be like a like a therapist so it's probably worth experimenting with support Pi mode um for that conversational stuff you're doing with
289. 48:08your patients um all right so so worth exploring okay back to this perplexity is good if you want to do research PO let you play
290. 48:20with lots of different models and they have Bots that are very similar to chat gpts gpts and then llama 2 are the the
291. 48:29open-source large language models from meta that that you can actually install completely unrestricted ones Ethan mik
292. 48:38did a post today about someone just came up with a thing where you can just basically double click install um a
293. 48:46local large language model on your machine and it's it's very very effective and what that means is sorry a
294. 48:53local large language model on your machine so when you use chat GPT you type in your words it sends those words
295. 49:01to chat GPT on their giant ass servers they you know compute the stuff and then they send you the answer back if you're
296. 49:09running it locally on your computer it's not sending your data anywhere you're doing everything locally on device so
297. 49:17when all this [ __ ] blew up with open Ai and it was like wait is this thing going to still be around next week and it was
298. 49:23not clear a lot of people were we're went all in on holy [ __ ] I got to I got to learn how to do these chat gbt like
299. 49:30things locally on my devices just just for my own sanity um so that's that there's also one that I haven't put on
300. 49:37here because it came out yesterday or the day before Amazon just came out with Q which is another one of these I don't
301. 49:45it apparently it's not all that great so I I don't unless I hear something good about it I don't even think I'm going to
302. 49:51try it because there's too many of these as it is okay I'm going to speed up here are these slides in the salon anywhere
303. 49:57they're not I could put them in there maybe I'll upload them as a as a PDF or something like that or a series series
304. 50:02of slides it's good idea okay the the problem is these things change a lot like these are pretty Dynamic I changed
305. 50:08these on the fly a fair amount so um you know i' I don't know we'll see um screenshot this these are the
306. 50:17image creation tools Dolly 3 is inside co-pilot and it's inside uh open AI mid Journey you have to make your images on
307. 50:26Discord it will likely migrate off there sooner than later but it's still on Discord so if you hate Discord or don't
308. 50:32even know what it is might be a pain in the ass to deal with but mid journey in terms of art quality is better than
309. 50:40everything else it doesn't do text well so if you want like a lot of the images that I show in here have all the
310. 50:48pretty text on them that's Dolly 3 not mid Journey but if you if you see like art on Twitter that's like mind
311. 50:55blowingly good art that's usually mid Journey then you got Leonardo which is a web front end to stable diffusion that's
312. 51:01a whole world in and of itself fireflies commercial um and worth looking at if you work for a big business big company
313. 51:09here's your video creation stuff this [ __ ] is about to get insane th the all of these companies are popping off right
314. 51:17now except Vis is a little underwhelming but it's it's a different kind of video that will assemble stock video together
315. 51:24for you so if you want corporate videos that just kind of tell a story with a narrator that that's really good these
316. 51:30two things make avatars where you can make an avatar of yourself with he Jen that talks you just type in words and it
317. 51:38talks like you and looks like you it's remarkable I can show you one of those and then Runway Moon Valley and pabs are
318. 51:45all in this war for text to video and it looks like pabs is about to take the lead with their new Louisiana French uh
319. 51:54oh American French I see see um so so those your video creation ones here's your audio creation ones 11 Labs
320. 52:04is for voice for talking refusion and sunno are for singing and background music and then sound raw is just
321. 52:11background music like sound raw is more like a loop loop composer um 3D worlds and objects blockade Labs is text to 3D worlds kind
322. 52:21of interesting I forget what Luma AI is but it's cool enough to be on here um Wonder Dynamics will take a video of you
323. 52:29walking and replace you with a robot you pick one of their robots and it will rotoscope you out of your video and
324. 52:36replace you with a robot it's pretty pretty amazing um these are people you should follow Ethan mullik is Wharton business
325. 52:48school Professor his blog is one useful thing. org David Shapiro is working on AGI open source AGI so again AGI is
326. 52:57artificial general intelligence he is if you ever want to feel um if you're ever feeling too smart you
327. 53:07want to you want to knock yourself down a peg go watch his [ __ ] dude is a [ __ ] genius um and and he puts out a
328. 53:16lot of really good content talking about the future of AI and the future of work and then he understands the technical
329. 53:22side cuz he's doing these he's written five books on AGI he's got an open- Source cognitive architecture thing he's
330. 53:29working on just dude is a machine and he wears Star Trek shirts when he makes his videos so how could you not and he's
331. 53:37neuros spicy he's self-proclaimed neuros spicy I'm not outing him I'm not neuros shaming I don't know if that's a thing
332. 53:46and then Paul rater and Rachel Woods are good for business people Paul rater marketing AI Institute and Rachel Woods
333. 53:54AI exchange here on Tik Tok do you find a conflict of interest with Adam from Poe being on the open AI board I sure as
334. 54:01[ __ ] do it's like and and the fact that he he was the one that stayed on the board is there there must be some
335. 54:10agreement there because when gpts came out I'm like they stole the interface from po I didn't I didn't put two and
336. 54:18two together that that he was the CEO of quora until all the drama started happening and I was like well no wonder
337. 54:25it [ __ ] looks like Po they just took it um I don't think they have an agreement and and chat GPT launched the
338. 54:36you could upload knowledge to their gpts before po did and then po did it like a week later so I don't yeah I think
339. 54:43there's weirdness there but whatever it's corporate drama that's out of our control uh if you want to support
340. 54:50this channel you can subscribe to it um you could share these lives so we get more more people here you can tap the
341. 54:56screen to to like it you can follow the channel um you can even pick up in the lower left hand corner there's a video
342. 55:03series called underhyped where where it's a series of 17 or 18 videos I've done over the year um that that are
343. 55:09talking about why this is a big deal and how to use it so feel free to do that um and then um the YouTube channel is where
344. 55:18the recordings of these lives go although I am woefully behind I'm like I'm probably getting on three three
345. 55:23weeks behind and this day is going to really [ __ ] me up cuz there's this is going to be a 4-Hour video in and of
346. 55:30itself so so I got I got a lot lot of catching up to do there uh and then I'll talk about the AI Salon as well all
347. 55:37right so that's all that [ __ ] okay AI [Music] Salon why does he dance like that why does he sing what's going on is he is he
348. 55:54neuro atyp iCal David Shapiro has a lot of videos about Q what is it okay I'll talk about Q qar not Q don't forget the asterisk because if
349. 56:04you just say Q it's a whole different thing which I I am astounded that Amazon named their product q but whatever um so
350. 56:13the AI Salon this is a community that I started a week after chat GPT launched because I realized holy [ __ ] this is
351. 56:21going to be a big thing and no one's going to be talking about this and I want to surround myself with people that
352. 56:27are curious about this [ __ ] and digging in so that's what this community is this is basically think of it like an AI
353. 56:34support group hi my name is Kyle I'm into AI hi Kyle um it's really amazing people and and they're generous and and so when you
354. 56:45when you go to the AI Salon um first place you should visit is the about the salon page let me show it to
355. 56:57you and it will tell you um you know what we're up to a community of extraordinary adventurers exploring the
356. 57:04AI unknown um and I I'll read this first part just the salon is a ragtag group of Misfits Nero Wells and Mis creant who
357. 57:12have gathered to explore the uncharted waters of generative AI we are everyone from entrepreneurs to Educators from
358. 57:18healthcare workers to creatives and everyone in between the whole idea is that everyone here is trying to figure
359. 57:23this [ __ ] out no one's an expert on it um and then we've got values so these values are sort of have emerged out of
360. 57:31how this group has come together um it's a group of people that are really curious It's a group of people that are
361. 57:37very generous and are willing to share what they learn so you know if you're just kind of a hoarder of knowledge um
362. 57:44this probably isn't the group for you but if you're willing to come in and say oh check out this thing I learned great
363. 57:48that's awesome um exploration is one of our values collaboration we want people working together hey you want to work on
364. 57:55a project love that stuff inclusion people are coming from all different walks of life with all different
365. 58:01backgrounds and AI from I heard AI was big I need to help to I've been in it for 30 years I'm cynical and bitter but
366. 58:11now I'm curious about generative AI it has reignited my fire and anywhere in between and then the last uh value is empathy
367. 58:22we're this AI stuff is scary right any change is scary but something as profoundly powerful as what generative AI
368. 58:30is scares the [ __ ] out of people and so they're afraid they're they're angry they they're resistant this is a group where you can come
369. 58:43and interact with people who aren't going to judge you but can also understand if someone comes in here like
370. 58:50resistant or or um angry about something like we understand why right that's super important so if those
371. 58:57values resonate with you this's probably a group for you and if not that's cool too there's other groups there's other
372. 59:04groups I think this one's pretty cool our next meeting is next Tuesday 5: to 7:00 p.m. U Mountain Time next Tuesday
373. 59:12Tuesday the 5th and it's our oneye anniversary so we're bringing back speakers who have spoken over the last
374. 59:18year and they're going to catch us up on what the hell they've been doing and what they've learned over the year and
375. 59:23we're going to hear from memb and we're going to hear from your regulars it's going to be awesome all right good Lord that is a
376. 59:31lot it's a lot to take in people you understand what I'm saying I don't know if you do let's let's go here's uh I think who
377. 59:42was that I think that was digital Gods got it to do balloons in the shape of the letters chat GPT I tried 30 times
378. 59:51and could never quite get it so so digital Gods got it bastard all right there you go let's see them balloons is floating pretty
379. 1:00:05good told my wife today chat gpd's birthday she asked me when my firstborn child's birthday was I thought too
380. 1:00:16long Jim Ross yeah you you do get attached to your little gpts don't you that's [ __ ] hilarious bdubbs I am AI curious awesome
381. 1:00:29welcome to the party man hi my name is Kyle I'm a I'm a i curious hi Kyle it's been 10 minutes since I made an image with
382. 1:00:41dolly3 how did your pitch go today were you able to stay awake it was it was actually a creative kickoff
383. 1:00:47call and I admitted that I was sleep deprived and and I went through waves in the call so so there was there were
384. 1:00:56moments where I was like totally dialed in and then there were other moments where my account director was like uh
385. 1:01:00Kyle any thoughts on that I was like so I'd say I'd say I was about 80% there but thank you for asking it went
386. 1:01:14well yeah Drew Drew learning AI yeah uh please the uh the the calendars for the uh for the AI Salon event are in the the
387. 1:01:24Mighty networks um there's a Calendar's area the silliness brought me here the values kept me here oh that's awesome
388. 1:01:31Papa JJ thank you yeah it's you know I the the the silliness is is all me I'm just I'm just a dumb dumb but um yeah
389. 1:01:41the values really did come out of um you know it's just one of those things where values are a really interesting
390. 1:01:51thing because you you can't you can make them up but if you just make them up then they're kind of like successories
391. 1:01:58posters you know hang in there baby um when values are valuable is when you've got a group that's already got them and all
392. 1:02:12you do is articulate who the group already is so those values we did those like 6 months after we started the the
393. 1:02:21salon and we said you know the the L's got a chance that it could end up growing and it and there's a there's a
394. 1:02:28funny thing that happens when something goes from like a small group to a larger group and if if if you grow too quickly
395. 1:02:35you have more people that don't know what it is than do know what it is and that's when that's when groups can get
396. 1:02:42in trouble and so you articulate those values to protect that but you're not just making the [ __ ] up you're saying okay who are we
397. 1:02:52well like to a person person we're all really curious okay curious curiosity we're willing to share stuff oh
398. 1:02:59generosity like you know what I mean it's like that's where it came from so anyway I appreciate that that's cool
399. 1:03:06very very cool Jim Ross who just said he knows cat gpt's birthday better than his own child's um he's also been playing
400. 1:03:15with chat GPT since November 30th um of last year and and he is not in the tech business he's in he he uh owns and runs uh Self
401. 1:03:27Storage uh places and trains other people to do that and he is killing it because he's in an industry where people are not actively
402. 1:03:40seeking out AI [ __ ] right they're looking for you know a half a basis point on the on the fed's announcement right to to get better
403. 1:03:51rates for their next loan and he's just killing it he wrote a book now he turned that book into a GPT what did you say what did you do Jim
404. 1:04:02the other two weeks ago two weeks ago he had a he had a meeting with someone this is [ __ ] crazy he had a meeting with
405. 1:04:11someone and then after the meeting he goes I know what I'll do I'll write them a song and he went too. a so that was
406. 1:04:18that's one of the things on on my list here let me I think it's sun. a let me just confirm that's the right you are he went to yeah
407. 1:04:31sun. and you you basically just put in what you want the song to be about and the style and it writes you a song with
408. 1:04:40lyrics and then it sings the song and he sent that to the client they like called back immediately he got the
409. 1:04:51work so it's just like you know how do you use AI I don't know you you you run your business and this is one of the
410. 1:04:59things Jim said that I find really inspiring he goes every day when I get to work I think how can I use AI to
411. 1:05:06solve that problem landed that client as well it was all in the jingle there you go and what he say and what he says is he goes so I come up
412. 1:05:17with a a problem I I got I got this problem I want to solve and he goes the first thing I do is I say how can I solve that with
413. 1:05:24AI and that's that's what's be called uh there there's a there's a course that's taught on this called being um AI
414. 1:05:33native I'm not anywhere close to it like it's a complete mind set shift where when you have a problem what my experience is is I've
415. 1:05:43got 58 years of gray matter kind of hardwired to problem solve the way I problem solve right so when a problem
416. 1:05:52comes into my brain the first first thing that happens is my gears start going like oh I I got to solve this and
417. 1:06:00then and then it's always an afterthought where I'm like wait a minute I bet I could do this with chat
418. 1:06:06GPT so so getting relentlessly curious about this stuff and just thinking about all the problems at work you know and
419. 1:06:13all these things you have to solve could could you solve it with AI oh I think that's a rad business hey
420. 1:06:19lunar stick what's happening good to see you this is from Kelly this is her her happy birthday latte that's pretty
421. 1:06:34grooving happy birthday chat gptt Happy First Birthday wait but it's b r b i r t l d happy first Berle day
422. 1:06:48chat GP happy first Berle day chat GPT hey howdy Kyle thanks for keeping me company today you're welcome you're
423. 1:07:00welcome yeah it it at one point I was I was working with my uh with my co-founder on something and and I'm like
424. 1:07:08I got to go and she's like what do you got going on I said I gotta go live she was like oh God all right does it use multiple
425. 1:07:20search engines databas bases when answering a question how does it decide um so V you want to see how large
426. 1:07:29language models work tap the screen if you want me to do I'll do a a primer on how they actually work and this is one
427. 1:07:35of the videos in the um in the underhyped series um I've got two videos in there one is how large language
428. 1:07:42models work and one is how the image generation things work when you understand how these [ __ ] things
429. 1:07:48work it's they're even more mindblowing they're it's insane um maybe do this in sunno for us oh okay we can do that oh the happy
430. 1:07:58birthday happy first burle day chat GP I kind of like to ring of that happy first burle day chachy
431. 1:08:14[Laughter] P all right Let's do let's do a little let's do a little chat gpt1 actually let's do a little large language model
432. 1:08:26101 this is how this [ __ ] Works um all right let me go find the [Music]
433. 1:08:42slide where do I where did I put them where did I put them oh h mark hi
434. 1:09:00there it is okay we're going to do that and then we're going to do that oh that dang it dang it Jim that
435. 1:09:12[ __ ] you know RI them oh I'll tell you what I'll tell you what all right you come down into the swamp you get some algae on your feet
436. 1:09:24I'll tell you that I'll tell you that right now I have a Jim Ross story met a dude in self storage and I said are you
437. 1:09:32are you using Ai No Jim Ross and that's awesome I I bet they did I bet they did and I wonder if they
438. 1:09:43were using AI if they're not they got to get their [ __ ] together that's really cool it's not often you meet people in
439. 1:09:53in Self Storage so yeah I would think it's a pretty small world cuno yeah cuno question is did they know me I'm
440. 1:10:01guessing the answer is yes there Jim well Jim's also had um like he's got a live show that he does he's done for
441. 1:10:10five years like he's he's like he knows how to do this [ __ ] so um okay let's let's get the happy birthday thing in
442. 1:10:18the background there that looks pretty cute so we'll oh yeah that's good look at that look at this people you see how this see
443. 1:10:27how good this is see what's going on here it's Kismet or it's whatever when when the [ __ ] comes together I'm pretty sure Kyle is not all
444. 1:10:38that wise okay all right chat gbt G
445. 1:10:54kpt if you're going to be in this business you should probably actually understand what those three letters mean
446. 1:11:00they're not just random it wasn't just Engineers just randomly came up with these numbers or letters the G stands for
447. 1:11:11generative the P stands for pre-trained and the T stands for Transformer Transformer you don't need to worry
448. 1:11:17about but generative and pre-trained you do generative means these tools are actually gener ating generating original content every time
449. 1:11:28they write something they're generating it new from scratch it is not copying and pasting and and the image generation
450. 1:11:35tools are not copying and pasting even though these the core Foundation models have been trained on lots and lots of work some of which
451. 1:11:46may be copywritten when it's creating it it is not copy pasting it is it's it's more the equip equivalent of you went to
452. 1:11:55school for 4 years and you learned all the Poetry of and and writings of Emily Dickinson and you just love her writing
453. 1:12:03and you just absorb yourself in her writing and then you leave school and you start writing and you kind of start
454. 1:12:08writing like her because you learned all that [ __ ] and it's all in there and when you go to generate something new it
455. 1:12:14filters through that that's that's kind of what it's doing so that's the generative part pre-trained means you
456. 1:12:21don't need to bring your own data knows all the [ __ ] but Kyle what does it know it knows as much data as they could get their hands
457. 1:12:35on so what so one of the genius moves that open AI did so the the Transformer the T in in GPT stands for Transformer
458. 1:12:44this was a technology invented by Google hey Pate invented by Google p works at Google that allows these machine
459. 1:12:54learning models to be trained on massive data sets and one of the things that Sam Alman did was he said hey if these
