Jun 17, 2024 - AI's Future: How Generative Models Are Transforming Creativity

Transcript

1. 0:01[Music] take Highway won't
2. 0:18you way your way and my way seem to be child mama don't understand it she wants to know where I've been I have to be
3. 0:34some kind of natur born fool to want to pass that way again you know that I [Music] feeling
4. 0:49[Music] country long Jesus you yeah you're singing good you're singing
5. 1:03good you got it [Music] going good evening good people [Applause]
6. 1:17[Music] w
7. 1:32[Music] [Applause] [Music]
8. 2:14[Music] good evening good people of the world Happy
9. 2:31Monday hope you had a good weekend do me a favor share the live Dr J shared the live yall got an early start we did we were two minutes
10. 2:45early listen Dr J be like Dr J share the live share the live you're like but Kyle who do I share the live
11. 2:53with you share it with people that would not want to hear about AI [Music] oh I would never send it to Jenny send it to [Music]
12. 3:28Jenny freom came my way that night just like a jet plane in and out side I was hauling ass at a million
13. 3:42miles an hour wondering how hard i' hereit when they came into the [Music] station they said I was bad Beyond
14. 3:58repair but I got no quals with my situation say here I [Music]
15. 4:14am [Music] so say sh sh Shere w't you dare to say Shere Shere Shere won't you dare to say
16. 4:29hey sh sh sh uhhuh yeah leave a message and your number please take your time to want to satisfy
17. 4:47me take all these old fantasies and send them care of [Music] me oh all right good
18. 5:01people play Shine On by the Cs if I knew more than seven songs I absolutely [Music] [Laughter] [Music]
19. 5:17would all right we only have 21 people in here Twi or Tik Tock does not like us tonight let's try to get some more folks
20. 5:24in here yeah the tapy Taps maybe those will help maybe if we got a different spokes model that would help people are
21. 5:32flicking along like nope nope nope just nope nope is that Alec bald when he got fat oh no that's
22. 5:48disappointing we got lots of good stuff to talk about tonight [Music] somebody floated another rumor that open
23. 6:03AI was going to launch something this week I stopped holding my breath for open AI about a month [Music] ago either they're playing
24. 6:16four-dimensional chess where they're struggling to release things although they did it it seems like they did release an updated dolly
25. 6:30three the notification did not go off oh now there's 40 people in here what was the [Music] notification who's knitting image is
26. 6:45that I forget whose knitting image that [Music] is oh that I was live it didn't it didn't go off that I was Live Well Tick
27. 7:01Tock are [ __ ] you know you go live seven nights a week for a year and then you main maintain that five nights a week for however long it's
28. 7:15been and Tik tok's like yeah whatever go [ __ ] yourself whatever I need to be a
29. 7:29young voluptuous Instagram Vixen don't I I need boobies and beach [Laughter] [Music]
30. 7:54scenes you and I keep wait you and I get all along [Music] Sunday morning hair at home sky blue and my C is strong it's
31. 8:11[Music] true you can have mine I'm sick of them all right yeah ship them over ship me the bikinis I'll you know I'll start going full
32. 8:21manini we'll go we'll go manini we'll do virtual Beach backgrounds and I'll just talk about how I'm manifesting abundance
33. 8:29we'll get we'll get this we'll get this puppy up to thousands we'll have thousands in here and they'll come for
34. 8:36the eye candy and they'll stay for the AI education it's going to be awesome he he doesn't look like an influencer and I I don't think I can
35. 8:48unsee what he's wearing um I should leave but I can't look away [Laughter] a that's funny my brain hurts I
36. 9:03think I had an epiphany on llms today do tell I on China what was your Epiphany I've had a couple recently as
37. 9:11well so I'm curious to hear what yours are Joey no on the maninis what I don't know I just I feel like you know I feel so not
38. 9:26objectified you know what you know what it is I feel like you respect me for my brain and and I just want to be you know influencer
39. 9:37meat maybe if you sold something Kyle yeah no [ __ ] have you tried the new what what's the new 3D printer the new bamboo Labs
40. 9:463D printer it's my favorite 3D printer of all time I'll tell you what hey I just got in my new shipment a Madagascar
41. 9:53vanilla we buy it by the pound yeah I know I'm not selling enough you can call in all right yeah hop on I in [Music]
42. 10:10China is there an AI to sketch out my product ideas or do I have to Forfeit my cop copyright if I spill all the
43. 10:18beans um that's kind that's a bigger interesting question like you're not you're not giving up your it's
44. 10:28just um open AI may use your wait Ian China are you here yeah he's here okay yeah um you may you may just you know
45. 10:38they may use your stuff for training but they're they're not actively going through people's [ __ ] and Mining it for
46. 10:44ideas what's happening what's your what's your so I was I was doing my walk today hey so I don't know if it's really
47. 10:51an epiphany but for me it is anyway so I was doing my walk today and I was thinking about llms and you know it
48. 10:56occurred to me that everything that I've been sort of trying to understand about them has been rooted in natural language
49. 11:03and how you know the statistical probabilities of the way the uh llm Works in terms of predicting the next
50. 11:10word and the next group of letters and the next words and phrases and everything like that it's all centered
51. 11:15around language but and you know the way it keeps vectors of numbers that that really are um abstract representations
52. 11:24of those predicted next words and so on and so forth and then I thought about what if you don't store natural language
53. 11:30in the llm what if you store data that represents some motion or some some aspect of reality for example the three
54. 11:39body problem which is very known difficult problem to try to predict the Motions of three bodies gravitationally
55. 11:46acting on each other how you could store that data in a in a huge I mean much bigger than the biggest llm now to help
56. 11:54predict where the Motions were going to be but that's just one example the other example is like autonomous driving car
57. 11:59like a Tesla you know how they threw out their old code and they mov towards an llm and they're basically training the
58. 12:06llm on the presentation of what's coming down the street and so and then I got into the thought of what else can you
59. 12:13make can you take that's abstract data not natural language and build a neural net an llm from it to predict and tell
60. 12:23you about something that's going to happen in nature next whatever that something is well that's so that
61. 12:29a lot of what's fascinating about the Transformer model is when you embed stuff when you create that Vector space
62. 12:37like words are just one kind of data so so how we're getting you know all these different kind of models that understand
63. 12:45language and understand protein foldings and things like that is they're basically just throwing different data
64. 12:50sets exactly how you're describing it at the Transformer so it's it's not just large language models it's large all
65. 12:57sorts of different kinds of models right the music stuff's created so yeah that that's it's it's it's something I've
66. 13:04been thinking about a lot is you know what are the data sets that you know data Geeks aren't super excited about
67. 13:12that are that are really obscure and valuable that no one's even thrown at these things yet right there's there's a
68. 13:18bunch of stuff that you know people are going to be creating these models around that we can't even imagine right now um
69. 13:26so yeah it's huge it's huge that's the thing is the Transformer approach this approach of creating this Vector space
70. 13:35and being able to sort of reach into it no matter what it's trained on that's that's how they're doing all these um
71. 13:41when they're when they're reading people's uh brain scans right right images they saw that's how they're doing
72. 13:47it they're just training it on different data not on words so that's exactly amazing that's why it kills me when I
73. 13:52hear these AGI you know um uh you know holier than now types talk about oh you know uh geni is a is a is an off ramp
74. 14:04you know it's an off ramp I feel like we haven't even scraped the God darn surface yet of this stuff and and the
75. 14:11other example that I was just reading about yesterday is there is a uh there's a fight between rare earth metals in in
76. 14:18the world going on right now and some of these rare earth metals are extremely strategically important so now they're
77. 14:24using uh gen llms to help predict where to find things like lithium and ridium and Palladium basically they're creating
78. 14:35a training a data model on Mineral other minerals that exist in other places and they're trying to teach it where to find
79. 14:43these rare earth metals based on what else is in the topography which is just exactly what I'm talking about it's
80. 14:49creating a an llm from data not from language yep yep yep and that's there's lots of stuff out there and a lot of the
81. 14:59the you know old school AI like like that's a lot of what they were doing it's just the Transformer model lets you
82. 15:04do it a massive scale right and it can be on anything yeah so that's coming my my Epiphany around large language models
83. 15:11I talked about it a little bit on here I think you've been on when I when I've talked about it but what's really hit me
84. 15:18I tried to figure out a way to articulate this idea that when you interact with chat GPT they call it a
85. 15:23chatbot but it doesn't feel very botlike it feels more human and but you know we call it artific IAL intelligence and we
86. 15:30call it AI Bots and things like that but um I saw I saw a tweet of this and and the Tweet was AR AI is the collective
87. 15:40intelligence of humanity and and what hit me is why chat PT and other llms feel so human is because you're you're
88. 15:50not interacting with the mechanism the mechanism is a calculator but what it was trained on is like the output of humanity right
89. 15:59yeah the Corpus the Corpus what you're actually interacting with is Humanity like the collective int of everyone that
90. 16:06came before us and that to me was like holy [ __ ] that's that's why being good with language and being good with people
91. 16:14like e Ethan mik put this out today that those people that are better with people are better at generative AI because
92. 16:23you're essentially interacting with like the ghosts of people it's bizarre ABS perect sense like that phrase to me
93. 16:29collective intelligence that's it that's what this stuff's all about awesome thanks so much Kyle I'm going to let you
94. 16:36get back thank you so much pleasure having you up here this great appreciate it cheers all right so yeah let me turn on my duly camera
95. 16:47again so bunch of things if you're new here welcome my name's Kyle Shannon this is the AI learning lab you know what we do
96. 16:56here we talk about trips to Barbados and manifesting our abundance let me go get my bikini um God I love it here does that
97. 17:10make me a nerd yes yes that makes you I'll tell you this channel is just we're gonna take it to the next level as
98. 17:20soon as I get twerking lessons I just got to find myself like a 20-some twerking influencer and then
99. 17:27I'll be like well we'll put on that song of like those Irish kids you know something I doubt it I doubt and then I'll do the dance
100. 17:39and then I'll start wearing maninis by the pool like it's gonna be we're gonna hook that [ __ ] up the guy from open AI said that they
101. 17:48aren't just stringing words together anymore yeah I know and Jeffrey Hinton said that as well Jeffrey hinton's been
102. 17:54on a couple of new interviews lately he's the Godfather of AI and he's saying that um the models used to PR predict the next word or next
103. 18:05token but they're now actually understanding the concepts so there's there's definitely a shift in the Jeffrey Hinton was also saying we
104. 18:13don't understand how they work but we know that they they're starting to understand things um I asked chat PT
105. 18:21give me the summary of collective intelligence of all of humanity into categories oh fascinating and what did
106. 18:28it say that's awesome that's awesome like what are the categories of all of collective intelligence there's food there's food there's
107. 18:45shelter cuno I have an abundance now in my bikini I need OIC yeah no [ __ ] actually you know what if uh if I
108. 18:56really wanted to make money i' I'd get on the uh the OIC uh influencers program get some get get some commissions and
109. 19:06lose some of this uh middle-aged manfat Hinton said they gave them the algorithm and they didn't understand
110. 19:16anything after that yeah I know I know it's [ __ ] Bonkers well and the the thing there so there's a couple of things what I want
111. 19:24to talk about tonight what I want to show is some of the new stuff that came came out from Runway ml they just
112. 19:30announced their gen 3 model and it looks like the [ __ ] I've seen from Runway gen 3 to me looks better than Sora like Sora
113. 19:45while it was really cool it still had glitchy like physics weirdness to it the stuff from Runway gen 3 which
114. 19:54which is apparently going to be here in days not [ __ ] weeks or months or years like Sora and then we had Luma labs and I
115. 20:02read I I learned something about Luma Labs that's absolutely [ __ ] mind-blowing and then um and then deep
116. 20:11mind today so deep mind if you don't know them they're the they're the AI arm of uh of Google um I heard an
117. 20:19interesting thing that Dennis aabus who runs Deep Mind he's the co- CEO I guess of of is he the co-ceo or he's I don't know he's up there uh
118. 20:30for Google that he might become the CEO of Google that would be interesting but anyway they announced today progress on video to
119. 20:43audio and you're like wait what so you know how all these video tools they generate silent movies well deep Minds
120. 20:51tool takes those silent movies and interprets them with a prompt and does all the sound effects for that clip it's like
121. 21:00[ __ ] so I figured we go through all that stuff because there's a there's a lot to unpack there um there's a lot to unpack there so so
122. 21:09we can do that so anyway if anyone has any questions or any other thoughts I asked meta AI sunglasses to evaluate my
123. 21:15golf stance they told me to straighten up instant Improvement no [ __ ] that's really cool Brandon how are