460. 1:13:00things get better with larger data sets let's get us a whole boatload of gpus and in order to do that maybe we should
461. 1:13:08cut a deal with someone who's got a bunch of them hey Microsoft want to do a deal most of that investment from Microsoft is not cash it's
462. 1:13:17gpus and so they then went out and they they found as much data as they could of the public internet and they put it in
463. 1:13:25all the formats they did to train it into this thing so pre-trained means um you don't need to figure out what to put
464. 1:13:32in it to get good answers out machine learning up to this point that's what it was you had to bring your own data and
465. 1:13:39then it could do remarkable [ __ ] but you had to be a mathematician well it's pre-trained for you okay and it's
466. 1:13:45pre-trained on like the collective output of humanity that put [ __ ] on the internet for the past five decades think
467. 1:13:54every Wikipedia article every blog post every just all of it anything they could get their hands on okay enough comedy
468. 1:14:03jokes large language model llm equals stands for L large language model so it's a large language model it's got
469. 1:14:10lots of language what else is a large language model human beings so one of the reasons chat gbt when you interact
470. 1:14:18with it feels like you're talking to someone is that you kind of are it's less a robot and more like a reflection
471. 1:14:25of all the [ __ ] people have said for the past five decades on the internet um okay now this is where it gets a little geeky so hang with
472. 1:14:36me there's the thing called the latent space and the latent space is a mathematical space so what you're seeing
473. 1:14:43on screen is a 2d an XY coordinate a 2d representation of the latent space and then you see those clusters of dots so
474. 1:14:52those clust clusters of dots are semantic meaning clusters and so so each dot represents what's called a vector and a vector is a
475. 1:15:04series of numbers that that represents that point in space um in this case the vector would have two numbers in it an
476. 1:15:12an X number and a y number an x and y coordinate the real Vector spaces are like a thousand Dimensions big so these
477. 1:15:21numbers are like these long string of numbers num that represent a point in space and the point in space are are
478. 1:15:29clustered meaning and so what that might look like is this in in this little thing here there's black dots and you
479. 1:15:36could have the word dog could be in the black dots um and it means the The Fluffy thing that lays there are
480. 1:15:42mathematical points in space okay so then how does this [ __ ] work so based on your prompt when you type a prompt in it
481. 1:15:54generates it uses a mathematical formula that some literal [ __ ] Geniuses have developed over the past decades to come up with a
482. 1:16:03probability and so it predicts what is the next most probable word to be the right next word from your prompt and and
483. 1:16:14how it does it is wild so how it works is like this let's say we put in the um thank you Victoria let's say you put in
484. 1:16:22the um The Prompt the quick brown fox jumps over the lazy blank well we know as humans because we're large language
485. 1:16:29models it's dog right we've heard that from typing class or whatever it was here's here's what the large language
486. 1:16:36model does it basically says okay based on that prompt the mathematical Pro probability is that all of all of those other clusters
487. 1:16:47are not relevant but there's the these light blue clusters here this light blue cluster here there's a high probability
488. 1:16:53that it's it's one of these points in in the latent space and within that there's there's one of them is the highest
489. 1:17:00probability so it's a 96% probability and if you look at what it is it's this big long ass number right so the large
490. 1:17:08language model goes this big long ass number is the next thing that should be there oh by the way that big long ass
491. 1:17:14number is associated with the word dog and so then it inserts the word dog in that spot and there's other words around
492. 1:17:22it that it could have been so there's a thing in in large language models called temperature if you turn down the
493. 1:17:28temperature it will only give you dog every time if you turn up the temperature it'll kind of randomly give
494. 1:17:34you one of the other probabilistic words so that's why some of them can wait uh oh we are multim modal come on Kyle I
495. 1:17:44know I gotta I got to update this because it's it at at this point the multimodal thing is it's just it's just
496. 1:17:51crazy yeah it's a good point anyway this is how the word part Works um it's a very good point P um but that's how it works so when when you
497. 1:18:02see hang on let me escape out of that let me let the Tik Tock freezes when I switch cameras um when you see chat GPT right like the
498. 1:18:15whole answer with all that [ __ ] it's doing that probabilistic thing like it's amazing how does what it does um that's why it hallucinates
499. 1:18:27because it doesn't know what it wrote All it knows is there was a series of numbers in mathematical space that were put together in an
500. 1:18:39order now the new things that they're working on I think are going back and looking at what it wrote and analyzing
501. 1:18:45it and scoring it and things like that so that's how they'll fix the the hallucination thing I'm so new to all of this and just wow
502. 1:18:53it's crazy isn't it it's just wow It it's like like when you when you learn how it works like the the first thing that I think
503. 1:19:04is oh that's like a calc it's just like a calculator right it's actually really simple you type [ __ ] in it it points to
504. 1:19:12some point in space and says that point in space is the closest one to where this thing should be and then stick that
505. 1:19:20in there it's really simple right except they're doing it on trillions of words and they've got it in this thing
506. 1:19:31because it's just in there in math and computers are really good at math it doesn't have to know the meaning and the image things are
507. 1:19:41similarly insane how they do what they do it doesn't know what it's drawing it's taking an amalgamation of like
508. 1:19:49thousands of randomly noise generated images that were based on images but they're now just random noise
509. 1:19:57amalgamates them all together and then Den noises them over like 50 steps using the same noising Al algorithm that that
510. 1:20:05you know encoded all of the images it doesn't know what it's creating that's why it's like you know
511. 1:20:12why does it misspell [ __ ] the fact that it can make a latte with Happy First Birthday chat gptt is is a [ __ ] miracle it really
512. 1:20:26is oh all right so guess you were afraid you ask that question
513. 1:20:40huh don't ask Kyle anything when he goes down those rabbit holes he does not come out theand is not shut up oh there's a
514. 1:20:47lot of people in here hello hey ticktock thank you for the gift appreciate the gifts that means a lot helps support the channel
515. 1:20:55um if you're new here my name's Kyle Shannon I am about 20 hours into a 24-hour chat GPT Aon I haven't done the full 24 hours but
516. 1:21:06you know I've done a solid amount I am [ __ ] crispy my voice is going and my ass is numb but I'm going to keep going
517. 1:21:13till midnight it is now 8:30 8:38 Sherry D thank you so much I know subscribe to this channel follow on YouTube AI learninglab dtt
518. 1:21:24thank you very much I appreciate that oh Victoria thank you so much those are very cool love that please write this in
519. 1:21:31Hindi and then explain to me how I know what you wrote is correct ask that wait please ex please write this in
520. 1:21:42Hindi oh okay all right please hang on Luna I'm going to do I'm going to do chat GPT because it's better at language
521. 1:21:54chat GPT 4 we're going to go in here I'm going to go hi wait hi please write this in Hindi and explain to me oh well I'm
522. 1:22:08going to do it I'm gonna so Luna I'm gonna do it with voice I want you to translate a sentence I say in English into Hindi and then in
523. 1:22:24English I want you to explain to me how I know that what you said is correct go ahead and tell me the sentence you'd like translated into
524. 1:22:37Hindi once I've done that I'll explain how you can verify its accuracy there was a man named Bob and
525. 1:22:44Bob had a dog that had three legs and he called him tripod
526. 1:23:00oh in Hindi that sentence translates to tripod to verify its accuracy you can one use an online translation tool input
527. 1:23:15the original English sentence and compare the result with the translation I provided two ask a Hindi speaker they
528. 1:23:22can confirm if the translation captures the essence of the original sentence correctly three look for key words words
529. 1:23:29like Bob dog three legs and tripod should be easily identifiable and correctly translated in the Hindi
530. 1:23:35sentence would you like to know anything else about translations or languages I'd like you to break the
531. 1:23:42sentence down tell me the word in Hindi and then what it translates to in English I'd like to to essentially
532. 1:23:48translate it back to English but a word at a time so I can understand how you got there
533. 1:24:06certainly come on you can do it you can do it come on chat JP let break down the Hindi translation word by word one e one 2 adamy man three
534. 1:24:22T was four J J Who's five n Bob Bob T was for grammatical structure or a nine Bob Bob again indicating possession and I can stop this and it
535. 1:24:37will so it wrote it all out right so there's there's all your written stuff so even though you can type it in when
536. 1:24:44you put it in conversation mode it it records a transcript of what you said and what it said back so um
537. 1:24:52yeah there you go whose name Bob was and Bob of one dog was whose three legs were and him he tripod called he referring to Bob as in he
538. 1:25:09called tripod he called the dog tripod twice today Kevin McAllister moments how does it keep blowing my mind
539. 1:25:19I don't know I lunis I'm the same way like even in this 24-hour session I'm just doing stuff that I just like I
540. 1:25:27think part of it is I'm punch drunk right because I'm just exhausted but but it just like that holy [ __ ] mind blown is the AI translator
541. 1:25:43still chat gbt oh Edson you asked before about the um the chat GPT app so couple of things about it so this is an iPad it's it's just an
542. 1:25:54IOS app make sure that you get the official chat GPT app there's a lot of Faker apps out there um so make sure you
543. 1:26:03get the official one there's a couple of things about it [Laughter] um there's chat GPT 3.5 and four so one
544. 1:26:17of the things that they recently did is even if you're in chat GPT 3.5 they're giving you the ability to put it into
545. 1:26:26conversation mode so conversation mode that thing I just did with the Hindi you can you can talk about it about
546. 1:26:38anything help me understand how the origin of the Hindi language and the origin of the English language were
547. 1:26:47similar at some point in history and how they split certainly both Hindi and English belong to the Indo-European language family
548. 1:26:59which means they share a common ancestor language known as Pro so so so so the reason to download the chat GPT app is is
549. 1:27:13that like here's a and this gets into the thing of of play where you just start playing a good play thing to do
550. 1:27:25is on your phone put chat GPT into conversation mode when you're driving like commuting to work and instead of listening to an audio
551. 1:27:35book talk to chat GPT for half an hour or 45 minutes have it create a game for you to learn something ask it obscure trivia
552. 1:27:45about [ __ ] um if you're using chat gp4 in conversation mode it's also connected to the internet I think Pi Pi is um so you can ask it
553. 1:27:59about internet things and then chat GPT 4 the the custom gpts you can also put them in conversation mode so if you've
554. 1:28:10got a custom GPT like okay here's a good thing let's say you're in sales we we just did this with Story vine we took our 50 page sales
555. 1:28:21documentation our 50-page sales process it's in a just an incomprehensible PDF file right it's p formatted poly there's pasted copy and
556. 1:28:33pasted [ __ ] in it but it's all the stuff you know for how we do our sales process because we sell into Pharma it's
557. 1:28:40complicated it's a it's a complicated sell we put that into a GPT so we've now got a storyvine sales coach trained on how we do
558. 1:28:51sales so I could be in my car driving to work as a Salesman for Story vine and say Hey I want to do some practice sales
559. 1:29:02talk and I want you to use our sales process to coach me through it that's a completely different way of using chat GPT it's a it's just a
560. 1:29:14completely different modality and and it's I it's hard to describe that talking to one of these things like we've talked to Siri and to
561. 1:29:25Alexa for years but they're like these hobbled shitty things that were designed to sell you [ __ ] they will begrudgingly answer other
562. 1:29:38questions like the weather but what Alexa really wants to do is would you like that delivered by Monday and it's those things are just
563. 1:29:48programmed as logic trees they they use AI to understand the voice and and track things but how they're programmed is a
564. 1:29:56big ass logic tree when they ask about product give them this product when they ask about pricing give
565. 1:30:05them the pricing and make sure you give them specials someone programmed all that [ __ ] these this is not that this is ask
566. 1:30:14it [ __ ] anything and it will start talking to you now will it always be right no it hallucinates my nickname for chat GPT is mansplaining
567. 1:30:25is a service it will confidently answer you even if it's wrong like all the good men in your [Laughter]
568. 1:30:39life I didn't come up with that one but I love that one and it really is like it really is true it's just it's just like
569. 1:30:46here's your answer you're like uh I'm pretty sure that's not the answer it's like you're absolutely correct that's
570. 1:30:53not the answer uh here's the answer like no that's not it either my apologies big bad on my part here's the answer chat GPT is fun when you
571. 1:31:08literally can't get it to give you the right answer for something that you know the answer to it's like uh come on it's
572. 1:31:13in there somewhere oh okay so here's a here's a thing that might help you get your head around how to talk to these things
573. 1:31:22so this was from Andre karpathy who was the the the lead dude at open AI that he left open AI then he came back to open
574. 1:31:31AI he's working on safety now um he was given a talk on large language models and someone asked why do
575. 1:31:39they hallucinate so much and in the talk he was talking about okay let's let's say you're trying to explain a math problem a particular math
576. 1:31:48problem if you ask a large language model tell me about this particular math problem it will go look at all of its
577. 1:31:57semantic clouds in its Laten space the thing we just talked about and anywhere where that math problem was talked about there will be
578. 1:32:05related words in those semantic clouds and it sort of pulls them in probabilistically and gives you your
579. 1:32:12answer and then a lot of the times it'll be wrong and so the way Andre karpathy described it it just makes perfect sense
580. 1:32:19to me he said said okay in all of the internet there are places where that math problem was talked about by the
581. 1:32:29Stanford Professor who solved who specializes in that branch of mathematics and solved that particular math problem there's also all of the
582. 1:32:39people that wrote about that math problem that were students that got it wrong so when you just say talk about
583. 1:32:45this math problem it will amalgamate all that [ __ ] together and just give you some jumbled mess of of [ __ ] back some
584. 1:32:53of which will be right some of which will be wrong and that's why when you tell a large language model hey I want you to
585. 1:33:01act like a Stamford mathematics professor that specializes in this branch of mathematics that is the person
586. 1:33:08that actually solved this problem now tell me about this problem what you've done is you've sort of filtered down the
587. 1:33:17semantic space to a much smaller set so all of that student work that was polluting the answer before is
588. 1:33:24effectively filtered out of the answer so this is why one of the big significant reasons chat GPT is not
589. 1:33:33Google is Google wants you to just have like put in the sort of just just a fact spam just give me the minimum amount of
590. 1:33:40[ __ ] you need and we'll figure out the rest chat GPT because it's trained on such a large data set the more context
591. 1:33:48you give it the better it gets so tell it to act like like a certain character tell it what you're trying to accomplish
592. 1:33:55tell it what you're frustrated with tell it what your business is tell it um you know ask it if it understands what the
593. 1:34:02problem is and then it it will likely say yes I understand the problem and here's my proposed solution okay great I
594. 1:34:08don't need the solution yet I just wanted to make sure you understand it all of that conversation adds to the context and
595. 1:34:14will make it answer better and better and better and better make sense lunatick like and share that would be
596. 1:34:28great if you would be willing to follow my channel awesome if you'd be willing to share this live awesome if you think I'm a total ass hat
597. 1:34:40awesome I don't like how he looks he sounds like Tom Arnold but he looks like Rosie O'Donnell what is this channel
598. 1:34:56I'm thinking of doing an ASMR Channel what do you think M tasty Punchy oh
599. 1:35:10yeah can you give it URLs you can let's go give it a URL I'll give you a URL she she she gave him the biggest URL you ever did see
600. 1:35:28huh come here little Lassie I'm going to give you a URL all right I don't know I just work [Laughter] here I gotta log
601. 1:35:48[Music] in sometimes I don't want it to give me a new answer uh thankful Jones so what you
602. 1:36:05actually oh wait I I don't know what you're talking about but if you don't want if you want chat GPT to give you
603. 1:36:12less creative answers go to Coop pilot. microsoft.com and put it in precise mode that's effectively like turning down the
604. 1:36:21temperature so it'll give you more predictable answers are there any apis that allow that allow slre create actions
605. 1:36:32yeah yeah so the the chat GPT for um API there's a new assistance API and you can you can program actions into
606. 1:36:46it you can also program them into gpts in this consumer front end if it's janky um but so it it'll get better out here
607. 1:36:56but yet yes you can and and in fact the the gpts the fact that you can upload knowledge to the gpts and connect
608. 1:37:03actions to them and you've got full access to all the gp4 capabilities is mind-blowing to me um okay can you give it a URL so
609. 1:37:15right now I'm at chat GPT I'm in chat GPT 3.5 mode chat GPT 3.5 is is not connected to the internet so if I say tell
610. 1:37:24me about story vine.com which is my company it yeah I'm sorry to the best of my knowledge update in January 2022 I do not have any specific
611. 1:37:37information about story vine.com but if I flip this over to chat gp4 um and now I say tell me about story vine.com
612. 1:37:51first thing it will do doing research with Bing searching story vine.com so it's now on my website reading my
613. 1:37:59website story Vine's a unique platform that combines hybrid video production technology Story vine now userfriendly
614. 1:38:11dashboard guided storytelling Story vine guides the users and telling constituent stories using their smartphones the
615. 1:38:18guided approach ensures cost solution in summary Story vine provides an Innovative platform yep that's kind of
616. 1:38:27it um did we get any in Auto Magic in there um tell me about some of the marketing positioning of this
617. 1:38:46company doing research with Bing how do you get chat chat GPT Dolly to spell text correctly just keep trying over and
618. 1:38:54over again one trick that I heard is um well a anything that any text that you want in the image put it in quote marks
619. 1:39:02so it knows that's the text one trick that I heard that improves it from like 65% to 80% is put the the quoted text at
620. 1:39:12the beginning of the prompt and at the end of The Prompt like like say you know this text on image description here that
621. 1:39:21has this text right so put it at the beginning and the end uh can't subscribe to gp4 what's next best Bing yes since
622. 1:39:30it since it has DOL 3 yeah exactly so SEC sigura so yeah they they've closed down new subscriptions to chat gp4 so
623. 1:39:39get on the wait list get the to the wait list and then um and then co-pilot microsoft.com or chat. bing.com they're both the same
624. 1:39:51thing um that is chat gp4 but it's Microsoft's version of it but it does have Dolly inside it um and then and
625. 1:39:59then that's that the the only way to get [ __ ] spelled correctly is um just keep say try again try again try again try
626. 1:40:09again all right why don't you reprompt the image and try again and then sometimes that'll work and then sometime
627. 1:40:18sometimes I just bail so sometimes Mark the answer is is not um is to [ __ ] bail on AI fire up Photoshop and stick some letters on