124. 21:23you liking the The Meta AI sunglasses and you can if you want you can hop up on stage and chat about
125. 21:33it got a meta Quest 2 headset for Papa's day still a little janky but fun interesting I have so many questions
126. 21:44about so many things tonight I know well so feel free pop the questions in and I'll do my best to answer them but I
127. 21:51just want to show some of the new stuff Brandon what's happening hey how's it going good so so how long ago did you
128. 22:05get those things uh I've had them for about two weeks now they're just the sunglasses though so I have to wear them
129. 22:10outside okay and what's the what's the verdict two weeks in are they are they have they have they made have they made
130. 22:17the Shelf yet or actually using them I am I am actually using them um they are I'm using them more for the camera it is
131. 22:26really nice especially when I'm playing out with my kids in the backyard to be able to just either say take a picture
132. 22:33or tap the shutter and the kids love it they're like Dad take a picture of take a picture of us with your uh glasses um
133. 22:42really high quality picture and video actually I'm really impressed with the quality of the camera um it does not
134. 22:49hold up from a a large language model standpoint quite yet um but I assume they'll be making it is it is but it's like I asked
135. 22:59it when I was out in the golf course I was really giving it it's trying to put it through its Paces for Father's Day
136. 23:04Happy Father's Day by the way um that I'd ask it like is this Poison Ivy and it's like I can't talk to you about
137. 23:11medical stuff like but that's not MediCal okay uh and and a couple of times I'm like what what am I looking at
138. 23:19it's like please reconnect your meta 3 to the or meta AI to the internet like I'm standing out in the middle of a golf
139. 23:26course I have five bars but other than that when it works it it is really nice like you know my kids are like hey what
140. 23:33is that plant and I ask what the plant is like that's basil I'm like oh that's cool and so how does it work so so you
141. 23:39just talk and you go hey meta or whatever the [ __ ] you say and then and then it's got Speakers by your ears how
142. 23:45like how does it actually work yeah it's like bone conducting headphones or whatever so it's over the over the ear
143. 23:51so I can hear it and people standing next to me really can't hear it that it's I wish it actually had like an
144. 23:57output like I could turn it on speaker mode because a lot of times I'm demoing it and people ask me to ask it a
145. 24:03question and I have to wait for it to answer and then read back what it said because they can't hear it but um it is
146. 24:10really nice I'm looking forward to using it this summer when we're out and about and it is really nice like hey what's
147. 24:15the forecast you know just the random questions that come to your mind where you would normally whip out your phone
148. 24:21unlock your phone go to Google type it in you can just ask and it's just there so yeah I like them they're I you know I
149. 24:29mean they were worth the 250 bucks or whatever they're charging for them now I wouldn't wear them as a regular
150. 24:36glasses yet though they're they are um they are not uh they're a little bit too nerdy for to there's because they're
151. 24:45really thick frames the sunglass Shades mask it really well but if you're wearing them like normal I look like
152. 24:51that emoji the nerd Emoji it's just really thick yeah they're really thick yeah yeah yeah yeah all this stuff's
153. 24:59gonna get gonna get uh smaller but anyway all right cool awesome well thanks for the review and by the way so
154. 25:05this this book right here M Mr Mr Brandon tid made this so one night we were doing I forget what I we were just
155. 25:12doing something in DOI and we did like do an alphabet and then you got inspired by that so this is my first journey in
156. 25:18the world of tomorrow and it's an alphabet book for all sorts of different geeky terms so and with with little
157. 25:26pictures that were made in DOI so y that was your custom GPT that did that authored that book and I just
158. 25:35copied and pasted it into a format for Amazon Kindle Direct publishing and then my kids love it that's awesome man so
159. 25:44cool all right I don't even remember what was the name of that GPT do you remember I don't remember I don't
160. 25:51either but I should also talk about Sin City sipper all right Brandon thank good talking to you yeah take care cheers
161. 25:59all right great book yeah no it's really cool [Music] um all right there's that um I updated Sin City sipper so let's let's flip over
162. 26:15let's go look at some [ __ ] will analyze the curvature the putting green and offer advice I'll tell you what you know
163. 26:21Tom with with the liar stuff um with the lar cameras like those those glass say they're going to get really
164. 26:29sophisticated where they'll probably have whole arrays of you know sensors that are beaming [ __ ] out so yeah
165. 26:36they're probably going to be able to generate 3D actually you know you could probably play a round of golf and then
166. 26:44not only will it analyze your swing but you can watch the replay of your ball you know going out of bounds you know
167. 26:50with all the curvatures it'll it'll do gaui and splatting and recreate the golf course for for you in
168. 26:563D wait this [ __ ] shit's about to get off the [ __ ] hook I'm telling you just so like tonight's tonight should be
169. 27:03fun because there's a lot of there's a lot of stuff to look at but the first thing we're going to look at is is me talking about Story
170. 27:12vine so what is Story vine and how does it work the video that you're watching I was talking about Story vine all damn day all
171. 27:21right let's get rid of that okay let's see let me go [Music] let me go to chat
172. 27:35JP oh I'm already on Sin City sipper so if you're on chat je and you've got access to either Plus or free go to explore gpts and search
173. 27:52for Sin City sipper and this is the very first Contour mapping of the golf course yeah that'll be coming that's it that'll be
174. 28:01coming if it's not already here I would imagine that that Contour [ __ ] you can do with the ldar cameras there's
175. 28:08probably a specific app out for that right now on on the iPhones um this was the very first custom GPT I made if you don't know what
176. 28:16a custom GPT is it's basically just chat GPT that does something special and what this one does is it will invent drinks
177. 28:24for you so um I just just the reason I'm showing it is I just updated it because um open AI shifted from gp4 to gp4 Omni
178. 28:38for the engine for all these gpts and they don't they didn't work the same like this stopped rendering HTML that I
179. 28:44had in the code and it uh it wasn't always making images it was kind of [ __ ] up so I went in and I fixed some
180. 28:51stuff and then I made I made it um I I kind of updated it so let's say we'll say um invent a festive Burbon based drink for the
181. 29:08upcoming 4th of July holiday and so this is called The Star Spangled sipper let's see what it is 2 ounces of bourbon good an oun of lemon
182. 29:24juice okay an ounce of simple syrup half an ounce of blue ciraa two dashes of bitters fresh blueberries interesting lemon twist for garnish
183. 29:37ice and then what it does oops I got to turn off my voice to text because [ __ ] happens is then it makes an
184. 29:47image of the drink and it now creates a custom coaster that gives you the name of the drink and it says by Sin City
185. 29:55sipper how [ __ ] cool is that and that drink actually looks really good although we got a we got a a star
186. 30:04in it we got to get that thing out of there curo H yeah that's interesting isn't it are the custom gpts in po open to all
187. 30:14users yes they are isn't yeah isn't the coaster cool Source camp and it says by City sipper so Source Camp when you're
188. 30:23demoing this to restaurant people um like this is easily modifiable for a for like for a bar so imagine like this was Source Camp
189. 30:35had this idea she was she was showing chat GP or showing Sin City sipper to a restaurant tour and said you could
190. 30:43create like an iPad where people while they're waiting for their tables could design their own drink based
191. 30:50on the booze you have the most of so here's a cool thing you could do with if you don't know chat gp4 has
192. 31:00Vision so for example you could take this was Source Camp told me she did this this blew my mind you can take a
193. 31:07picture of the bottles on the bar and have it determine which ones are at what percentage of emptiness so so you know
194. 31:15which one to you know to reorder you could similarly say take a picture of the bar and then say give me drinks that are only based on what I
195. 31:27have have in the bar right so it's always using the booze you've got available um and then yeah you could make it so so that the image it
196. 31:36generates has the name of your company on it something like that crazy right just have the star change to a thin
197. 31:45slice of star fruit yeah there you go that's perfect all right I have it making drinks with booze we need to get
198. 31:52rid of at the restaurant there you go Silver Fox that's it and and that's that that um that use case um someone shared
199. 32:02with me I forget who it was they did a Super Bowl part party this year and they took a picture of their bar and they
200. 32:08said come up with 50 Super Bowl drinks and you know present the menu so that we give it to people and then create the
201. 32:15recipes for the bartender so they had 50 custom cocktails made by Sin City sipper and uh and it used only the booze
202. 32:24they had available I think that's really slick all right is isn't that the purpose of a special to get rid of the
203. 32:30Overstock well it's still good there you go yep exactly is that an anast star no that's just well it should be but it's just it's
204. 32:43just confetti it's the same as the [ __ ] on the on the ground but Isn't that cool and so now it correctly says this
205. 32:52GPT was created by Kyle Shannon co-founder of the AI Salon Creator and host of the AI learning lab on Tik Tok
206. 32:58so it's a little ad for you know for the community and and this and that so anyway go use Sin City Seer all right
207. 33:06enough about me let's talk about [Laughter] me all right what's that oh yeah that's that okay okay
208. 33:21so let's do some history here d
209. 34:00oh that's cool Lord digital Gods what did you make oh that's pretty slick except you know it would be much
210. 34:14better if I actually went in and put titles in all these things when was the last time I actually did a title
211. 34:22December November 30th is the last YouTube video I actually have um a description of we'll get that going Brandon and I
212. 34:33were working on that for a bit I think I dropped the ball on that all right let's talk about you what yeah exactly enough
213. 34:39about me what do you think about me um okay so for for many moons um Runway ML and Pika were the big
214. 34:54texted video things and then Sora came out and Sora was announced does this say when it was announced read the technical report February
215. 35:0815th so many many months ago open AI comes out with here's the this is the famous famous woman famous Japanese woman walking down
216. 35:23you know Tokyo Tokyo streets in the rain 59 second video and and we still haven't [ __ ] like this thing still hasn't
217. 35:34launched so February March April May June 15th so almost exactly four months and we still don't have anything
218. 35:50then Google announced Vo our most capable generative video model
219. 36:08and no word on when it's coming out and then last week was it only last week I think it was last week last
220. 36:24week this Sora killer called Luma Labs showed up on the scene and so like here's a video this is a video I made I I went and got a
221. 36:38um just a picture off the Internet this was of The Shining like a Lego kind of thing and so this is an image to video
222. 36:47with Luma Labs it's called Dream Maker and it's Luma labs. and so like holy [ __ ] like out of nowhere this Luma lab showed up and you
223. 37:01know this might not be Sora level good but it's pretty [ __ ] good and you know if you if you look at some of the
224. 37:08videos like this is a video I did of Alvin ay dancers that just morph into monstrous monstrous [ __ ] aberration of nature uh so it
225. 37:21still got all the sort of morphing and weird ass hallucination stuff but like this was a picture I did of I did a
226. 37:28mockup of a of the musical that I'm working on on that little picture and then it just made all this New York City
227. 37:36it it panned down to the street and then this whole part of the scene is completely invented um so that's really impressive and
228. 37:47then I read this no this no where the [ __ ] is it did I put it here hang
229. 38:05on Luma Oh Know by the way this tool I think I learned this from Riley Riley what's his name um
230. 38:20Luma yeah this one okay so this dude research scientist working on video generation at Luma laabs previe
231. 38:36previously video poet at Google AI so Benjamin decracker deer says I realiz Luma dream machine is better
232. 38:49anything Google has shipped and Dan said now you know why I left Google to join Luma I was in the team that developed Vo
233. 38:57early on but knew it would never be shipped to the masses for quite a long time same as Sora not until a company
234. 39:05like Luma forces their hand he quit Google went to Luma they [ __ ] launched Luma in three months three months three months
235. 39:22so um we have a model that's doing this kind of [ __ ] that was launched in 3 months so so we're now starting to see
236. 39:32like little whipper snapper of companies you know sort of outclass and outpace these big research companies now
237. 39:41is this stuff going to be as safe is it going to be as um coherent is it going to be I I probably not like this is probably not super um
238. 39:52responsible but I think that's independent of the fact that this stuff is accelerating is starting to happen faster [Music]
239. 40:05um yeah so that's the next thing coming okay but what I wanted to show on the Luma stuff did I lost all my hang on let me go back
240. 40:20to this [Music] thing okay okay so Luma came out a week ago after three months of nothing they just show
241. 40:33up on the scene they've already added the ability to extend the video so it'll make like a 7-second video then you can
242. 40:42extend it to 15 and then you can extend it to 22 and then you can extend it to 28 right so you can just keep extending it so that's
243. 40:52cool and then they where is it yeah this thing
244. 41:12oops so coming soon to dream machine um here's a sneak peek of something powerful editability and intuitive