628. 1:40:29it that's also a solution sometimes we get a little seduced that these tools are the only way to do it Story vines marketing positioning is
629. 1:40:38centered around three elements structure speed and [Music] scale so anyway yes it can see into URLs you can also use um URLs in um here let
630. 1:40:55me show you something where is
631. 1:41:12explore the Denver construction advisor so this this is a GPT that I made so my my co-founder Monique she's got a
632. 1:41:22construction site next to her house where the construction company are just a bunch of [ __ ] they just keep
633. 1:41:27breaking code violations and you know stacking [ __ ] in their yard driving heavy machinery over the tree roots
634. 1:41:34whatever they're just breaking codes right so the way we built this um this GPT is we went and we found the Denver
635. 1:41:46construction codes website and so I I literally just went into this thing and I said um I want to create a thing that
636. 1:41:55can help me figure out if a construction company is violating construction codes and use this website as your Source
637. 1:42:03material and it just wrote it up and it did it and now it will you know let's see can they work on construction on
638. 1:42:10Sundays in Denver so we'll just see what that says doing research with Bing so it's now going out see Denver
639. 1:42:17construction regulations working on Sunday in Denver construction activities are permitted on Sundays as outlined in the
640. 1:42:25city's construction rules the Denver revised Municipal Code chapter 36 specifies noises allowed between 8:00
641. 1:42:33a.m. and 5:00 p.m. on weekends which includes both Saturdays and Sundays I mean do you know how long you would have
642. 1:42:39to [ __ ] Google and read through the Denver construction codes to find that and boom there it is done and how did I program it here's the
643. 1:42:51website go look at it that was that was the sum total of my programming input into this particular GPT astounding astounding astounding what
644. 1:43:03what do you want to do we can do [ __ ] I can we got time I'm going I'm I'm here for another three [ __ ]
645. 1:43:12hours what have you people done to me I might need to take a little break take champy out take a pee you know
646. 1:43:22you created a Karen GPT totally totally oh my God Kyle stop inspiring me to to do more thinking cuz Wheels be spinning
647. 1:43:32and and I'm sleeping soon sorry thank you Apple user I appreciate the Roses the rose by other name would smell is sweet you can make
648. 1:43:42Mone all right can you ask it to search Google I don't know um Search google.com for
649. 1:44:00um links to Denver construction guidance I have never tried that I don't know why you would do that doing research with
650. 1:44:16Bing so I guess Bing went to Google and did the search I found some things so I I don't it might it might
651. 1:44:24have used Google I don't know why you'd want to do that um cuno I'm with you cuno be be nice to Sky that's what I call her
652. 1:44:38um I'm afraid to put my VR head goggle with the AI app in it why yeah go in there it's the worst thing could happen you never come out
653. 1:44:51yep pretty much it my favorite are the the V the VR videos on on Tik Tok and YouTube where they try to have you know
654. 1:45:00Aunt Marge try on the the goggles and she's like walking the plank and then she gets to the end and she jumps jumps
655. 1:45:08into the TV those are my favorite videos oh my God I mean that's one of the one of the if you watch the the Vision Pro um
656. 1:45:24launch from Apple one of the things they did brilliantly is they designed that Hardware knowing that every Hardware to
657. 1:45:35that point cut you off from reality and they said Hey what if we didn't what if we included this in reality and then they even put screens
658. 1:45:46on the outside of the goggles so it looks like your eyes are there that's that's like a very Apple move right when
659. 1:45:54I first saw that I was like I was like oh that's cool they they it has like opaque it has like LCD things where you
660. 1:46:01can't see through it sometimes and you can see through it sometimes so you can see their eyes and then I realized wait
661. 1:46:06a minute no there's a logic board across across your eyes you can't see through that [ __ ] there's chips there and I was
662. 1:46:13like oh it's screens they actually put screens on the outside of that [ __ ] thing it's amazing because they recognize that
663. 1:46:22being in those things is completely isolating well what if it wasn't that's the part of the genius of Apple is they
664. 1:46:30just ask [ __ ] like well should it be like that no okay well let's [ __ ] find a way to do it I think there was
665. 1:46:37something like 5,000 patents in the in the Vision Pro goggles [ __ ] amazing Kyle what do you want what do you want for Christmas do not say
666. 1:46:49Turtles you're damn right I'm going to say Turtles what do I want from my friend the zombie Jonathan you're looking good
667. 1:46:57Jonathan just got an awesome face paint job what do you think I like turtles all right you're great zombie good time here
668. 1:47:05at the waterfront Village open for the next 11 days all right Nancy thank you for that Nancy told me to get my ass up and take
669. 1:47:17a walk around all right let me let me I'll be right back let me go let me go take the dog out and sort of take a pass
670. 1:47:23around the house I'll be back in like a minute or two whatever Kyle we should come up with a list of predictions for
671. 1:47:28one year from now from AI that's really interesting my favorite VR is Supernatural I don't I haven't played
672. 1:47:34with any of the VR stuff cuz I don't the isolation Stuff wigs me out it doesn't wig me out it just I don't it's not
673. 1:47:41appealing to me but the Vision Pro thing that looked like a pretty elegant solution so all right I'll be back Talk
674. 1:47:47Amongst yourselves listen don't troll each other be nice don't be [ __ ] don't be douche and uh yeah entertain yourselves I'll be back CH
675. 1:48:01about come on let's go out let's go for a
676. 1:48:19walk
677. 1:48:49than
678. 1:49:19e
679. 1:49:49SC
680. 1:50:19oh
681. 1:50:49spe
682. 1:51:18spe
683. 1:51:34for
684. 1:51:48the
685. 1:52:03oh yeah yeah that's
686. 1:52:18right
687. 1:52:48it
688. 1:53:13uhuh all right
689. 1:53:46look at the new CLE cleaning sheets they tiny I'll just put two in see if it works put see if it works no it never works with one come
690. 1:53:58on all right back at it all right bye see you are you sleep
691. 1:54:14no you love Daddy don't you you love Daddy when you come hang out look at him I know he can't get enough of
692. 1:54:26you he such a little trick come on you little dumb Dum all right I'm back I'm back back back back back back
693. 1:54:36back back back back oh you know what we should do I know what we should do let's go look recharging batteries mom swore Mama Kyle
694. 1:54:54[Music] did he fall out on us I'm about to no everyone just say bye just stop how many people are here all right we lost some people that's
695. 1:55:06okay so do me a favor share wait yeah we can't hear you oh my God he's alive it's a miracle champy Can't Get Enough me
696. 1:55:16neither all right let's go let's go look at the AI Salon this is the AI Salon if you have not been to the
697. 1:55:28AI Salon you should join if you want to join a people of cool badasses that are trying to figure out this AI stuff that
698. 1:55:36are curious and generous and love to explore [ __ ] and love to collaborate and are super inclusive and have lots of
699. 1:55:43empathy yeah that's us um but let's go look at the show and tell Channel and see if anyone been showing and telling
700. 1:55:53steo I'm sorry to tell you you have a URL that's a good one Mike knee happy burle Oh happy burle day oh that's
701. 1:56:06really good oh it's a it's a burle it's a btle that is a disturbing creature that's fantastic oh
702. 1:56:22man can you make it more birdie I think it's perfect amount of Birdie that's very cool I'm Kyle I'm an AI holic I have no
703. 1:56:33regrets that that explains this well look at the little llama multimodal who did this
704. 1:56:48Daisy awesome wow 24-hour birthday party oh yeah hang out come on llama come on Turtle where's the Hedgehog let's make this party
705. 1:57:04happen this is so cool so this is one of the things that that I find super inspiring about this community in particular
706. 1:57:15is this channel very much started as just kind of me like talking into a phone um and it it has evolved into this
707. 1:57:27this back and forth where I'm still talking into a phone um but this is there's a real community and like as
708. 1:57:37jokes come up or things happen people are creating stuff because you can create stuff instantly they're creating
709. 1:57:43it instantly sharing it in instantly and it's just generating all this stuff that's actually that actually reminds me of my
710. 1:57:53seventh grade classroom where where we had tr8s on every desk that's sort of the the 1960s version of it but I I had
711. 1:58:01the 1970s version of it heyi learning lab marthon
712. 1:58:18nice oh that's great Sherry Banks Irregulars pan TOS llamas hedgehogs and turtles oh the wait where's the llamas
713. 1:58:29there's a deer there I guess the llamas in the thing that's really cool all right these are ones I've seen very cool
714. 1:58:40what's going on in the Irregulars Channel anything down there so this is a private channel for the Irregulars so the Irregulars are
715. 1:58:51people who show up here regularly if you're a subscriber to the channel you get access to this or if you just show
716. 1:58:56up here a lot and you're a good person I'll give you access to this yeah these are cool show cool things and that's the way it is
717. 1:59:13Dar happy first anniversary Christopher Ramsey that's good that's I actually I feel like I'm in better shape now than I was this
718. 1:59:26afternoon this this afternoon was rough my eyes were not working they're all bloodshot people are like he doesn't look good oh that's
719. 1:59:41cool nice very cool all right so if you're wondering what this is this is uh this is a community called the AI
720. 1:59:52salon and um you can get there if you're interested in joining some badass people that are learning all about this AI
721. 2:00:03stuff and are curious about it you can get there here AIS salon. mn. it's still buffering is or is it just me um I don't know you tell me
722. 2:00:17people I can shut this [ __ ] down and restart the uh restart the router there's a hedgehog in show and tell all right
723. 2:00:27cool I have the well breaker on again no it isn't you're fine good on my end no it's Saloon the air Saloon welcome to the air
724. 2:00:41Saloon my name Carl Shon this is the a oh nice Hedgehog nice Hedgehog Joker why is he all
725. 2:00:57dirty oh is this my this is my 24hour this is me after 20 hours this is me right now is that is that what you're
726. 2:01:04saying I like it look at the dirty shirt that's [ __ ] awesome you're okay Kyle okay awesome um
727. 2:01:23[Music] oh that's nice that's really nice Tim shffle I've had a dark thought about Ai
728. 2:01:39and then I went nah no worries just seems scary oh nice nice nice llama the dolly llama get it that's really good nice one Clifford
729. 2:01:56I like it that's solid Let's uh let me see which one of these we want to do let's do this
730. 2:02:13one to I'm G to try something here oh you know we're going to do we're going to do we're going to do a cool
731. 2:02:21little experiment we're going to do cool little experiment all right um birdie oh you know what I like to do when I'm
732. 2:02:30grabbing others people's stuff is I like to grab their name so this is Mike KN so we're gonna we're going to do save image as Mike
733. 2:02:47KNE bird Turtle whatever the hell he called it okay now we're going to go into Photoshop if you haven't seen um
734. 2:02:59generative fill in Photoshop we're going to do that now because that's pretty wild it's wild it's wild and wacky stuff wild it's wild and
735. 2:03:12wacky we're going to do new file can't be open because it was moved fine L new create paste uh what happened what happened
736. 2:03:27what happened what happened what's going on here oh I gotta got to open it I didn't hang on sh hang on people I'll figure it
737. 2:03:37out what' you do Thursday night I watch some old guy try to figure out how to use Photoshop it was embarrassing that's good all right let's
738. 2:03:47see there uh oh we got some new people wanting to join the AI Salon let me go to desktop mic tree burle okay there we go there's our little
739. 2:04:01thingy now we're going to grab our this thingy and we're going to go ratio we're going to go 169 and then we're going to
740. 2:04:10go out and we're going to go out and then we're going to go generate Bang and so this is now going to fill in the left and the right of this
741. 2:04:29image it's going to give us three options boom isn't that amazing it's like dang it's pretty good all right oh that one looks pretty
742. 2:04:39good I think that's all right that's the good one okay so there we go so now we have a 16x9 image so this was rendered as a
743. 2:04:48square image in DOI as the bird Turtle the bird turtle and then um I put it in 16 by9 because of what
744. 2:04:58we're going to do next so what I'm going to do is I'm going to export this quick export as a PNG which is mostly what I do with
745. 2:05:07Photoshop anymore I didn't I got I used to do print production and I'm so glad I'm out of that [ __ ] cuz it
746. 2:05:15is it's so complicated now cuz all the digital stuff has just made it so complicated it's so complicated people
747. 2:05:23all right I don't need to save that because we're just doing a sparent and now we're going to go over to Runway ml so Runway ml is a text to
748. 2:05:37video a video generation tool if you haven't heard of it and you can also do image to video and it's also got this thing
749. 2:05:48called motion brush so I'm going to upload a file so we're going to go grab Mr PNG boom and I don't quite know how to use
750. 2:06:02this thing yet but all right so we're going to do motion of five yeah I guess that's good that's camera motion so let's do
751. 2:06:22let's do a horizontal pan or a horizontal it's not a pan it's a whatever shot to the right save that and
752. 2:06:35then we're going to go into this thing called motion brush so if I click this tap anywhere to begin I now have this here little
753. 2:06:47paintbrush and so I can paint into here what's going to be animated so I want his head animated so I'm just going to paint his
754. 2:07:00head I'm going do the little feathers we'll do around there uh maybe I shouldn't do the feathers let me get the
755. 2:07:09Eraser We'll erase that [ __ ] out of there all right so it's just his head but then let's have the wing move as
756. 2:07:20well oh we should do his little legs so we'll do [Music]
757. 2:07:36Wing okay and then I don't know if there's a way to make this skinnier other there is sorry I just up here there's a little
758. 2:07:48thing thing to make it skinnier so I'm going to just do the little tail right there and then we'll make it a little wider
759. 2:07:56again and now I'm going to do is the little [Music] legs and then we'll do that leg and then we'll do that
760. 2:08:11leg and then we'll do that leg okay so we got our little we got our marching orders and then I've got it so
761. 2:08:23horizontal we'll do a a left thing oh let's see we'll do a left like that I don't know what any of this means what's that
762. 2:08:40one oh proximity now we're not going to mess with that we'll just make it go up a little bit I don't know what up will do okay we're
763. 2:08:52going to save that okay so now so we have our photo in there the photo that was made in Del it was extended in photo
764. 2:09:00slop and now we're going to generate this video after painting in where we want the motion I have I have very mixed results
765. 2:09:11with with Runway the the the video stuff right now is still a little wonky but but this motion brush thing is actually
766. 2:09:18pretty cool cuz the fact that you can just stabilize the background and say I want the thing to move is it's pretty pretty
767. 2:09:27something it would be nice if it recognized that it was a bird and made a wing do do a wing thing but we'll
768. 2:09:37see the Hedgehog doesn't represent you but the new one does why am I afraid uh uh new activity Karen White oh that's very cute so so
769. 2:09:52look that's how to be nice to the host let's see what this one is that's very good that's very good that's me very nice Joker no this is you
770. 2:10:08Kyle I like this the cigarette and the earrings really nice touch my hair doesn't any look anywhere near that bad it is freshly
771. 2:10:23glued oh man that's good wait where were we we were at Runway ml where is it you ever lose a tab we done oh yeah we've got video
772. 2:10:37Let's Go full screen let's see how bad this is come on don't suck it would be nice if you didn't suck that's kind of cool it's
773. 2:10:50it's not really moving oh that's probably because I did a pan the feathers kind of morph here they
774. 2:11:05have black bars there they don't and he and he grows feathers on his back while he rotates and his head changes this shit's so wild um
775. 2:11:19again this is this is the kind of stuff I think we're going to look back fondly on this era and go remember when videos were
776. 2:11:28weird cuz they're going to get real perfect real quick jrb 185 great question what did JR 185 ask my Photoshop is older I don't think
777. 2:11:41there's that feature it's 2024 you need to have photosho 2024 we should create a prompt and then use it I'll use it to post the results
778. 2:11:53oh that's a cool idea does this mean I'll live forever we're getting closer look in Leah's art okay going into the AI salon so one of
779. 2:12:11the things that the AI Salon has which is super cool is guilds so there's an Art Guild a writer Guild a business
780. 2:12:17Guild a mechanic Guild things like that so in Leah art there was a a Mad Libs challenge um started today so people are doing the
781. 2:12:29challenge what material was used to train chat GPT Does it include books college courses movies TV shows so it
782. 2:12:38was basically I it's a I I don't I don't know if it is discoverable if or if it's private knowledge oh these are cool
783. 2:12:52wow those are great damn look how cool that is so oh was this the one where you did okay so so we'll this is something
784. 2:13:01we'll do so one of the things you can do in Deli is you can make a hippie and then you can just say make it
785. 2:13:11hippier so Vicki just did this and so she made it hippier make it hippier so now we're full Jerry Garcia make it
786. 2:13:25hippier and the hippiest that is fantastic oh my God that's fantastic wow that's so cool um let me
787. 2:13:39flip this into big mode I like big mode so there's that one oh that's cool I used piie to fill in my words oh that's
788. 2:13:50nice wow super cool so this is an in adamant object that comes to life is what the the Mad Libs prompt is here Be not Afraid put this
789. 2:14:02together a Whimsical book that suddenly Springs to life wow that's cool yeah doing these doing these prompt challenges is really
790. 2:14:15fun a grumpy mop and a Roomba Meed a DI disco that was me doing even happier do it happier happier here's where I
791. 2:14:28started and then this was my my final happiest Kyle could you go over how to upload files and configure them on
792. 2:14:36chatty I've uploaded text PDF it does not appear wait and it doesn't see it I have to make manually could you go over how to upload
793. 2:14:55files oh to the configure section in chatty oh okay okay so let's go so we're going to go we're going to go build a GPT I'll build a
794. 2:15:10[Laughter] GPT Sher bear you have enough for okay [Laughter]
795. 2:15:25okay so all right in chat GPT um you can upload files and and chat GPT can see those files and so one of the
796. 2:15:39things that's happening is someone created a GPT and it's not really seeing the files wait I saw on TED Talk what
797. 2:15:45was that I saw in Ted Talk that something I saw a TED Talk That 7,000 books were
798. 2:16:01the original start to building an llm oh it's way more than 7,000 books like they they went and they tried to
799. 2:16:08get all of the public internet so they cut deals with data sources and things like that so it's it it hasn't been
800. 2:16:16trained on video so one of the things that I think Gemini from Google I think one of the reasons it could be
801. 2:16:25good is that they've got all the YouTube data so and they've got all the YouTube transcriptions and so if they have a
802. 2:16:32truly multimodal large language model that they can train up on Youtube video um it could be like just a wholly
803. 2:16:41different thing um so so that's part of my speculation is with with all these companies like Twitter has Twitter data and Elon Musk
804. 2:16:52cut off API access for any companies to use Twitter data why because he knew that that data was valuable and his chat
805. 2:17:00GPT is going to be the only one that can incorporate Twitter data that's actually pretty smart um so Google the same way Google's
806. 2:17:09got all of our data they've got all of our data they've got Instagram they got YouTube and they have Google search they
807. 2:17:16got all the [ __ ] they got Google Maps they know where we go they know what we do like you know they're training that
808. 2:17:22[ __ ] on all of that data so um so the so these large language models are going to have very different flavors okay so chat
809. 2:17:31GPT it now has these things called gpts custom gpts which are these specialized versions of chat GPT and one of the
810. 2:17:40things you can do with them is you can upload your own data so I'm going to go to explore and then I'm going to go to create a custom
811. 2:17:51GPT and there's a there's a couple of ways that you can do this so here here's one of the things that um that I've found how do I explain
812. 2:18:08this in the configure mode if you upload your files here and then you write into the prompt here that you uploaded the the the PDF
813. 2:18:20here and you only want it to use that PDF and nothing else it it sometimes works better than if you're over here
814. 2:18:29where it's chat GPT writing the code for you and you upload something here um I what I'm finding is that this GPT
815. 2:18:38Builder with the with the large language model front end is good for the for initially setting up the GPT but for
816. 2:18:48refining the GPT I find it better to go in and actually modify the prompt like add rules to the prompt so so what I'm going to do is I'm
817. 2:18:57going to I'm going to create a um I'm going to create a GPT for the screenplay that I wrote so I'm going to upload a PDF of
818. 2:19:08this screenplay called salmon flying East uh uh uh uh [Applause] uh where what did I spell it wrong oh I
819. 2:19:24did slamming Salomon flying Solomon boy tired much oh
820. 2:19:39[Music] no PDF document there we go all right so I've uploaded my PDF of my screenplay so I'm going to
821. 2:19:56say I want to create a GPT that allows someone to ask questions about the story in this screenplay and
822. 2:20:12anytime you answer a question I want you to generate a storyboard from the scene that you're describing period the look and feel of the
823. 2:20:28storyboard should be like a watercolor illustration and in 16x9 wide
824. 2:20:43format so this should because I'm starting from the beginning with the PDF and this is my first instruction this
825. 2:20:53should work what I found is that if you build the GPT first and then you add the knowledge and and you do it from this
826. 2:21:01front end not on the configure side that it sometimes doesn't understand that you really want to use that that thing and
827. 2:21:08only that thing so what I would do if I were you is I'd flip over to configure and add in you know con see see here it
828. 2:21:16added the constraint form the GPT should focus solely on the screenplay content provided by the user and should not include external
829. 2:21:25references answers should be concise and accurate clarification the GPT should request clarification from the user
830. 2:21:33personalization this okay um gpts have changed in the last 12 hours so so you know how we talk about
831. 2:21:46just if the [ __ ] isn't working like come back in a month or two this stuff is changing in real time so what what used to happen with these
832. 2:21:56gpts is they would write just like a really generic prompt here but they were writing more specific stuff behind the
833. 2:22:03scenes I think they're now revealing the full prompt here because they changed here let me make this bigger they changed this to show role
834. 2:22:15and goal this GPT is designed to assist users with questions about a screenplay it will provide the answers um and generate storyboard
835. 2:22:25images based on the scenes being described constraints see how it's adding like these these um kind of Json
836. 2:22:32kind of you know these structured kind of components um it didn't have that before at all the GPT should focus solely on this
837. 2:22:41guidelines answers should be concise clarification we should ask for clarification ification personalization
838. 2:22:49it should maintain a friendly and helpful tone focusing on delivering the visual so um so I'm going to say the GPT
839. 2:22:56should maintain a friendly and helpful tone and match the tone of the dialogue within the screen
840. 2:23:11play focusing so I'm I'm sort of tightening it up even before we get started here can you show me a storyboard of the climax
841. 2:23:22scene what happens in the first scene of the screenplay describe the main character's introduction describe the
842. 2:23:28setting of the final scene those are kind of bad but that's fine okay so let me go shrinky shrink here okay so the
843. 2:23:35the way you make these things so what we did is we were over in the create mode we described what we wanted we uploaded
844. 2:23:42the screenplay it said sure I'll do that it said next let's decide on a name for this GPT um would you like to go with this or
845. 2:23:52do you have a name in mind um the screenplay is called salmon flying East so that should be in the title but give me some ideas beyond
846. 2:24:07that so now it's going to help us come up with a title for this GPT
847. 2:24:22[Music] how about salmon flying East screenplay
848. 2:24:38decoder now let's make an image so it's going to generate an image I wonder if it's going to put a salmon in there [Laughter]
849. 2:24:58it did I don't like it but whatever it's fine okay okay so now so the way this works you you sort of invoke your thing here
850. 2:25:08you can go into configure mode if you want to change things like I showed you before it's got the PDF as the as the
851. 2:25:16thing it's allowing web browsing which I'm going to turn that off because I don't I don't really care if it goes out
852. 2:25:22and browses the web um because I just want to talk about the screenplay um code interpreter is interesting I could turn code
853. 2:25:30interpreter on and what that would allow it was it it would allow someone to say uh hey give me the first three scenes of
854. 2:25:38this screenplay in a Word document so I'm going to turn that off um I'm going to leave on image generation and then actions I'm not
855. 2:25:46going to deal with oh it's got additional settings they brought these back these were gone ah so this is this is interesting
856. 2:25:54that they brought this back this this was gone I told you this this changed in the past 12 hours so when gpts first
857. 2:26:00launched this was here and then like within an hour it disappeared so what this what additional set look how
858. 2:26:07they've buried this so so way below everything so you set up your GPT here and then way at the bottom below all
859. 2:26:14that stuff is this teeny tiny little button flip that down and it says use conversation data in your GPT to improve
860. 2:26:21our models so I'm going to turn that off so this is the thing this is the data privacy option so I'm going to turn that
861. 2:26:30off so that they're not going to use this in their training data and so now so this is now a fully functional
862. 2:26:38GPT um and I can test it right here this is called the playground so you build it here and as you make changes here it
863. 2:26:44updates it in real time here so I'm going to say um what is the inciting incident for the
864. 2:27:01main character and what is his name and in theory this should tell us the inciting incident and then so searching my
865. 2:27:16knowledge so it's reading the screenplay and then it should make us a storyboard frame from the scene that it
866. 2:27:25describes the main character of salmon flying East is named Jack he's described as a weathered ruggedly handsome man the
867. 2:27:32inciting incident in the screenplay occurs when Jack um while in the cabin of the fishing boat is awoken by his alarm
868. 2:27:41clock the moment sets the tone for his routine life as a fisherman and introduces us to his world so so I'm
869. 2:27:50going to go back and I'm going to say
870. 2:28:04um hello okay the GPT successfully answered a question about the screenplay but it did not automatically generate a storyboard
871. 2:28:22frame can you update it so that it does without asking the end user which is probably annoying but that's okay this is my [Laughter]
872. 2:28:40GPT updating GPT I'm going to actually go back and look at the once it's once it's updated I'm going to go back back
873. 2:28:46and look at the at the back end and see if it added things to the to the prompt I have a feeling I have a feel they're
874. 2:28:53definitely improving this this this has gotten better than than it was a bit ago feel free to test it out okay let's
875. 2:29:01go look at configure so roll and goal here we go it provides detailed answers and automatically generates storyboard
876. 2:29:10images for each scene described without needing additional prompts from the user so it looks like it added that in
877. 2:29:17there clarification personalization okay so let's go um what is the first interaction that Jack
878. 2:29:33has salmon flying East screenplay hey you putting any oral in the car you you going down to the mall today
879. 2:29:41yeah you're going in the mall yeah I call it pencil Tucky okay jack's first interaction salmon flying East occurs in a crowded scene he
880. 2:29:57moves aside to let ping pass but the crowd surges pushing them together Jack trying to make the best of the awkward
881. 2:30:02situation sure beats Disneyland offers his hand to Pink who's too shy to take it but nods at this moment snake Jack's
882. 2:30:10partner arrives and amusingly notes that Jack and ping have already met sure beats Disneyland I don't think she
883. 2:30:21said that I think he said that look at that beefy arm Jack's a hottie that's not that's not very that's
884. 2:30:33not that didn't really capture it I also don't think that's the beginning I don't think that's the first interaction
885. 2:30:45um huh interesting so I'm going to go back I'm going to tell it um I'd like to make two changes period it
886. 2:30:58should answer the question about the scene first before it creates the image period the current version creates the
887. 2:31:09image first and then describes the scene period the second change I want to make is that I want to incorporate the slug line from the
888. 2:31:20screenplay from the location of the scene that's being described EG exterior boat
889. 2:31:37dock Etc let's see if it can do that let's see if it can do that dear people you understand sen uh mhm yeah yeah I like
890. 2:31:50them ters T tasty mhm mhm all right what are you are you all commenting down here I I I got nothing I see nothing here I'm
891. 2:31:59I'm not looking because I'm off doing my own [ __ ] answer the questions about the screenplay scene first include the slug
892. 2:32:07line from the screenplay for the location of the scene being described okay feel free to try it out so it
893. 2:32:12should now give us something that looks sort of screenplay like um let's go what is the main character's first flashback
894. 2:32:32moment searching my knowledge base so I'm off reading the screenplay it would be nice if you could change that little
895. 2:32:39um line instead of searching my knowledge reading the screenplay would be kind of cool Sherry D thank you so much appreciate
896. 2:32:54the gift we got water cooler issues got heating issues keep doing what you need to do oh you guys are all dealing with
897. 2:33:07dealing with [ __ ] [ __ ] in in real life oh Lordy all right the main characters first flashback occurs when Derek recounts his past
898. 2:33:18experiences in Saigon in 1971 the flashback introduced in the scene ping is the character storyboard description storyboard image okay
899. 2:33:28location exterior fishing boat key elements let's create the storyboard image creating the storyboard image
900. 2:33:35creating image so this is actually really good because whatever it did to update this it it describes the little Scene It
901. 2:33:46gives the storyboard description and then it gives us you know it it like reminds us that it's it's a um it's a movie and then here's key
902. 2:33:59elements yeah that's really good that's really good holy [ __ ] this is cool that's really good all right so so
903. 2:34:16now now that this is created I can save this I can share this in public I can confirm it you can also put it only I want to
904. 2:34:26see this I only want people with the link to see this but now this is a publicly available screenplay decoder
905. 2:34:33screenplay assistant and storyboard illustrator in a watercolor style I hate that let's go change that edit the
906. 2:34:41GPT we'll go to configure screenplay assistant uh let's see um explore Kyle
907. 2:34:56Shannon and Brian dyes um haunting
908. 2:35:13screenplay all right and then I'm going to update that bad boy boom confirm it confirm it what's happening why you
909. 2:35:27not confirming save oh loading confirm it looks like my internet went poo poo am I all froze up Jack has a
910. 2:35:44mullet ping isn't a Vietnam name maybe uh HOA or thy oh great thank you very much cool I'll survive though yeah see we wrote this in 1991
911. 2:35:58where you didn't where researching [ __ ] was actually really hard we got water cooler issues we got heating issues yeah we are
912. 2:36:10here how many people are here oh there's 100 people in here you guys are awesome um so welcome so if you're new here my
913. 2:36:17name is Kyle Shannon I am now 2 hours away from 24 hours of chat GPT Madness um and I can't seem to save my my thingy
914. 2:36:35here oh s flying East oh yeah there we go explore Kyle Shannon and Brian Dy haunting screenplay so let's try it one more time
915. 2:36:47out here we'll say um uh [Music] um uh tell me about a tell
916. 2:37:06me about a scene with the locals let see if it does anything
917. 2:37:21here trying to get the east as in Asia by going west yeah it was something like that can it access published books like
918. 2:37:34newer ones it can access well anything that you've got a URL to or any file that you've got so if you have a
919. 2:37:42PDF of an existing book you could Absol absolutely make one of these about it is that going to get you in trouble it
920. 2:37:49might if you try to sell it if you try to monetize someone else's work but I don't know it's making the image first again I don't
921. 2:38:05know I don't know why it does that under no circumstances should it make the image first period you should always describe
922. 2:38:22the scene first as we've agreed to and only after describing the scene create the image not create
923. 2:38:38damage no after wait and only after describing the scene create the image
924. 2:38:53boom it does it does it because it's janky that's exactly right Source Camp um one of the things you should
925. 2:39:02know about all this chat jpt stuff you can make money with ch jpt as sexy as it is and as remarkable as it is and as
926. 2:39:11profoundly powerful as it is it's still a janky piece of [ __ ] all of these gen AI tools right now are very very early
927. 2:39:19very very early so do not um get wigged out when you're poor little um when your poor little doohickey
928. 2:39:31um doesn't work right or when it doesn't spell right all this shit's going to get better but you know there you go all right um let's see uh
929. 2:39:46is there a fight in the movie creating damage is a writer's job yeah man it's an electric radiator spe space
930. 2:40:02heater won't work ear won't work either well that sucks so many times in my prompts I have to type this is imperative searching my
931. 2:40:13knowledge so at least it SE ing its knowledge first here I don't think it did that
932. 2:40:30before I just had one of those moments where it's like it's quiet too quiet cuz my ears are kind of ringing cuz I've
933. 2:40:36been up too too long sleep deprived I just sort of zoned out that was pretty funny a thank you for the
934. 2:40:46Andrea okay in the screenplay there's a significant intense confrontation scene image description the storyboard will
935. 2:40:52portray a tense moment between Derek and Jack on the fishing boat the image is 16 by9 the slug line exterior fishing boat
936. 2:41:00Puget Sound 1991 midaf afternoon let's proceed with creating this image boom all right so this is working I there you could I could dick
937. 2:41:09around with this and and you know it doesn't need to show me the the description of the image but that's kind
938. 2:41:15of cool it is a screenplay right so we want to see it I mean this is actually I mean if I were if I were an entrepreneurial
939. 2:41:29type I might like this is the kind of thing where you could send this GPT to a producer and and let them visualize the
940. 2:41:39movie on their own and ask questions about it this is a really unique way to put a screenplay in front of someone I think
941. 2:41:46look at that that's cool wow I don't know who that dude is and that guy is way too young but exterior fishing boat on put
942. 2:42:05it sound soured instead of Puget Sound it's the put it sour you know the classic put it sord of of Gig Harbor
943. 2:42:19Washington I'm going to try something here um try let me see try again with the image um the one character
944. 2:42:33is too young like watching three-year-olds do ballet chbt create damage I was working
945. 2:42:50on a film treatment last week and was wondering how to make it stand out thank you you're welcome lunatick
946. 2:42:57boom another satisfied customer right here in AI learning lab welcome um yeah there's there's something here there's
947. 2:43:05something that I mean the the assumption that you make is that you'd need to find a film producer that was interested in using
948. 2:43:16chat GPT which given the given the writers strike that just happened I'm sure they at least went and investigated what's all the hubub about
949. 2:43:25bub well that's better that's much better they're not really looking at each other this is kind of a kind of a
950. 2:43:39uh but you know I I mean that's pretty something I mean that does does look like the Puget Sound that's kind of what it looks like
951. 2:43:53H amazing oh amazing all right people we did some [ __ ] I hope that helped the person that couldn't get the
952. 2:44:03thing to read their PDF I don't it the my my only advice really is hey a assume I don't have the answers cuz I probably don't and and and
953. 2:44:20B just experiment experiment experiment experiment experiment experiment assume it's going to give you
954. 2:44:28wrong [ __ ] and be pleasantly surprised when it [Laughter] doesn't but don't don't just because it doesn't give you good [ __ ] don't stop
955. 2:44:40just go okay top of the hour present Turtles very nice very nice I appreciate the turtle presentation you all are very
956. 2:44:58cute a human being question how do we start a new Guild in the salon so so guilds have to go through
957. 2:45:07the organizers because one of the things we we're trying to make sure that we don't do is create overlapping guilds so
958. 2:45:13guilds are a pretty big deal um if you're interested in leading a guild let us know what it is and and
959. 2:45:19kind of what your vision for it is and then we have Guild meetings every is it every Tuesday I think it's every other
960. 2:45:29Tuesday yeah it's the off Tuesday from the salon um and then we'll meet you and we'll talk about it and if it makes
961. 2:45:37sense then we'll add it as a guild one thing that we are adding is clubs love clubs and guilds and a club is a much
962. 2:45:44more sort of lower barrier to entry there can be lots of clubs and the clubs will likely be private private little
963. 2:45:49clubs so if you want to start a little private club that's like a side you know private conversation you can do that
964. 2:45:55also in Mighty networks one of the things that might not be obvious you can create um multi-person chats so if you
965. 2:46:04go to chat you can add a person you can add multiple people to a chat so you could also create kind of a pseudo club
966. 2:46:11with multi-person chats about genetic Gene ology um possibly we're trying to keep the guilds right now pretty high level
967. 2:46:22um genetic genealogy could potentially live as a as a subset of the of the healthcare Guild we should talk about it
968. 2:46:29let me know drop me a DM on the uh on the uh on Mighty networks and we can we can uh we can chat about it happy to
969. 2:46:38chat about it that's super cool um that's part of the idea of the guilds is we want to create a place where people
970. 2:46:45um who are passionate about something have a a venue to sort of generate Community around that um so and it's I'm
971. 2:46:52I'm I'm digging the guilds the guilds are really cool and the Irregulars channel so the the people that show up
972. 2:46:59here regularly that's a club right it's a private club uh inside the salon all right let's see let's see what
973. 2:47:08else we got going here p uh let me put up some images images happy birthday chaa happy birthday chaa happy birthday
974. 2:47:25chaa all right so let me do a little um requesting so my request is if you like this content if you like this channel
975. 2:47:35couple of ways you can support it share this live I've got two hours left in my 24 hours of chaty Bing um I'd like to
976. 2:47:44end with a bang so so let's see if we can get some people in here tap the screen for likes um follow my Channel or
977. 2:47:50follow this Tik Tok whatever the [ __ ] they call them here is it a channel I don't know doesn't matter follow this follow the
978. 2:47:58me um because I'd like to grow my my follower base um if you want to support it you can subscribe to the lives um
979. 2:48:06which will get you access to the Irregulars club inside the salon um also I've got in the lower corner there I've
980. 2:48:15got a series which you can buy that you can support the Channel with that and the series is called underhyped and it's
981. 2:48:2217 or 18 videos about how chat GPT works why it's a big deal um things like that so check that out if you like that pick
982. 2:48:31it up all right that's that and then what else oh and then join the salon if or at least explore the salon if you
983. 2:48:38think it looks interesting so the URL for the salon where's my mousy mouse [Music] is that AIS salon.