245. 41:24controls so let me make this video full screen and then let's we'll stop and start okay so they're doing multiple
246. 41:35aspect ratios I think they all do that okay so let's see share more related ideas so it's it says dreaming
247. 41:52some ideas about a young Apprentice discovering magic so rather than creating one video it's creating three so that's
248. 42:03cool explore Concepts look at all these new things there's a little swipey finger thing there's a starry thing
249. 42:11there's a other thing there's a paintbrush thing and there's a whatever the [ __ ] he's clicking on here [Laughter] [Music] thing wait
250. 42:27so so he clicks the button I guess this was the magic button he clicks the button and they're doing segmentation Young Apprentice is here
251. 42:38bluish tint is here afternoon is here Forest is here magical book is here and then they're they're changing that
252. 42:52stuff look at this he's he's over changing the background and the The Apprentice is staying the [Music] same precise
253. 43:17edits so so that's Luma so that's the thing that's like a week old that the guy [ __ ] said [ __ ] it I'm leaving Google I'm going to go
254. 43:30make some [ __ ] that can actually launch 3 months later they launched this thing now was it in development before he got
255. 43:38there I don't know possibly probably which platform is this this is Luma lab's dream dream machine then today Runway I can't keep up anymore
256. 43:55lunatick this is [ __ ] Bonkers because that's just that's just that's just Luma Labs that's just dream dream
257. 44:05machine now Runway which we know and love today they announced gen 3 Alpha and it's supposed to be launching in a few
258. 44:14days so none of this Sora will show you the coolest thing ever and then maybe Ashton K Kutcher gets to play with it
259. 44:24because he's [ __ ] Ashton Kutcher what have I been punked how about the AI learning lab I've got nearly 35,000 followers
260. 44:36like what's Aston kusher got like 60,000 [ __ ] so anyway they're never [ __ ] launching that thing so this thing is
261. 44:50supposed to come out in a few days so we' so we've got Luma sort of goes from nothing to Something in three months and
262. 44:57and is already iterating features rolling out features you know they are they just added extend 3 days after they
263. 45:03launched and now they've got this this editing coming that's got this layered ass editing and then you've got gen 3
264. 45:10Alpha and and again this should be here in a few days okay so let's go look at some of these videos because every one of the Sora videos
265. 45:19that I saw when I watched it slowly I was like eh it's it's like 80% right like there was always some weird
266. 45:29physics [ __ ] up in it it was good and the coherence was really impressive but it was like 80% these examples that I've seen from
267. 45:39gen 3 Alpha I like I don't I don't see anything wrong with these I just don't oh the other thing six million what's six million
268. 45:57I don't know what six million is in reference to Luma is two years old by the way okay that makes sense I I'm like they
269. 46:06couldn't have made it in three months from the time that guy got there it was three months till they launched
270. 46:10something great Ashton's followers oh six million oh so what so so Ashton's got a few followers Kutcher Kutcher six million listen listen
271. 46:22Ashton you think just CU you're like like famous and B been in movies and like we're married to Demi Moore and
272. 46:30like have 6 million followers like somehow you're more deserving of access to Sora than me fine fine be that way all
273. 46:43right so there's that IM there there's that video there's this this is a little weird and wonky in the some of like physics around the
274. 46:53knees but like the coherence of the stores going by are pretty crazy it's pretty good his face isn't morphing and moving weird his hands
275. 47:04aren't like turning into [ __ ] octopi are we looking at Runway yeah this is Runway gen 3 Mary Mary to me more oh yeah we're on Runway
276. 47:18now six million followers less than the AI learning lab exactly this the 35,000 that are here is it's quality not
277. 47:27quantity I could have 6 million followers easy that's that's I reserve the right to wear a manini we will
278. 47:37have dozens of followers leave on that first day please do not denigrate yourself comparing yourself to Mr Kutcher I
279. 47:50actually think Mr Kutcher is he's very smart and he's he's got a a really good sense of humor I he's one of those guys
280. 47:57that you really want to hate but he's actually kind of a nice guy and really [ __ ] smart so I I I actually
281. 48:03appreciate uh Mr cucha um so anyway that's that video this one is stunning The Prompt here is fpv flying through a
282. 48:15colorful Coral line streets of an underwater Suburban neighborhood and it like starts out there's like a Toyota
283. 48:21Prius in the corner and like the the corals grown up around the houses like bushes and trees it's like a little
284. 48:27established and look at the little road and look at the like the bubbles like again I don't see any major physics fuckups here like this
285. 48:40seems pretty [ __ ] good the turtle's weird that fish is weird so there's there's little things but like I
286. 48:49wouldn't notice that this one's a little odd the the shape of the balloon but there's a balloon floating down a street in some European
287. 48:59alley is this text to video or do we need to start with an image I think it does both but I'm not sure um let me see oh by the way
288. 49:13um gen 3 Alpha is the first the first of an upcoming series of models trained by runway on a new infrastructure built for
289. 49:23large scale multimodal training multimodal training m multimodal training it's not going to be just video it is a major Improvement in
290. 49:33Fidelity consistency and motion over Gen 2 and to step towards building General World models this was the thing Sora was talking
291. 49:42about Sora was talking about we're not making video we understand and are simulating the physical world Runway is
292. 49:52building General World models moving toward multimodal training so this thing's going to do sound this thing's probably going to do 3D
293. 50:03worlds I wouldn't be at all surprised if it does language and if it does language it probably does programming and if it
294. 50:10does programming it probably does game mechanics and simulations like just if you think about what they're thinking
295. 50:22about where everyone's headed is systems that understand the world understand and cons simulate the world the output is video but what it
296. 50:33actually understands this goes back to what ion chy was talking about that wait you throw different data at these things
297. 50:40they get good at crazy ass different [ __ ] okay so let's go look at some more examples um again as it's zooming in there was a
298. 50:53little weirdness there but that was was pretty good this thing is very very coherent like the people in the background are
299. 51:04staying pretty consistent no one just materializes that balloon floating pretty good that dandelion doesn't
300. 51:16really look like a dandelion but it's not bad there's a Tokyo train running through a Tokyo neighborhood
301. 51:27fine grained temporal control and again so if you're new to if if you're just joining this is Runway ml's gen 3
302. 51:36preview which is apparently coming out in a few days the whole we're living in a simulation argument is starting to make
303. 51:45more sense Brandon you know what's funny a year ago like people have been talking about we're living in a simulation for the
304. 51:53past [ __ ] decade like the evidence points to and like physicists like just some smart people are like it's possible
305. 52:00we're living in a simulation like what the [ __ ] are you talking about like we would know it's a game and now I'm looking at this going
306. 52:08would we would we I was just tinkering with that too I know [ __ ] Bonkers it's [ __ ] Bonkers we still haven't learned our
307. 52:19lesson of taking press releases as the truth have we well but what what I said Pate was uh apparently this is coming out in in a
308. 52:29few days so if by the end of the week today's Monday a few days says to me that we will have this by Friday if we
309. 52:37don't have this by Friday I will put it in the [ __ ] pile of Sora uh where you know we can make fun of it four months
310. 52:45from now for still not existing but anyway fine grain temporal control it's been highly trained with highly
311. 52:53descriptive temporal dense captions enabling imaginative Transitions and precise key framing of elements in a
312. 53:02scene don't quite know what that means but it sounds like they can do some pretty sophisticated isolation of
313. 53:12elements this is really similar to the um the Tokyo train video that Sora did that's really interesting with the
314. 53:22little flashlight on the walls dude on a mountain mountain isn't
315. 53:38morphing yellow sky is weird but cool water's looking decent it's a
316. 53:53little weird in here but decent again like I'm looking at this stuff pretty hard to try to see is there anything here that's like an obvious
317. 54:05giveaway if you didn't know you were looking if you weren't looking for weirdness there's nothing here that jumps out to
318. 54:19me these tunnel tracking shots are always they kind of invite you to find the weird ird Ness in them the the fpv Drone drone
319. 54:37things actually this one's actually quite impressive because normally with um with the video models like the pen would
320. 54:47start out looking like a pen and the fingers would start out looking like fingers and then the pen would go Z and
321. 54:53you know it would turn into a finger or it would like be you know intersecting the finger and then the other fingers
322. 54:58would be doing this look at look at the hand here like that's a hand drawing a line little weirdness in the lines right there
323. 55:17but crazy from the shelter of My Mind through the window of my eyes Ryman Simon some bugs this is another Sora shot
324. 55:30knockoff clearly these these guys are are uh [ __ ] with the Sora team like that's a that's a pretty impressive shot start zoomed in on bugs
325. 55:44pan out to the Suburban neighborhood none of the houses are moving the bugs didn't turn into three bugs or five bugs or 10 bugs
326. 55:58this is tsunami in the streets which water and waves are always weird there's what's interesting about
327. 56:07this that moment right there water just materializes is weird and right there the water had weird structure to it but
328. 56:15like this stuff and that you could absolutely use you could there's there's usable pieces of this that you could use
329. 56:22even though some of it's weird this is an impressive shot I think this is where we started all right so that's that photorealistic
330. 56:32humans excels at generating expressive human characters with a wide range of actions wide range of actions gestures
331. 56:46emotions unlocking new storytelling opportunities no weird weird body proportions well maybe a little weird
332. 57:00leg there but generally pretty impressive what does this mean temporarily dense captions um I definitely know what that means but I'm
333. 57:13going to turn it over to Pate to answer that because um densely what was it densely accurate captions means they
334. 57:22they have act they're they're not sparse uh Pate why don't you why don't I could easily explain it but I think you do
335. 57:32better with Concepts like dense uh captions I don't [ __ ] know um lots of data tags and annotations throughout the
336. 57:44videos for training oh for training on oh okay cool interesting oh that's actually what it means so he did answer it great um
337. 57:57yeah these humans are looking pretty human the camera movements are weird but not bad I've been getting amazing Clips with
338. 58:09Haier and pixiverse temp oh temporarily Through Time captions words oh I see oh dense dense captioning over time got it got it got
339. 58:19it okay so when they're tagging the videos for training they're tagging a bunch of [ __ ] oh no I guess probably when they're
340. 58:28tagging the videos after it was rendered as well that's probably how they're doing that interface where they're
341. 58:34popping up all the different words and they can change out the backgrounds and the foregrounds and things like that that's
342. 58:40fascinating so dense temporal uh what was it descriptions is similar to the it's similar conceptually it sounds like
343. 58:56to what Sora was using with their with their little uh um their little grid of Cubes I forget what they were called points no I
344. 59:07don't know whatever interesting okay let's see what else we got here looks realistic and again to pate's point these are the these are the
345. 59:19marketing clips that they chose so these are good but it's not bad like the hands on the piano they're a little weird his
346. 59:29right hand's a little weird that one but they're not
347. 59:44bad like that's impressive because normally you know a Spaceman and a suit like that his face would be all over the
348. 59:54place this one blew me away watch this the hair is good it's consistent prompt a middle-aged sad bald man
349. 1:00:07becomes happy as a wig of curly hair and sunglasses Falls suddenly onto his head around his head that one's pretty [ __ ]
350. 1:00:21impressive it does dark and M Moody cinematic really nicely apparent L okay yeah that's where we started four artist
351. 1:00:30by artists was a collaborative effort from a cross disciplinary team of research scientists engineers and
352. 1:00:35artists it was designed to interpret a wide range of styles and cinematic terminology so giant strange creature Joker GNA give reality a
353. 1:00:46one-star review today's janky is certainly nothing like yesterday's janky I know like we're really having to like search for the Jank
354. 1:00:57that's what I said embrace the Jank and you know enjoy the Jank while we got it all right this is another one that's a
355. 1:01:05SORA a man made from rocks walking through the forest full body shot so there's weird like the rocks are like on
356. 1:01:14fabric because they move when he walks but like there's no additional rocks rocks aren't merging together he's basically staying coherent
357. 1:01:25that's pretty amazing slow cinematic push in on an ostrich standing in a 1980's kitchen
358. 1:01:41damn that was not an ostrich beak but we'll forgive them for that a giant humanoid made of fluffy blue cotton
359. 1:01:48candy stomping on the ground roaring to the Clear Blue Sky zooming through a dark forest with neon neon light Flora lighting
360. 1:02:00up I'm so sick of this neon aesthetic like everything is this neon aesthetic that's cool a cyclone of broken glass in
361. 1:02:10an urban Alleyway Dynamic movement I mean that looks like a [ __ ] shot from a movie right where someone's got a dream