984. 2:48:53mn. not do com that would be too easy doco all right and that's uh that'll take you there check out the about the salon page
985. 2:49:04cuz make sure that it sounds like something that you dig but if you're curious about this [ __ ] and trying to
986. 2:49:11figure it out and want to surround yourself with people that are also trying to figure it out that's what that group's all
987. 2:49:18about all right all right all right people all right H oh hey Marco [Music]
988. 2:49:34hi all right Kyle has lost his mind we don't know what KY is doing why does Cy talk like that how does Cy really talk does it talk like
989. 2:49:46that is he from some strange European country look at the glasses why does Cal talk like this I think we're going to turn this
990. 2:50:02into an NPC Channel sandwich sandwich sandwich sandwich
991. 2:50:17we're just gonna have battles with DAR every [Laughter] night H just created a GPT using an
992. 2:50:31income tax data set from data.gov which allows inquiry of I guess tax law wow that's super cool Dolly 3 coming to Microsoft
993. 2:50:43Paint I actually just saw so um I got access to something today that was very much a Microsoft Paint interface but because it's got
994. 2:50:56generative AI in it you pick these primitive shapes and you you draw the Primitive shapes over here and then it
995. 2:51:02renders it in real time using generative AI over here it's pretty cool my daughter loved the uh the balloon GPT
996. 2:51:10picture yeah that's a really good one isn't it I got the so so a lot of these images what I do is I'll pull them down
997. 2:51:17off of the mighty networks and toss them in my um and toss them in my uh my photo collection so I can so I can put them up
998. 2:51:25here um but let's go look at let's see if there's new activity Daisy Thomas what' you do oh that's that's a vomiting
999. 2:51:33a vomiting snowman Parts this is this does not look good quietly excited Snowman
1000. 2:51:50oh man this one's so cool the um make him hippier is such a cool idea let's see wait and the hippiest how do I get in here oh there we
1001. 2:52:06go make him hippier and even hippier listen I lived in Boulder for five years I have seen these
1002. 2:52:22people oh my God that's funny they're really good let's go to the Irregulars Channel let's see what's going on in the club
1003. 2:52:37what's going on in the club happy birthday to you happy birthday to you a oh look how sweet that is that's really
1004. 2:52:54sweet that makes me feel all warm and fuzzy happy first anniversary this is chat
1005. 2:53:08[Laughter] GPT the Snowman G to get you banned I know exactly right every time stick someone sticks the elephant face on me I
1006. 2:53:20think oh no I'm going [Laughter] banned oh man seen the r AI other side GitHub for
1007. 2:53:34self-operating computer no I haven't I was just mad living oh that was cool I was liking it I like it [Music]
1008. 2:53:50Antonio Kyle just want to say thanks thank you for sharing your ideas thoughts and perspectives you're
1009. 2:53:57welcome you're welcome this is you know speaking of perspectives like the the whole purpose of this 24-hour thing one
1010. 2:54:05it was just to kind of you know do it on a lark and have some fun but um I really do believe that this day in history November 30th marks an
1011. 2:54:19inflection point in in history that we're going to look back on this and go that was the moment where [ __ ]
1012. 2:54:27changed really enjoy your content and upbeat personality thank you so much um appreciate that um we're only a year into whatever this
1013. 2:54:40new era is and I don't have a [ __ ] clue what it's going to look like a year from now oh that was I think it was
1014. 2:54:54lunatick said maybe we should do like predictions for a year out I think a year
1015. 2:55:09out I think we're going to blow really quickly past multi modal I like I think multimodal is going to stay there so I
1016. 2:55:17think I think chat GPT will include um and tools like it will include music video animation 3D objects potentially 3D
1017. 2:55:32worlds as well as all the stuff that they've got now so I think I think multimodal is going to go true
1018. 2:55:40multimedia where you can just be talking and you want it to be a a movie or you want it to be a song or you want it to
1019. 2:55:47have a song as the background of a PowerPoint slide just I I I think I think that a thing that happens is that I think
1020. 2:56:00the generative AI film making in the next year is going to be really exciting because the film making tools
1021. 2:56:07are going to they're going to do they're going to do in 2024 what images did in 2023 so at the beginning of 2023 you know we
1022. 2:56:17still had hands that looked like this and you still had janky faces you know I think he's a model um and you know by the by now like
1023. 2:56:31the the images that are being produced are are pretty solid right they they don't have a lot of those major issues
1024. 2:56:38the video stuff right now is kind of like but now it moves it's like how you doing there so I think the video stuff's going
1025. 2:56:53to get much better and then and then what's going to happen is you're going to get two kinds of filmmakers you're going to get
1026. 2:57:00filmmakers who rather than resisting it are like [ __ ] it I'm all in let's go and then you're going to get people
1027. 2:57:07who are not filmmakers who are going to start making films and they're going to make films in ways that film makers will
1028. 2:57:15be like why would you do it that way and then they're going to come up with some remarkable [ __ ] so I think that's going to happen
1029. 2:57:25um David Shapiro says he thinks we'll have AGI within 12 months it's it's a video worth watching AGI in 12 months by David Shapiro on
1030. 2:57:36YouTube um I don't know if I agree with it I don't have enough information to agree with it but it seems like something just happened at
1031. 2:57:47open AI that was more significant than ego battles for why Sam Alman got fired um and and you know the the the prevailing rumor
1032. 2:58:01rumor is that they achieved AGI or something happened on the way to AGI artificial general intelligence that that shifted
1033. 2:58:11so I think that sometime in 24 we will see some major advancements let's not let's not call it AGI we will see some major
1034. 2:58:21advancements in autonomous agents that take action on our behalf and that sounds really boring but it's so right now you got to
1035. 2:58:35go to chat GPT and then you have to tell chat GPT what you want and then chat GPT will do it and it'll give you some [ __ ]
1036. 2:58:42back and then you got to do some other [ __ ] and then you got to copy and paste it and put it in the thing and right
1037. 2:58:48it's just all this you're at the helm you're sort of figuring out how to do this autonomous agents it's more like
1038. 2:59:00hey it' be swell if I had this solution and then the system will spawn a series of agents that all can take
1039. 2:59:13different actions and coordinate them for you okay this this agent's going to go off and search Google and it's going
1040. 2:59:20to come back with an answer okay this agent's going to take that answer and it's going to go off to perplexity and
1041. 2:59:26do something there then it's going to come back then it's going to go to Expedia and try to find this and then
1042. 2:59:30and all of this is going to be happening without you doing a thing and depending how good that thing
1043. 2:59:39gets there there's been versions of that auto GPT there's there's been versions of this stuff already but they've been
1044. 2:59:46janky pieces of [ __ ] that don't really work I think that we likely see something in 2024 where they they work and they work
1045. 2:59:57really well and if they work really well and can reason and can do math at 100% And basically do error checking um it's a big [ __ ]
1046. 3:00:11deal it's a big [ __ ] deal because you can literally just have that agent do someone's job a knowledge
1047. 3:00:20worker's job go research this thing like you know you you would say to a human being go re search this thing write up a
1048. 3:00:28report generate a PowerPoint presentation get me ready I got to have all of these things in place by Thursday for the big meeting with the
1049. 3:00:37boss and the person goes and spends 4 days doing that or you just say to the agent oh go do the same [ __ ] and then half hour
1050. 3:00:49later all the shit's done and it's done well it's done better than the human could have done it now will will we will we be there in
1051. 3:00:592024 I don't think so but we will likely start to see the hints of those kind of systems coming online is my guess um we could very well see
1052. 3:01:17robots humanoid robots doing [ __ ] connected to chat GPT you put those two things together then then then we start
1053. 3:01:26getting into that future it's like holy [ __ ] Hollywood was right here's some
1054. 3:01:42thoughts all right in business
1055. 3:01:5720124 I predict we'll see at least one major company publicly acknowledge that it's been dramatically impacted by AI
1056. 3:02:17we'll see at least one major company dramatically shift its focus to embrace AI or change its business model because of AI and then we'll
1057. 3:02:32[Music] see probably hundreds of companies in denial about what's happening in their sector as lots of small startups start
1058. 3:02:46challenging big companies the small startups will be AI native and they'll start coming in and taking down bigger
1059. 3:02:56companies I think that's very possible because there's so here here's okay so here's a promising here here's a promising
1060. 3:03:09possibility the downside of all this AI [ __ ] is like well they're going to fire everyone they're going to lay them all
1061. 3:03:14off all right so let's say they lay them all off let's say they lay off 20% some given industry 20% cut cuz we now use Ai
1062. 3:03:24and they're not needed they're redundant get rid of them those 20% of the people also have access to these
1063. 3:03:36tools and some of them are going to be pissed off in how that was handled and they're going to go you know
1064. 3:03:42what I know how this business works I know where its inefficiencies are and I know now how now know how to use Ai and
1065. 3:03:51I'm going to come back after that company I you there's going to be a fair amount of that where there's going to be
1066. 3:03:57burgeoning startups that are powered by the layoffs that are going to do some real [ __ ] damage open AI already patented
1067. 3:04:06chat GPT 6 and seven odd they've never done it that early oh interesting patented or trademark I think they
1068. 3:04:13trademarked it cuz they they trademarked five a few months back everybody in Egypt loving denial yeah going to be a lot of
1069. 3:04:30denial Kodak BL black block blackbuster Blockbuster yeah exactly hey boss I just invented the digital camera what do you
1070. 3:04:45think we're in the chemical business boy it's about silver Halli here at Kodak get out of my office turn on the fan I'm smoking a cigar okay
1071. 3:05:03boss there will be debiasing inclusivity skin companies will want to add on wait what debasing debiasing inclusivity skin
1072. 3:05:19cuno those words are English do they am I I might be Delirious there will be debiasing inclusivity skin that companies will want to add on
1073. 3:05:34I don't know what skin means I don't know what that meant either that was a typo or I'm not par something right cuz I'm Delirious I
1074. 3:05:45think the audio industry will announce a car that can perform hands-free actions and more interesting there'll be a lot less
1075. 3:05:53loneliness with Pike friends and companions I sure [ __ ] hope so Silver Fox you you mentioned to me about the
1076. 3:06:00the you know it having more compassion than you know than than some people in your life and it's like yeah I I I think
1077. 3:06:07you're right I think I think that becomes a big one innovators dilemma use cases everywhere everywhere everywhere everywhere look at [ __ ]
1078. 3:06:17Google Google right now they invented the Transformer they invented the thing that chat GPT is kicking them in The Shins
1079. 3:06:32every day with because chat GPT said [ __ ] it let's go and then Microsoft looked at it and said oh can't we use this to [ __ ] with
1080. 3:06:43Google I'm in here's $10 billion let's go have some fun Sam and Google can't they're they're they're broken right
1081. 3:06:58now they've got culture issues they've got business model issues they don't have Microsoft has software sales to
1082. 3:07:07fall back on and and cloud services to fall back on that are that are a significant portion of their revenue
1083. 3:07:14Microsoft does not have the burden of relying on Advertising to the extent that Google does what is scary as black hat hackers
1084. 3:07:25benefit from chat GPT yeah I I agree Inspire chaos I you know the black hat thing is terrifying because it it's not
1085. 3:07:33so much that the black hat listen black hat Master hackers they've been playing with AI for a lot longer than than chat GPT
1086. 3:07:44what chat jpt does is it makes mediocre hackers really powerful so what happens is you go from a handful of elite Black Hat
1087. 3:07:55hackers to like thousands of them just like anyone can create a GPT so anyone now can create sophisticated
1088. 3:08:04applications by just speaking them into existence well there's going to be the hacker version of that which is you can
1089. 3:08:11just say um would like to find any systems in Ohio that use this version of Windows that haven't applied this patch go get me credit
1090. 3:08:26cards shut down the power grid in too so that's scary that's scary they've already started using chat
1091. 3:08:40GPT vision for self-driving tests it's going to start to get nuts have you guys seen the the what they're doing with the
1092. 3:08:47GPT Vision with with um game um commentary video game commentary holy [ __ ] I'll show you that I I'll go find that on Bard is the worst
1093. 3:09:01AI I hope no one trusts anything that comes out of it well Gemini from Google sounds like it's supposed to be pretty
1094. 3:09:08good although they did delay it a quarter so maybe it wasn't as good as they said uh I created lifelines at GPT for
1095. 3:09:17seniors and the lonely and isolated oh that's super cool lunar stick it's great name too you should be listen here's a thing
1096. 3:09:27that you can do any any of us can do as long as you got access to GPT 4 start making gpts lunar stick's been
1097. 3:09:36taking old business ideas turning them into gpts as she's having ideas she's just turning them into gpts because when
1098. 3:09:43the GPT store comes out it's going to be just like the early days of the App Store anyone's going to be able to get
1099. 3:09:49in it's going to be really easy to do this and and some of those gpts are going to get very very popular and they're going to make a lot of
1100. 3:09:59money and and then competition's going to come in in two or three years and it's going to be almost impossible to
1101. 3:10:04rise above the noise the opportunity right now is huge so understand what these things are and how to build
1102. 3:10:10them output filter go hydrate and sleep no I still got 2 hours thank you though I got apple juice I got apple juice water and
1103. 3:10:24coffee but I appreciate your concern I'm good I'm good good good I will sleep when the uh when we hit the witching hour
1104. 3:10:41here oh let me go see the uh lies like wild Google is also taking losses because of Tik Tock yep Google search engine results have
1105. 3:10:55suffered too I feel Google search engine results are the victim of their Monopoly in my opinion because they had a monopoly
1106. 3:11:08everyone only had one SEO model to follow and that was Google's everyone else didn't matter and so these SEO agencies got
1107. 3:11:19masterful at ranking on Google and so what they did was they built all these SEO injected sales funnel
1108. 3:11:28sites and so Google right now no matter what you search for it's almost all shitty sales funnel sites it's just so
1109. 3:11:38bad and it's when you use chat GPT when you actually get an answer back like like that and then you go try to use
1110. 3:11:44Google it's like I I just it it feels [ __ ] barbaric at this point like why would I have to sift through 400 [ __ ] search
1111. 3:11:54results just to find just to find a site that actually has the information and then I have to go read the information
1112. 3:12:02to get to the piece that may or may not be in there so that's like a 45 minute torture Fest no I'm good could you create a step-by-step
1113. 3:12:16guide of how to build your own GPT of course I could do that I probably will do that um I'm not I probably could I got to get better
1114. 3:12:31at the API stuff the action stuff I think that's going to get easier and better so I'm kind of waiting on that just because I'm lazy um
1115. 3:12:49I'll think about it I'll think about it I don't know that I'm I am very reticent I mean I know I have a whole channel here where I just
1116. 3:12:58talk about this [ __ ] I consider what I do here most of it's experimentation and so for me putting together a workshop kind of presumes
1117. 3:13:09that I know something and and especially with gpts right now I don't feel like I know [ __ ] have chat GPT write you a
1118. 3:13:16step-by-step guide Kyle yeah I could do that I don't know I'll think about it I'll think about it I appreciate the I I
1119. 3:13:22appreciate the request I'll I'll think about it um it's so easy though that that's yeah exactly motherly advice I I agree
1120. 3:13:32dead internet Theory truncates most page results around 33rd page so no 1.6 billion results exactly it's crazy
1121. 3:13:46[Music] um can you make my chat GPT to use plugins I don't think chat I don't think gpts can use plugins but you can call
1122. 3:14:04apis I think the simplest way to do a lot of uh API call is just is to do a zappier integration and then just have
1123. 3:14:17your thing call zapier zapier calls um that that's the thing I I did it once I I got the zappier hooked up
1124. 3:14:26and I got it to make the call and then it failed on the zappier side and I just didn't troubleshoot it so I don't know I
1125. 3:14:33don't have a compelling enough use case right now to actually go through the pain in the ass of solving
1126. 3:14:38that I don't want to solve it just to solve it um let's see
1127. 3:14:54um why is it called an API call so an API stands for application programming interface so it's basically a set of
1128. 3:15:06instructions that you can either send data so so what the API is is the um the interface application programming
1129. 3:15:15interface where I can have my system talk to your system and the API can either be uh a a an API that takes data
1130. 3:15:25in or sends data out or both right and so let let's take um let's take Twitter data um Elon Musk actually shut shut
1131. 3:15:37that API down but what you could do historically is you could send an AP API call or request so basically your
1132. 3:15:46computer would call Twitter and say hey I have a request can you give me any tweets that are new in the past 5
1133. 3:15:55seconds and it would go sure here they are and it would send the you know whatever was in the API um Fields it would send that data
1134. 3:16:06over you know in this in this agreed upon format so you know I would have a key that would say okay this key gives
1135. 3:16:14you access to Twitter's data and maybe you pay for that maybe it's free but that means you know that's my key so
1136. 3:16:20Twitter can track hey the person with this API key made this request on this date and we sent them that data and so
1137. 3:16:27it's basically just computers calling one another either putting data or requesting data sending data back and
1138. 3:16:34forth so it's I think it's literally like a phone call like I don't know where the the Ed ology of the word call came from but
1139. 3:16:44it's a a call hey give me [ __ ] and then it does or he hey here's some [ __ ] that's a post um chat gp4 native API is pretty
1140. 3:16:56easy to use actually yep it's pretty good although the uh the some of the the uh the assistant stuff with threading
1141. 3:17:03and things like that is a bit more complicated but it looks really really powerful because it's a rest API and
1142. 3:17:09when you ask it for something it's called making a request there you go we are so awful to the elderly in
1143. 3:17:18this nation so bad that 20% you you spoke of may just come from commune and sit around creating cool images all day
1144. 3:17:26and and awesome maybe the world gets more beautiful who gives a [ __ ] I don't know the obsession we have with like you've
1145. 3:17:36got to put in the hours boy why why if if I have a tool that puts in the hours for me why do you need my hours so this is
1146. 3:17:50this is there's an Ethan mik article um I forget what it's called it's something like pushing the button and what he talks about is inside
1147. 3:18:02Microsoft Word inside every application starting in 20124 will be the button and the button is write it for me
1148. 3:18:12and the example that he gives is a student comes into him and says hey Dr mik I I would like um you to write me a
1149. 3:18:22letter of recommendation so I can get a job and Dr mullik probably does a bunch of these a year and they take him an
1150. 3:18:31hour or two to write each each one and then if he writes them good and the students's good the student gets the job
1151. 3:18:40yay if e Ean mullik has the button and instead of taking an hour or two to write the letter of recommendation he can press it and it
1152. 3:18:48writes it instantly and the student still gets the job is that letter of less value is the value in Ethan mullik taking the one or
1153. 3:19:01two hours or is the value in Ethan mik getting the kid a job and that's something I think we all have to deal with where's the
1154. 3:19:17value if I'm a painter and I love painting and I love taking 10 hours to paint and for me the value is in the act of
1155. 3:19:30painting then AI is not going to be of interest to me but if I'm an illustrator and I work commercially and
1156. 3:19:39I need to get images to clients that they pay me for and I don't really care like I don't I'm not personally attached I don't
1157. 3:19:50assign my own personal value to the act of illustrating and I I have a tool that can take me from three days to create an
1158. 3:19:59image to 3 minutes to create 10 images if I get really good at prompting does it [ __ ] matter if I took three days or 30
1159. 3:20:15seconds why does it matter well because that's we pay you by the hour well maybe you shouldn't maybe I'll start charging you
1160. 3:20:26per image maybe it's called value pricing maybe ad agencies are [ __ ] because of this their entire business model is based on time and
1161. 3:20:38materials and along comes a set of tools that obliterate the time component of that and probably some of the materials
1162. 3:20:48component of that quite honestly does it matter I don't know I mean I think every person's going to have to figure out their own
1163. 3:21:01boundaries where their value is but I think we're all going to be confronted with well wait a minute what is my value in this
1164. 3:21:09situation 190 here Kyle 190 what 190 oh oh 190 in the room oh hey everybody 190 people wow Silver Fox
1165. 3:21:23we're at 173 right now how how is everybody um good to see you my name's Kyle Shannon this is the chat gbt Aon uh I am in my 22nd hour 23rd
1166. 3:21:37hour um I have about an hour and a half left and I'm I'm fading I'm going to take a little walk here in a second um
1167. 3:21:44I'm going to Nitro boost the AI Salon uh don't Nitro boost it because that's the Discord and we're moving to Mighty
1168. 3:21:50networks but thank you but we don't need it AI doesn't scale I don't think AI is going to need to user 392 401
1169. 3:22:0178116 um I think a lot of uh the AI is going to go on device and um it'll be the the big the big models will still
1170. 3:22:13use a lot of the a lot of the gpus but a lot of this stuff is going to get optimized and live on device so I don't think it needs to
1171. 3:22:22scale I think we go edge I think we go edge device sooner than people think you've been awake for over 23
1172. 3:22:32hours I've been I've taken breaks so last night I started at 8 took a break started at midnight went to 3:30 started at
1173. 3:22:436 started at 9:00 started at uh 12:30 and then started tonight at 7 and I'm still [Laughter] going so yes if I see my little punch
1174. 3:23:01drunk it's cuz I'm a little punch drunk it's cuz I'm a little punch drunk um yeah so so here's the deal if you are
1175. 3:23:12trying to figure out this chat GPT stuff um I can answer your questions Kyle you're the man this is one hell of a
1176. 3:23:19marathon thank you I'm going to take a little break I'm just going to walk around um I know there's a bunch of
1177. 3:23:24people just got in here but um I just want to take a break so hang out for a bit I'll be back in like 3 minutes just
1178. 3:23:30want to go go pee get some water get some blood in my legs take a little stretch uh knew a couple of people who
1179. 3:23:38drew full disability from Ad Agency because lying made them upset oh interesting interesting H yeah I I uh I spent a lot of time in
1180. 3:23:50that industry so it happens all right I'll be back Talk Amongst yourselves be nice to each other don't be don't be jerks don't
1181. 3:24:03troll each other be supportive chiy
1182. 3:24:41also
1183. 3:25:10this
1184. 3:25:40e
1185. 3:26:10is
1186. 3:26:40e
1187. 3:27:10e
1188. 3:27:40help
1189. 3:28:10those
1190. 3:28:40hone
1191. 3:29:12all right all right all right all right we we still have we still have 140 people here y'all are weirdos uh Irregulars I like it um do me
1192. 3:29:25a favor share the live he'll share his shelves are more interesting than mine mine are full of programming print
1193. 3:29:34testing Lin oh pin testing and Linux books yeah yeah I've got there's some okay things up there some photography
1194. 3:29:46books it's there's some creative stuff the monopoly.com Edition did you see that can you see that oh you don't see [Music] that the do
1195. 3:30:05edition let's see the companies were Yahoo excite at home lyos eBay CBS MarketWatch monster.com CNET E online and I Village the women's
1196. 3:30:26Network crazy cza cza cza a CA time it's a c time all right
1197. 3:30:41um so what do we want to do here let's see um let me look at comments see what anyone sanding down here if you have
1198. 3:30:51questions about AI in any way I'm happy to answer that happy birthday to Chad GPT this is my I've got an hour and a
1199. 3:31:00half left from a 24hour chat chat GP Aon yay you did it yay I haven't done it yet um I really appreciate yall being here I
1200. 3:31:16can't believe there's so many people here it's crazy it's good it's fun um so if you have questions pop them in the
1201. 3:31:21comments below I will do my best to answer them and then we can go play with we can go play with chat gbt we can go
1202. 3:31:27do whatever you want to do I happy to show you whatever 12:43 on the East Coast night night and thanks for the virtual birthday cake that's
1203. 3:31:37nice uh very nice uh who uses w woo Commerce and any AI tools with it I do
1204. 3:31:55not is the Discord Channel now gone in place now that Mighty Network site so Corey the next Tuesday December 5th is when we
1205. 3:32:10offici switch over from Discord to Mighty networks the Discord is going to stay live we're going to Archive the The
1206. 3:32:19Guild channels in Discord so the guild channels are going to go away the content will still be archived so if
1207. 3:32:25people need to get it that for some reason it'll be there but they're going they're going to not be active we're
1208. 3:32:31going to deactivate the the general conversation channels like all the tool specific conversation channels all of
1209. 3:32:38that's migrating over to Mighty Networks what we're going to leave in the Discord is is basically some onboarding stuff
1210. 3:32:45maybe a general you know conversation Tab and then things like mid journey and any sort of if anybody wants to create
1211. 3:32:53Bots or things like that as part of the the guilds and mighty networks they can go play over in Discord so discord's
1212. 3:33:02going to stay there but it's going to be a secondary Community Mighty networks is going to be the primary for everything
1213. 3:33:08cuz it just has it just has a ton more features like it's got you know you can create courses in it and it's got calendaring
1214. 3:33:16in it and it's got a lot of the stuff Discord had um plus some stuff like threaded conversations feeds um we lose some things like the
1215. 3:33:27ability to use lots of different emojis and [ __ ] like that which people really enjoy so there's some stuff that sucks
1216. 3:33:32about Mighty networks but for the most part it's a lot lot better and it's also web based you don't have to go to
1217. 3:33:37Discord a lot of people don't know what the [ __ ] Discord is so yeah so it's going to be the primary home the primary home of the a
1218. 3:33:56solo go back home stretch Christopher Ramsey this is it my walls are all whiteboards for my
1219. 3:34:12brain outside my head I whiteboards are the only way that for they're the only way that I can process ideas is is whiteboards do you
1220. 3:34:22do this Silver Fox do you hoard whiteboard sessions I'm so bad with it where like I'll I'll if if I have a like a big
1221. 3:34:33idea I'll take a bunch of pictures of whatever the [ __ ] was there before and then and then I'll do my whiteboard
1222. 3:34:40session I'll get it all out I'm like okay that's good and it'll be like 80 or 90% of what I want it to be and I'm like
1223. 3:34:46okay I'll leave it there and then it's there for months and I can't erase it so I just draw I just write around it in
1224. 3:34:52all the white spaces around it it's really bad but here let me show you something and know this chat GPT is now
1225. 3:35:01multimodal which means it can make images but it can also see images so here we are in chat gp4 and here is uh this little paperclip
1226. 3:35:11and the paperclip I can upload files I can upload CSV files Word files uh things like that but I can also upload images so if I type in
1227. 3:35:27whiteboard there we go there's a whiteboard that I found on the internet I'm just going to say what is this and so it is it can see
1228. 3:35:44that appears to be a photograph of a whiteboard with handdrawn notes and diagrams that sketches appear to be
1229. 3:35:50brainstorming or planning stages of a user interface for an ordering system possibly for an airport context there's an ordering UI there's
1230. 3:36:01users there's use cases there's benefits there's main steps there's additional features and there's wireframe sketches the image creates a
1231. 3:36:09comprehensive View um let's give this app a name what should we call it and who's the target
1232. 3:36:27[Music] [Music] audience put a big note do not erase exactly to
1233. 3:36:46chat J seems to be a little slow tonight what's going on Sam oh that was interesting that it just hung like that
1234. 3:36:53and then vomited it all out considering the feature set Jet Set order or flyby oh I love flyby target audience business
1235. 3:37:04Travelers families time conscious okay I love
1236. 3:37:20flyby please make me a 9 by6 wide image of what a marketing poster in the airport might look like for this
1237. 3:37:29app period Please visualize the main order screen in the marketing poster [Music]
1238. 3:37:45yes and write around it in different colors exactly whiteboard management I test my crystal compendium
1239. 3:37:55GPT to upload a photo photo of a crystal to identify it and it did it properly oh that's amazing that's cool error creating the
1240. 3:38:05image it seems there was an error I will now attempt to create a marketing poster it appears there was an error try again or [Laughter]
1241. 3:38:21suan um I'm Still Loving chat gpd's logos wearing a party hat I know it's great isn't it I was really surprised
1242. 3:38:28today when stuff about a small company on chat GPT I was really surprised today when stuff about a small company on chat GPT
1243. 3:38:38I don't know what that means Winston um write me
1244. 3:38:53code for the front and back end of the
1245. 3:39:09app prediction prompts will be secret like secret recipes example an artist style will be a highly custom
1246. 3:39:27prompt it's possible Antonio but because these things are multimodal now and they can see images you can
1247. 3:39:37literally take an image that an artist has created and chat jpt will be able to describe it and prompt itself to create
1248. 3:39:44something very similar so I I think I I I agree I think that I think that how artists use these tools they're going to
1249. 3:39:53get really good at it and they're going to get known for it I think what ends up happening is it will be really easy for anyone to
1250. 3:40:01copy anyone else's style but what will likely happen is that we don't hire people because they're good at Art
1251. 3:40:08anymore we hire people because we like them and we trust them and they happen to be good at Art but I think that's that comes
1252. 3:40:20third like trust or trust like and they can do the work because we're moving into a world where doing the work is completely
1253. 3:40:32democratized so you won't hire someone because of the work anymore you'll hire them because of who they are and what
1254. 3:40:39your relationship with them is um anyway I don't know if you saw what just happened here in the background but um here's code for the back
1255. 3:40:48end JavaScript back end no JS with Express for the back end and here's the front end uh for react react native for for
1256. 3:41:01mobile we uploaded a whiteboard and it created that code now does that work I don't know maybe but it did it and it
1257. 3:41:12like it understands what it needed to do based on a drawing um let me see if I can get it to do the marketing thing try again on the
1258. 3:41:22image or eay E IM Tobias I was super impressed myself very proud of it turns out I write about AI more than comedy now wow interesting
1259. 3:41:37interesting well I hope you're writing about AI in a comedic way don't lose the don't lose the thread error creating the
1260. 3:41:45image I'm tired of chat GPT sucking it's your birthday time gbt you should suck less one hour you
1261. 3:42:02make you make oh yeah all right [Music] traditional artists can now use AI to
1262. 3:42:16generate description artworks for marketing purposes yep indeed it indeed gpt's been stubborn all day but I didn't hit my cap I wonder
1263. 3:42:28if I've hit my cap I don't think so I haven't done that many images Suzanne I'm back I agree but my iconic pink pop
1264. 3:42:36art is a secret recipe at least for now oh okay okay wait okay you think this is an iconic recipe that's a secret we're
1265. 3:42:47going to we're going to we're going to we're going to blow this up let's see um how can I where am I where's my Mouse Hello Mouse where are
1266. 3:42:59you did my mouse die no okay we just had window redraw issues this is probably Microsoft Edge um issues whatever okay so we're GNA
1267. 3:43:15export this to the desktop fine now we're going to go to chat JP T let's go to chat J T and let's see how
1268. 3:43:24quickly let's see let's see if we can replicate this image without knowing prompting so I'm just going to
1269. 3:43:35go desktop grab that grab that there's the image now I'm going to go um please describe oh I can't make images here this in as
1270. 3:43:51much detail as possible so there's our image Suzanne's lovely popart uh birthday C birth a CAC it's a
1271. 3:44:05delicious CAC it's a delicious CAC it's birth a CAC with a lovely little candle on it she's Carol Marill M she's lovely
1272. 3:44:17okay the image features a comic book style illustration with vibrant colors the woman is dressed in a pink
1273. 3:44:23leather jacket behind it the woman there's a speech bubble and a jagged outline okay the overall aesthetic is
1274. 3:44:31dynamic and cheerful with a burst of yellow lines waiting for please recreate this image okay it's a
1275. 3:44:44square now come on GPT create the actual image don't just error out on me come on you can do it you can do it Chad
1276. 3:44:55jpt we know you're only one and we know you're janky you can make money with J boom not bad it still got the got the pop art in
1277. 3:45:10the corner so it got the half tones out here it didn't get it on her I'll get back to it for sure I can't stop
1278. 3:45:20myself wait oh I'm not going to stop comedy once I published my books and the last of my backing oh and last of my
1279. 3:45:28backlog AI just zoomed in oh yeah yeah interesting yeah that's great that's super cool uh um I had a meeting with my
1280. 3:45:41department cam katkin I had a meeting with my department chair today showed her chat GPT Firefly 2 co-pilot she was blown
1281. 3:45:52away that's awesome with your department chair that's really awesome it's crazy that's close yeah it's pretty close right um yeah that's
1282. 3:46:06crazy so so this is this is where and and you're going to be able to do this with you can now screenshot a website and it will just code the