362. 1:02:23sequence or they discovered their magical powers and they put together the ball of glass and send it down to the cops and cut them
363. 1:02:33up the Matrix Salon okay that fire is really
364. 1:02:48impressive things fluid dynamics are normally really not good this is this is super impressive highly detailed closeup of a
365. 1:03:05bacteria ultrawide shot of a stone hand reaching out of a pile of rocks at the base of a mountain yeah that's pretty much that isn't it like
366. 1:03:17that's I see nothing in that not other than just it's a weird if someone took the time to make a hand that big it
367. 1:03:25would probably have better proportions it looks like a very thick man's hands he was very short with stubby fingers but you know other than
368. 1:03:39that nothing this is trippy aerial view of a cloaked figure elevating in the sky between skyscrapers oil painting of a natural
369. 1:03:54Forest environment with colorful maple trees and cinematic Parallax animation so it's doing The Parallax animation
370. 1:04:02that's where [ __ ] in the background moves slower than the [ __ ] in the foreground so it gives it depth there you go that was
371. 1:04:13that lady in a car can you see her reflection you can't really see her reflection but she stays really coherent so does everything in the
372. 1:04:24background doors opening up on this ship or whatever the [ __ ] it is industry customization as part of the Gen 3
373. 1:04:36models we've been collaborating and partnering with leading entertainment and media organizations to create custom versions
374. 1:04:43of gen 3 ah we talked about this mark my words we're going to be able to license official models so imagine an
375. 1:04:58official Toy Story model or an official Disney model where it understands the animation style and you'll be able to generate licensed
376. 1:05:09content from from your favorite companies customization of gen 3 models allows for more stylistically controlled
377. 1:05:17and consistent characters and targets specific artistic and narrative requirements among other features like say
378. 1:05:28licensing for companies interested in fine tuning get ready to write a [ __ ] [Laughter] check running toward the rocket launch
379. 1:05:40I'm pretty sure you're not allowed to do that ma'am ma'am ma'am ma'am please please ma'am please step back from the
380. 1:05:47ma'am ma'am don't run toward the rocket [Laughter] a dragon toucan walking through the serengetti oh
381. 1:06:04that's nice ah an empty Warehouse where flowers start blooming from the concrete I'd say they bloomed beautiful
382. 1:06:19all right so that's that and then I got one other thing to show you what's this yeah that what's this that's just nothing and
383. 1:06:29that's Sora okay so the last thing to show you is also [ __ ] Bonkers so you know how all these here
384. 1:06:42videos are silent well this is Deep Mind so this is Google Deep Mind and this is just a how did they put it a
385. 1:06:57they're G us giving us an update but they want us to know what's coming and again they haven't launched vo so I
386. 1:07:05don't know when this is coming but what this thing does is generates audio from the video clip so you give it the video
387. 1:07:16clip and then just give it a prompt so prompt cinematic Thriller horror film or horror film music music tension Ambiance footsteps on the
388. 1:07:32[Music] [Music] concrete
389. 1:07:49crazy so this is cute baby dinosaur chirps jungle Ambience egg cracking
390. 1:08:04have you ever tried to [ __ ] do sound design do you know how impossible it is to find something even remotely close to what you want
391. 1:08:18ever just tuck that [ __ ] into existence [Music]
392. 1:08:37crazy cars getting oh this this one I was really impressed with this one's so cool [Music] [Applause]
393. 1:08:51[Music] one minute power tap folks thank you slow mellow harmonica plays as the
394. 1:09:06sun goes down on the Prairie [Music]
395. 1:09:23[Music] w [Music]
396. 1:10:00there's how the [ __ ] works this was bad this is the only one I saw that was really kind of [Music] eh and the only thing bad about it is
397. 1:10:16that if you were playing that high in the neck and you like you you would I don't know it just it just wasn't good music
398. 1:10:32I want to be a fley artist when I grow up I was shook when I heard breaking celery is is breaking bones oh that's
399. 1:10:38awesome I want it's funny I wanted to be a stunt man when I was a kid a stunt man or a microsurgeon went for acting this tery
400. 1:10:49looks amazing though I am so hungry all right so anyway isn't that [ __ ] Bonkers people it's [ __ ] Bonkers
401. 1:11:03like I don't know quite how to process all that so we should have Runway so we've got you've got Luma Labs right now
402. 1:11:16um if you don't pay for it it's slow as [ __ ] like hours um and you're limited to 10 videos a day if you do pay for it it's $23 minimum I
403. 1:11:33think you get 100 or 150 or something like that there's no daily limit and it's faster you get priority um but none of these media
404. 1:11:47things are going to be cheap because it just takes a lot more processing but you can go use that for free and then Runway
405. 1:11:53ml we should have that this week we and then at some point Sora is going to [ __ ] launch like it's not going to not
406. 1:12:01launch um apparently gen 3 um gen 3 from Runway is 90 seconds for a 15 second clip which is not bad a
407. 1:12:16minute and a half to generate video of that quality um that's not bad um and then I look at things like okay so
408. 1:12:28you've got this vo model you've got the Sora model you've got Runway saying they're not just making videos they're making World
409. 1:12:37models and that and that Gen 3 is the first in a series of multimodal models that are going to create this world thing system that they have
410. 1:12:56you you take the little hints of video to audio that Google put there higher quality video things
411. 1:13:11we're I'm guessing about a year away from being able to just say make me a five minute movie of XYZ and it'll just go do
412. 1:13:22it with actors and dial dialog and music and sound effects and good [ __ ] and it'll just be done probably a year before there's a
413. 1:13:33tool that can do that they'll be jankier than [ __ ] within three years it'll probably be fulllength feature films or
414. 1:13:44documentaries documentaries with whatever reality you know you put in there but probably within three months
415. 1:13:56you're going to be able to Cobble together all these different tools we're we're going to start to see some really remarkable
416. 1:14:06work where there's either no actors or there's an actor that provided voiceovers or maybe not even that maybe
417. 1:14:14just provided their model and their likeness if you're a Storyteller if you're a filmmaker you shouldn't be
418. 1:14:25pissed off well you can be pissed off there stuff's coming whether you want it or not it's funny I ran into one of the
419. 1:14:32guys from the AI Salon here in Denver I ran into him today and he he keeps moving AI competency he's he's been in the video
420. 1:14:42production business for like 25 years he keeps moving AI competency further and further up on his web
421. 1:14:48page and my advice to him today was get aggressive because a Storyteller who's been telling stories with video high-end
422. 1:15:01video for 25 years that gets AI literate first they're going to be better at this [ __ ] than people who are like oh I made a