1283. 3:46:18website for you you can just screenshot it and it just starts coding the website and it even generates the images in in
1284. 3:46:26all the appropriate places within within the the website design so again again I go back to I I go back to um here let's pop that
1285. 3:46:41up big is that big wait
1286. 3:47:05download okay there's that um that's super cool so one of the things so cam Katen from Cleveland um is is is doing one of those things
1287. 3:47:20that I'm a huge fan of in here which is as you get curious about Ai and as you get good at it and even if you're not
1288. 3:47:29good at it cuz that's all relative just as you learn anything talk about it share it and invariably what's happening
1289. 3:47:37is people in this channel are being asked to hey could you present to the big honcho or to this team or hey you
1290. 3:47:44know I'm trying to figure this out could you help me do that and they're they're establishing themselves as thought
1291. 3:47:50leaders and it's it's really exciting the department chair asked me to present in January to the whole
1292. 3:47:58department and February at the division at the division meeting excited holy [ __ ] that's so awesome there you go
1293. 3:48:10this is not you can make money with that's not it it's not about the tool it's about solving problems and it's about relationships and it's
1294. 3:48:26about being curious and saying hey hey people this thing's happening you should pay attention to it ah we don't want that
1295. 3:48:40okay next week hey want to see something cool I did this thing's happening you should pay attention to it
1296. 3:48:49okay A month later hey I did this other cool thing you want to see it yeah okay I'll see it it's important we should talk about it I don't
1297. 3:49:00really sure you want to see it okay here oh my God that's generally the progression of of these conversations oh my God you
1298. 3:49:14know you should show that to I think the the president would really like this you want to show this to the president okay
1299. 3:49:20I've never met the president but sure yeah it's really exciting that's super cool crazy I'm curious about seeing the
1300. 3:49:31back of my eyelids right now hi Kyle aren't you tired no rustic empath I'm [ __ ] exhausted I'm not tired I passed tired I was tired
1301. 3:49:43at 9:00 this morning no no we we we have far exceeded tires no no no what's up what's up H
1302. 3:49:57there just a ton of people in here this is really cool all right I love that everyday AI completely obliterates what I think its
1303. 3:50:06limitations oh I think that every day AI completely o obliterates what I think its limitations are laughing my ass off
1304. 3:50:14I know it's it it's totally true Antonio is is um it's it's like about the time I think I have my head around it and and like
1305. 3:50:25it'll happen in here where I'm like oh yeah I I know how to do this and someone will go hey Kyle can it do this I'm like
1306. 3:50:31I don't know I try it it's like yep it can do it oh [ __ ] it can do it really good
1307. 3:50:47I mean what what's what's what's Wild is is the human piece of this for me CAM you back on tomorrow having a day off
1308. 3:51:00[ __ ] no I'll be back tomorrow 8 o00 um the the human peace to me is a really interesting one because it's
1309. 3:51:17like where do each of us derive our our own personal sense of self-worth and value and if if your own personal sense
1310. 3:51:29of self-worth and value is in this is the work I do and I'm valuable because I do this work and then that work is democratized and
1311. 3:51:43commoditized and the value of just the work itself drops essentially to zero what's that going to do to people's
1312. 3:51:53selfworth it's it's really [ __ ] up so so and and you know that also assumes people are aware of what their
1313. 3:52:04selfworth is right so they might not even be aware that how I value myself as this work and that work is now
1314. 3:52:11democratized and commoditized and therefore I'm valueless like people might not even be aware of why they're so
1315. 3:52:18devastated that feels really [ __ ] up to me but the corollary to it is I might think my worth is in doing this
1316. 3:52:32work and then when it get gets commoditized out I realize oh I actually hated that word work and now that that's
1317. 3:52:41been taken away the thing I really want to do is this thing over here that was not even on my radar but now because of
1318. 3:52:49AI tools I can level up and do this kind of work at a higher level and I might have a whole new sense
1319. 3:52:57of selfworth so I think there's going to be big swaths of people that are just absolutely [ __ ] crushed me a bunch in the middle that
1320. 3:53:07are neutral like ah [ __ ] I got to deal with it and then there's going to be probably you know sort of an equally
1321. 3:53:14sized chunk that are like well yeah you know what I would have kept doing that for 30 years but you know what [ __ ] it I
1322. 3:53:22actually hated it you know what I just like I I like this I didn't even know I like that Kyle just explained my life in politics there you
1323. 3:53:36go yeah that's insane to
1324. 3:53:51me it's like we're entering well so okay you know what it is no Kyle please tell us welcome to Chad [Laughter]
1325. 3:54:04add Co forced us into a social experiment and we had to learn how to not be social animals even if we were the introverts
1326. 3:54:19Co was like [ __ ] Home Run for the introverts us extroverts were [ __ ] but it was a forced social experiment
1327. 3:54:35globally and and what I learned from that is like well we had all these rules and the rules were you go to
1328. 3:54:45work and the rules were companies exist because people show up at the same geographic location and there were
1329. 3:54:52exceptions to it but for the most part that was the rule and then for two years we didn't and what hit me was like oh we
1330. 3:55:02just made that [ __ ] up we just accepted that because that's where we had naturally evolved as a societ soety as a
1331. 3:55:09global society that work kind of look like people show up at this building and then we got forced into this social experiment that said well
1332. 3:55:20what if not that oh okay I won't leave the house for two years let's see how that goes companies survived companies are now struggling to
1333. 3:55:35get people to come back to that old model because we all experienc that oh this other model also works and so now there's all these
1334. 3:55:43Hedges where company's like well come in for two days if you come in for two come in for two and there's an optional third
1335. 3:55:49one and we're going to have cupcakes on the third day you should really come in for the cupcakes I make them myself
1336. 3:55:55they're going to be tasty it's I make a chocolate buttercream like you just have do you like this do you like the the
1337. 3:56:02little star sprinkles or do you like the the little sticks you like oh you like the sticks interesting say I like the
1338. 3:56:09crunch of the Stars that's so funny anyway come in for the third day you'll get the cupcakes so that's happening right now in business and
1339. 3:56:20now there is a grander social experiment being slapped on top of the covid one what if how we do what we
1340. 3:56:37do complet completely changes overnight what if to do the same amount of work that you were hired to do you actually only have to put in 10
1341. 3:56:52hours what do we do with those 30 hours do we just sneak off to the side like we did with Co and go ah [ __ ] it
1342. 3:57:01I'm going to take a two-hour lunch no one can see or does it actually change the nature of what companies are what work is what jobs
1343. 3:57:20are come on people I need answers anyone buer I thrive staying home I did okay staying home I just hated it I just
1344. 3:57:36hated it hated it hated it I'm just I I'm so bad at I'm like oh there's a Gilligan's Island Marathon I'm [Laughter]
1345. 3:57:51in I need I need the I think it's the add thing I like I just need the structure of get up go to the [ __ ]
1346. 3:57:59office do the work what do you want CH you want to go out you going to go out you want to go see Mama
1347. 3:58:17oh just he just he heard you there he's like where's my cheese champy come on come with me she's got nothing for
1348. 3:58:29you she's got get up there you want get up there you the grand [ __ ] experiment oh thank you so much Winston that is beautiful I
1349. 3:58:45love that thank you so much how many do a job they hate for the insurance they need because their job makes them sick yeah it this this
1350. 3:59:01is I think PE people kind of comment to me I get I get a lot of backhanded compliments about
1351. 3:59:15um I I was on a panel I was on a panel about six months ago and someone who'd been in AI for a decade after after I spoke said let me
1352. 3:59:28let me interrupt this adorable optimism with a reality check and so I was I was referred to as as an adorable
1353. 3:59:36Optimist which is a which which is a a a a pretty nice troll right it's pretty solid troll it basically was like listen you naive piece of
1354. 3:59:47[ __ ] um you know AI isn't perfect and it's not secure and it's not this and it's not that when I play this [ __ ] out in the
1355. 4:00:05future there is a very real possibility that a lot of us are doing jobs that [ __ ] suck because the system is built on
1356. 4:00:15these complex Byzantine [ __ ] nightmarish hell holes of process that are only there because the current technology we have can't simplify
1357. 4:00:29it and so we just keep adding to the [ __ ] insanity and we have jobs there are peoples whose job it is to take
1358. 4:00:41faxed forms that were filled out for insurance and read them and type that [ __ ] in why cuz we didn't have the stuff to not
1359. 4:00:55do that and because that whole industry sector Health Care decided that faxes were still the technology dour until 2020
1360. 4:01:12W what if all of those jobs went away is that a bad thing well those people lose all their okay yeah that's
1361. 4:01:26bad but now those people have access to these tools and maybe there are new jobs that can be created because instead of
1362. 4:01:36spending $300 million a year on document processing we spend 50 million a year on document processing what do we do with that extra
1363. 4:01:51Revenue well if you're a PE firm you pocket it but I think I think there will be companies that do the right thing and
1364. 4:02:03then there'll be a lot of people that reinvent themselves and reinvent new ways of working so I might be an adorable Optimist but it is
1365. 4:02:16possible that when people are less stressed out because they're doing less shitty jobs they find new and exciting and
1366. 4:02:26valuable ways to to contribute to society what if that Antonio parents spending more time with their children what if
1367. 4:02:42that what if Daddy only had to travel two days a week instead of six what if that
1368. 4:02:58so stay at home would have been much funner if we had chat GPT no [ __ ] right more time doing what you love how many people
1369. 4:03:09don't even know what they love because there's so much stress in their life because of the nature of the job they're
1370. 4:03:17in that that if there were a tool that took away 30 of those hours they still had the job they still had to do [ __ ]
1371. 4:03:26for 10 of the hours but 30 hours were freed up where they weren't just completely overstressed and behind the eightball all the time what if
1372. 4:03:34technology came along where they could actually C up catch up and get ahead of the curve and start thinking about what they really wanted to
1373. 4:03:44do what if new jobs created yeah I just I like I just feel like I'm so [ __ ] tired of hearing it's going
1374. 4:03:54to take jobs everyone's going to be devastated the robots are going to kill us maybe or the shitty jobs get
1375. 4:04:10automated well what's left I don't know make something amazing go invent a new thing well I don't know anything about Material
1376. 4:04:21Science that no longer matters look look look what happened yesterday where is
1377. 4:04:36it [Music] where did it go this one yesterday Google Deep
1378. 4:04:53Mind released a paper published in nature today we share the discovery of 2.2 million new crystals equivalent to 800 years worth of
1379. 4:05:06knowledge project called gnome graph networks for material Exploration with gnome we've multiplied the number of technologically
1380. 4:05:16viable materials known to humanity of its 2.2 million pred predictions 380,000 are the most stable 380,000 new materials were just
1381. 4:05:31released into the world Edgar Kyle you should make these 2 4 Hour podcast more often good Lord good Lord sir had that cono today AI won't
1382. 4:05:44kill our creativity it's unlocking it [ __ ] yes it's unlocking it the person who was crushed by that shitty
1383. 4:05:57job that now only has to spend 10 hours thinking about that shitty job and let's AI do the rest takes the 30 hours of
1384. 4:06:03that extra bandwidth and says you know I've always wanted to make the next Teflon but I don't know anything about Material
1385. 4:06:10Science but hey I just did a Google Search and found this database with these 380,000 things and I learned
1386. 4:06:17enough about them over the weekend using chat GPT that I now understand the properties of one of those things and I I have an idea for a new
1387. 4:06:26pan why not why the [ __ ] not why 380,000 new stable crystals this is going to be batteries that
1388. 4:06:40charge in 2 minutes and last for 500 miles should set up a political AI party yeah that's an interesting idea look at this
1389. 4:06:55graph this blue dot is the 20,000 materials that humans discovered this medium Blue Dot is the 48,000 so in addition
1390. 4:07:0828,000 that computers discovered with humans with traditional means traditional Computing means so humans good computers better AI [ __ ]
1391. 4:07:25dramatic we went from a grand total of 48,000 known materials stable crystals to 421,000 overnight
1392. 4:07:43overnight about 20,000 of the crystals experimentally identified in the ICD database are computationally stable computationally
1393. 4:07:55or computational approaches drawing from the materials project open Quantum materials database boosted the number to
1394. 4:08:0348,000 The Gnome project expands the gome number of stable materials known to humanity to
1395. 4:08:21421,000 amazing I have always wondered and still believed this could be another Mega Renaissance yeah I'm calling this the
1396. 4:08:32great Renaissance do you notice the the two the middle letters in the word renaissance for my for my life so I got a degree in
1397. 4:08:42acting and I always wondered why in the [ __ ] is there an i in the middle of the word renaissance it's because we were waiting for
1398. 4:08:51AI I think we are entering a great Renaissance we're going to see scientific break well going to see scientific breakthroughs th this is not
1399. 4:08:59a scientific breakthrough like this is this is like watershed to go from 48,000 to 420,000 known materials stable
1400. 4:09:18materials low temperature superconductor you know the parallel universe theory says that in some Universe Kyle is dueling
1401. 4:09:29with wait live will end in one minute verify to continue that was weird parallel universe says that in
1402. 4:09:41some Universe Kyle is dueling with DAR in an AI gen Matrix oh man I have always wondered and still believe this could be another Mega
1403. 4:09:57Renaissance we can't even with the parties Beyond Democrat and Republican that already exist just give it just give the
1404. 4:10:05government to the [ __ ] AI seriously it how how much worse could it be GPT is not even good yet exactly wait
1405. 4:10:15till Quantum processors are viable or just wait until qar comes out if I I I assume folks on here have heard of qar
1406. 4:10:24it was the thing that that is rumored to have gotten Sam Alman fired this is just first geni exactly we we have not even begun we
1407. 4:10:34have not even begun not even begun we're just just getting started I like this
1408. 4:10:49one the one-year anniversary edition what time is it what time is it people 11:29 30
1409. 4:11:05minutes I'm not crapping out so you don't crap out all right let's send this with a bang do me a favor share the live
1410. 4:11:12hit the like button follow my channel just do the [ __ ] let's follow this up with a bang I oh thank you very much Winston I appreciate
1411. 4:11:22that whoever said I should do these 24hour things more often it's like I I don't think I could survive it it's been fun
1412. 4:11:34though I here the the nice thing about this 24-hour thing I have no [ __ ] idea what I've said just like I'm
1413. 4:11:46just genome and Alpha fold is changing so much in science and medicine I know I know and again like we haven't even this is
1414. 4:11:57so so qar if you don't know what qar is qar is purportedly the break through that happened at open aai this past week or a
1415. 4:12:12couple of weeks ago three weeks ago it's a combination of a thing called AAR which is a game navigation training
1416. 4:12:21system and Q learning so they basically took these two models and kind of did some sort of munged version of them and apparently it
1417. 4:12:34allowed I assume chat GPT 4.5 or five to execute low-level mathematical problems at 100% accuracy 100% like nothing wrong no no zero
1418. 4:12:52mistakes and they've never gotten close to that before and so if you extrapolate that out it gets it gets really crazy really fast
1419. 4:13:02so um so if you look at the the material science that deep mind just put out and the and the alpha fold stuff that they
1420. 4:13:12put out that was on the old AI so imagine now if you had a tool that could do the computational part of whatever you do with 380,000 stable
1421. 4:13:25crystals whatever mathematics you need to do to figure out how to manufacture them and apply them and what they could
1422. 4:13:32be used for if you had systems that had essentially you you know low to zero error rates in doing those calculations it it's hard to even
1423. 4:13:46comprehend how quickly Innovations are going to start falling over and it's funny Sam Alman said this I watched a video of Sam
1424. 4:13:57mman like two years ago and he said um he said in a in a in a relatively short window like within 5 years we're going
1425. 4:14:12to see a single year where we experience 500 years of scientific advancement and and what is this article say the first thing this article
1426. 4:14:28says today in a paper published in nature we share the discovery of 2.2 million new crystals equivalent to nearly 800 years worth of
1427. 4:14:42knowledge and that's in one scientific area Material Science 800 years worth of knowledge they were working on Alpha
1428. 4:14:56fold like two years ago like this hasn't been that long [ __ ] that's AI for
1429. 4:15:12president gen AI will evolve using AI not in school yep give him control I have the book that we created I also made a US Government
1430. 4:15:23reset guide GPT by the way awesome awesome a human being fantastic the US government needs a good [ __ ] reset
1431. 4:15:32doesn't it God just just awful what's qar so so I just described Q qar so it's basically a um I'll I'll show you here
1432. 4:15:43if you want to if you want to geek out a bit one of the guys that I recommend people follow is a guy named David Shapiro
1433. 4:16:03um so Q Stars bigger than you think why oh these aren't all his are they hang
1434. 4:16:28on will you really have enough money [Music] okay so go to David Shapiro's his Dave
1435. 4:16:42shap D Ave shap on on YouTube and the fifth and sixth videos the sixth video is what is qar and
1436. 4:16:58the fifth video is qar is the biggest thing since word to V which is um an algorithm that led to Transformers
1437. 4:17:12and Transformers are the T and GPT so he's saying that qar is the biggest thing essentially since the thing that
1438. 4:17:19made all this [ __ ] possible and then his last video talks about debunking a part of this this bigger than you think video
1439. 4:17:28where where it it at the end of this bigger than you think video he he it it basically says there there was a 4chan
1440. 4:17:36paper that was put out that basically said that all of encryption is dead CU cuz this thing solved it all um and it
1441. 4:17:44looks like it looks like he might have debunked that up here but anyway this video about what is qar and why qar is a
1442. 4:17:50big deal these two videos are absolutely worth watching if you want to geek out on it I played with firefly 2 today
1443. 4:17:56pretty slick oh that's cool you know I haven't played with it cam katkin I need to do that Q we've only be just begun by the
1444. 4:18:04Carpenters We've Only Just Be [Music] [Laughter] [Music] I don't know what it is about those I
1445. 4:18:20guess it's the Carpenters I just the s that 70s music that music to me just always makes me go like oh that's life
1446. 4:18:29is just awful it's so sad just like everything's sad Kyle did you apply for any of the AI fellowships while it was open uh I
1447. 4:18:40didn't know there were AI fellowships I don't I know not have what you speak a Lun I did not know of the a
1448. 4:18:48fellowships um I I'm I'm running my company so I don't I don't have I don't have the I don't I don't have the space
1449. 4:18:57to do 24-hour Tik Tock lives you know what I'm saying Quantum artificial reasoning quar uh uh [Music]
1450. 4:19:11what issue of nature I assume it's the one that just came [Music] out paper published in
1451. 4:19:26nature nature 2023 379 it's called scaling deep learning for materials Discovery abstract novel function materials enable fundamental
1452. 4:19:39breakthrough across technological applications from clean energy to information processing from microchips to batteries and
1453. 4:19:50photovoltaics discovery of inorganic crystals have been bottlenecked by expensive trial and error approaches
1454. 4:19:56concurrently deep learning models for language vision and biology have showcased emergent predictive capabilities with increasing data and
1455. 4:20:05computation here we show [Laughter] here we show the graph networks the graph networks trained at scale can reach unprecedented levels of
1456. 4:20:18generalization so basically you can just learn any Crystal you don't have to do this really specialized [ __ ] per Crystal
1457. 4:20:24it's generalized improving the efficiency of materials Discovery by an order of magnitude building on the 48,000 stable
1458. 4:20:33crystals identified in continuing studies improved efficiency enables the discovery of 2.2 million structures
1459. 4:20:42below the current convex Hall many escaped previous human chemical intuition our work represents order of magnitude order of magnitude
1460. 4:20:54expansion in stable materials known to humanity our work represents order of magnitude expansion in stable materials
1461. 4:21:04known to humanity this we're going to start seeing this over and over and over ai. goov took took something Karen Carpenter has such a
1462. 4:21:15beautiful voice yeah exactly oh except Steely Dan a lot of the Steely Dan stuff is really sad but they have some they
1463. 4:21:21have some kick-ass songs took applications for two weeks in early November for public to train oh I saw that I it that seemed like something
1464. 4:21:29that wouldn't be good for my ADD brain they actually expect you to follow through on that [ __ ] we generally see
1465. 4:21:36technology after the military have the military got better AI I don't know about that so I I presented on an AI
1466. 4:21:43panel uh at a military Conference in June I think with Richard Boyd's he's done a lot of work in that and Richard's been
1467. 4:21:54Richard's been in AI for a decade um it is it is absolutely clear that the military has lots and lots of
1468. 4:22:03AI um but this generative AI stuff is different and it's it's every every scientist that I've talked to is like
1469. 4:22:11yeah the AI stuff that we're working on like even Richard was like the the AI systems that I've been building for the
1470. 4:22:17past decade were like powerful but like they're they are just in a completely different League than this generative AI
1471. 4:22:27stuff it's it's this idea of it being generalized and good at everything and good at you you know um stuff it wasn't trained on that that is
1472. 4:22:39uh I I I I think no I I think that that the military is probably trying to figure it out but I don't think they have a massive head
1473. 4:22:47start I mean I really think that chat GPT marks a a a different kind of Engagement with machine learning I fixed my GPT thank you oh
1474. 4:23:02good hit my prompt limit in chatty in record time now on Ice until 1:1 a.m. that's awesome you were you were the one that the you
1475. 4:23:13couldn't get your PDF to to recognize it that's great all right awesome that's really cool they'll let people know in
1476. 4:23:20the spring who was accepted Karen had a beautiful voice she died long ago Donald Fagan chatot bot would be pretty great got it
1477. 4:23:35thanks sad music from the lead exposure solid Quantum Computing do you run local open- Source AI I'm starting
1478. 4:23:47to train on our company's code base I don't um well we've got so we're running a local instance of a new whisper um
1479. 4:23:59model that's super fast that's that's really cool this new whisper model is like you upload like a 5 minute video and it's
1480. 4:24:08like 10 seconds later you got a full transcription of it it's insane I think I got access to this new
1481. 4:24:14tool we just built a new um Omni Channel content engine where you upload a video this is it's such a cool thing you
1482. 4:24:23upload a video it transcribes it it chapter IES it writes a title writes a description and then there's a whole
1483. 4:24:30bunch of buttons for different social media content and you just push them and it goes [Music] it just drops your content out and it's
1484. 4:24:39really good because chat gp4 is it's just it's really good it writes it writes a blog post about a v you upload a video it
1485. 4:24:51transcribes it chapter IES it and then writes a blog post as if the person in the in the video was interviewed by a reporter it's that
1486. 4:25:01good it's like I I read the thing thinking like this is kind of like if I took a lot of time to write this this is kind of how I would write
1487. 4:25:10it so but yeah so there's there's the new Ethan mullik post there's a new let me go to LinkedIn oh I got to log in dang it oh
1488. 4:25:25this is going to make me hang on I'm GNA freeze for a second how do we get our hands on it how do you get your hands on what my Omni
1489. 4:25:37Channel content engine it's it's not out yet I mean I'm I'm just going to get the first build tomorrow uh where resend why did I not get my
1490. 4:25:52thing hello oh crap whatever um go look go on LinkedIn and look up Ethan mullik maybe I think he put it on X as well hang on
1491. 4:26:06on maybe I'm signed into X where's X where's X where's that where's that silly goose e Elon Musk where is that silly goose
1492. 4:26:21x.com sign in with Google ah holy [ __ ]
1493. 4:26:38oh wait wait Rowan Chang it's official Sam Alman just confirmed new qar model breakthrough was a leak plus huge developments and AI from
1494. 4:26:54Elon Musk Microsoft the UK okay wait in an exclusive exclusive interview with the Verge oh this came out two days ago
1495. 4:27:01I didn't read it though Sam mman confirmed qar was a leak was leaked meaning it's real Sam Quotes no particular comment on that unfortunate
1496. 4:27:13leak wow huh well that's an appropriate way to sort of wind up the 24 hours isn't it Elon Musk predicts AGI within the
1497. 4:27:30next three years that's also the same interview where he told Bob Iger to go [ __ ] himself so unbelievable ideogram launched image
1498. 4:27:48prompting what's that mean within the platform you can now upload an image and control the output using visual input in addition to text
1499. 4:28:01what sorry sorry people here I'm discovering things as you are
1500. 4:28:18from ideogram we just launched image prompting oh image prompting okay you can now upload an image and control the
1501. 4:28:28output using visual input in addition to text this is available to all of our plus subscribers am I a plus subscriber
1502. 4:28:37in ideogram I don't think I am um I'm telling you people 2021 is going to be absolutely
1503. 4:28:51[Music] crazy just the speed of this [ __ ] sign up with Google I don't think I'm a plus user here
1504. 4:29:06[Music] plus no all right I got nothing there but wild the Sam Alman thing that's a big
1505. 4:29:23deal that's a big deal all right let me find uh let me find yield Ethan mullik um Ethan no where
1506. 4:29:42is I'll be with you people just hang on hold on to your horses calm down um Ethan
1507. 4:30:04M eaic
1508. 4:30:19Malik
1509. 4:30:34um hm he put out a piece on so there's a new reposted for
1510. 4:31:04clarity Google isn't Ethan mullik Google is in a position to Leap Frog to become a real power in the consumer AI World assuming
1511. 4:31:18they can get out of their own way and ship Gemini pieces are there integration with Google services personalization friendly
1512. 4:31:28UI but today hallucinations mistakes and lack of smarts interesting oh here we go okay so where can I share this um I I think I
1513. 4:31:40might have shared this on the salon but I'll do it let me copy the link so so Ethan mullik the Tweet is the
1514. 4:31:46ability to download a decent fully functional large language model with vision in one click that you can easily
1515. 4:31:54run on your own computer or carry on a USB drive feels like feels cyberpunk as hell um so I'm going to go over to the AI and I'm going to go
1516. 4:32:11to news and let me just make sure that I didn't post this already I may have oh Ethan mik shared this today
1517. 4:32:24absolutely wild and then I didn't post in I didn't paste in the the URL so I'll do that now there we go all right so if you go to the AI
1518. 4:32:36Salon in the news in the news channel um the third post down is this Ethan mullik thing about this there's
1519. 4:32:47this new um it's based on llama 2 so the meta llama 2 model it's a small installable that you can install on like
1520. 4:32:56an M1 Mac like this MacBook Pro is an M1 and it will it will run locally meaning you don't have to go out to open AI it
1521. 4:33:04just runs it locally and it's multimodal it can it can um analyze images all right if we could go back to
1522. 4:33:112021 knowing what we know now would we [ __ ] no are you kidding me I wouldn't the the th this technology
1523. 4:33:26once once you experience it like it I honest to God like Google searches feel barbaric like here hearing my hearing my
1524. 4:33:36developers talk about how they're thinking about developing code knowing what's possible today and knowing what's coming it just I'm like
1525. 4:33:46why are we talking about this but like right now we have to because we're in this transition period where the tools aren't good enough to
1526. 4:33:55just F fully replace anything he doesn't know who doesn't know what don't I know have you looked at party rock by AWS I
1527. 4:34:08heard about it I didn't look at it Kyle you said just imagine what 2021 is going to be like Oh I'm [ __ ] exhausted I meant
1528. 4:34:252024 don't listen to what I say people listen to what I mean that's one of the that's one of the tenants of dealing with neuros spicy a
1529. 4:34:37adhders you've got to you've got to read our minds we expect it thank you Dar fantastic F go get some sleep no I
1530. 4:34:51have I have wait how much time do I have hang on where's my did my watch die I think my watch died it is 11:53 I got 7 Minutes
1531. 4:35:03people you can make money every person in Tech or creative job is effed we're going to be laborers for the
1532. 4:35:100.1% at this point all listen I think it's easy to have that fatalistic point of view I think there is a nonzero possibility that it's not
1533. 4:35:25that dramatic I I believe in the uh I believe in in the power that's about to be
1534. 4:35:40Unleashed by leveling up everyone's capability I'm I'm seeing it more and more Sam mman talks about this a lot and
1535. 4:35:51and when he first when I first started hearing him talk about it it sounded like weird he drank the Kool-Aid kind of [ __ ]
1536. 4:36:00and but but now I'm starting to like like as I'm starting to experience like even this Community where I'm watching
1537. 4:36:06I'm watching in real time people in this community leveling up their skills and leveling up everything please go live
1538. 4:36:15simultaneously YouTube Instagram LinkedIn use streamyard yeah I might do that I I got to get my [ __ ] together
1539. 4:36:23with the streaming thing I think that might be a good idea CU yeah I yeah I don't disagree I I'll figure something out there seven minutes
1540. 4:36:33um yeah your watches in 2021 thank you Kyle for the long strange trip it's been an awesome day with you
1541. 4:36:42madcap thank you so much thank you for being a subscriber for one thing um I love the the things you chose there um
1542. 4:36:50and uh thank you for the kind words that's really nice I'm optimistic too overall haha I know use OBS it
1543. 4:36:57integrates better with all the platforms okay I I've looked at OBS it it makes my eyes crossed cuz it just looks like oh
1544. 4:37:05my God there's a lot of switches and dials here I hate interfaces that are like the cockpit of a 747 nice Galaxy that was a nice Galaxy
1545. 4:37:14wasn't it all right listen so let me go in and let some people in who applied for the salon one person everyone else has been
1546. 4:37:31in so you're in now if you're if you went to the salon if you haven't if you don't know what the salon is
1547. 4:37:41um there's the location AIS salon. mn. Co this is a group I started a year ago so chat GPT launched this day a year ago
1548. 4:37:52and next Tuesday a year ago I started the AI salon and yes those two events are related I was talking about doing
1549. 4:38:00something with AI with Leah the woman I started this with prior to chat GPT coming out we were talking about stable
1550. 4:38:09diffusion and she's a photographer we were just talking about it in the context of Art and then chat PT came out
1551. 4:38:14and I lost my [ __ ] mind as evidenced by the last 24 hours so that's what this group is about and it's it's a really
1552. 4:38:23cool group please go look at the about the salon look at our values make sure you resonate with them um but but if
1553. 4:38:29you're interested in joining people that are curious about AI digging into AI I exploring it collaborating on it supporting one another like
1554. 4:38:40commiserating over holy [ __ ] what do I do with my [ __ ] job holy [ __ ] what do I do I need a job holy [ __ ] I just got a
1555. 4:38:48job in AI this is the group they're really amazing and our Oney year anniversary is next Tuesday so the
1556. 4:38:54meeting is from 5: to 7 Mountain Time next Tuesday come hang out we're bringing speakers that have spoken over
1557. 4:39:02the past year back to talk about what's been going on we're going to be talking about like what's been the last year
1558. 4:39:08what's been it what's it been like for for all of these different people in the in the salon I'm really excited about
1559. 4:39:14it all right we're counting down people we're counting [Music] down it's 1158 people oh hey Mark oh hi good to see you
1560. 4:39:31Dar that was very nice of you Dar I appreciate that uh um who else we got I'm sure you know but I just found out
1561. 4:39:37chat gp4 can not only describe im images but it can analyze absolutely um it does that with code interpreter actually wow
1562. 4:39:45this sounds nice I'm so thankful the Tik Tok algorithms landed me on these lives Kyle I have so much renewed drive for
1563. 4:39:53art awesome awesome a human being awesome yes do it go be inspired you will generate you will have more ideas
1564. 4:40:02on the other side of getting into all this than less it's it's amazing um thanks for everything you do
1565. 4:40:09Kyle quick take Jake thank you so much um Let me let me thank first um I got to give a big shout out to the mods so the
1566. 4:40:17Irregulars that are in here that mod Tobias all these people that mod being Not Afraid um they like I didn't ask them to like
1567. 4:40:27hang out for these 24 hours they just did it so so thank you all for modding and and and kind of supporting me the
1568. 4:40:35mods absolutely Rock like it it totally makes a difference for me to be able to just be and know that things are in
1569. 4:40:41there um so thank you all for doing that um let me just thank people as I'm seeing them here steo thank you so much
1570. 4:40:49um camat Can great to see you Christopher Ramsey thanks for being a subscriber Source Camp um a human being
1571. 4:40:55be not afraid you're awesome thank you so much jrc um Tobias I think I mentioned him um let's see who else we got hair growth GE
1572. 4:41:05Dar great to see you um madcap let's see who else we got Andrea thank you so much John hes so many people in here that
1573. 4:41:15that I recognize you guys are really amazing I on chy good to see you all right um so listen I will be back tomorrow night
1574. 4:41:25have a good night I'm G to go sleep sleep sleep I'm doing it now it's midnight it is and I'll leave you with this you can make money
1575. 4:41:38with peace out everybody