423. 1:15:13movie it's with all these tools if you've got competence in a given discipline and you get AI literate first
424. 1:15:22you're going to be able to [ __ ] just kick serious ass for the next two or three years then everything will
425. 1:15:35normalize and work will become optional I'm assuming I can't wait for importing a video source and having it modify I
426. 1:15:43think we're pretty close to that there might even be one or two of them out there that do that there might be something in the
427. 1:15:49stable diffusion Line St stability AI Tech that does that I'm a novel list and the only issue with this AI Jen's
428. 1:15:58content is that will it will not allow for a risky scene yeah that's so [ __ ] annoying Relentless Aaron um there's a
429. 1:16:06couple of things you can do you could try if you haven't played with custom instructions here let me show you oops
430. 1:16:14what did I just do I just got rid of my camera hang on [Music]
431. 1:16:31so in chat J you can go to customize chat GPT and in your custom instructions you could say you're an author that does racy edgy content and
432. 1:16:47then you could say how should how would you like chat GPT to respond you could give it custom instructions that tell it
433. 1:16:52to be nasty um or whatever like specifically how you want it to act it likely will not do it but um it's worth a
434. 1:17:06shot then the other thing to do is if you really want something that is um edgy and [ __ ] up is learn how or
435. 1:17:22have someone technical you know have them install an unrestricted version of llama 3 and while the writing won't be
436. 1:17:30as good it's an unrestricted model um one of the guys in the AI Salon um put an unrestricted version of
437. 1:17:39llama 3 as a character in their slack Channel and he said it's absolutely [ __ ] Bonkers he said sometimes it'll
438. 1:17:48just swear all day sometimes it'll sometimes it'll just go into speaking Spanish and speak Spanish for like 3
439. 1:17:53days no matter what what they prompt it to do it will not stop speaking Spanish he said it's it's a [ __ ] trip um so
440. 1:18:02so so feel free to uh if if you really want to get edgy you can do that with with some of the some of the open source
441. 1:18:10models um but yeah it's it's it's a real pain in the ass it's I kind of feel like you know at this point all of these
442. 1:18:18major companies have over indexed for safety um and it's just [ __ ] annoying like I should be able to describe an
443. 1:18:26image as blood spattering if I'm making say a horror film I'm going to talk about things that might be upsetting in
444. 1:18:35the context of a conference call with a Fortune 500 company but I'm not on a conference call with the Fortune 500 company I'm making
445. 1:18:45a slasher flick let me make some crappy fraking bloody you know so anyway that's that's the kind of stuff that's just
446. 1:18:53annoying cuz what what they basically are doing is they're doing safety to the lowest common denominator rather than
447. 1:18:59doing it optin and context specific right most of the time I want chat GPT to be nice I want to though be able to say hey
448. 1:19:12I need an [ __ ] critic to kick the [ __ ] out of me so that I can you know practice you know giving a speech to a bunch of tough
449. 1:19:23investors and I and I want it to be an absolute [ __ ] so you know I I get the worst of it you know before I get in the
450. 1:19:30room with the real people just say ketchup splatter I know but we shouldn't have to say ketchup splatter that's my
451. 1:19:36point if I'm a director of a horror film I should be able to say whatever the [ __ ] I want but we'll get there just not there now
452. 1:19:46sucks chat GPT after dark can you instruct chat gbt to allow political images I certainly not this year
453. 1:19:58tck I mean if you want to do political images I would just go try all of them and then you can I'm guessing I'm guessing almost definitely
454. 1:20:10do it with some version of a trained uh stable diffusion instance I'm I'm sure someone has created a stable diffusion
455. 1:20:22Laura you know a little model for for political [ __ ] guard rails are what brought me here yep um it's too public and too many
456. 1:20:34messed up people out there yeah I know I listen I get why the safety's there but I'm just saying just ask it for
457. 1:20:40Tarantino mode like I kind of feel like you should be able to opt into the crazy [ __ ] because the fact that
458. 1:20:49there's a model as powerful as llama 3 out there unrestricted thanks to meta is Bonkers to me but it's out there so
459. 1:20:59let's you know let's let's figure out a way to let responsible people maybe you have to register to do crazy ass [ __ ] that
460. 1:21:09you're not going to do bad things in the world I don't [ __ ] know but it's it's solvable I can go into Photoshop and
461. 1:21:20make any [ __ ] image I want Photoshop doesn't go oh I'm a i hey well that's a little risky isn't it there that's a
462. 1:21:27that's a lady with two heads we're we're we're talking about a genetic malut Mal malfunctions there Mal formations that's
463. 1:21:36uh let's go ahead and not not create a photo of that shall we right that would you know artists couldn't do art if it was doing that so
464. 1:21:46they should they they've got to figure this [ __ ] out I know they have bigger fish to fry like copyright and staying in business
465. 1:21:55competing but God damn it um Claude refuses to write a lot of my video prompts that have people in with intention emotions yeah exactly
466. 1:22:10like I understand that we don't want to talk about self harm but the fact that I have to on [ __ ] Tik Tock say that instead
467. 1:22:20of the actual word is ridiculous we can't say the word because it might upset someone like if you're saying the
468. 1:22:35word to do harm in in in a bad context yeah we should not allow that they got AIS that can figure out the
469. 1:22:48context but they're just saying this word bad that word bad you can't say in the name of a drug bad because if you
470. 1:22:57say the name of a drug someone could get strung out on it before this live ends so let's cancel it I don't
471. 1:23:10know one of my one of my favorite people from my agency.com days Aon Sugarman when I said God I was I was frustrated
472. 1:23:18someone it's it's common sense and Aaron just said to me he was very calm he was very Zen always he said Common Sense isn't common it's like
473. 1:23:29[ __ ] it should be it's crazy all right mid journey and GPT of lockdown political images yeah No One's Gonna
474. 1:23:42none none of the major companies the okay here here's the thing I on China any company that has raised 10 billion
475. 1:23:49doll or more is not going to touch politics with a [ __ ] thousand foot poll this year no [ __ ] way did you hear um the
476. 1:24:01press secretary for Biden today said that the videos that are out there of Biden where they're all dancing and he's
477. 1:24:09all Frozen they said that was a deep fake like it it what's GNA happen is there's going to be deep fakes of people
478. 1:24:17doing bad [ __ ] then there's going to be real people of real videos of people doing bad [ __ ] that are like I was just a
479. 1:24:26fake we're there we're there I called this I told you this year was going to get weird and I still feel by the end of the
480. 1:24:36year there's going to be enough weirdness and there's going to be enough um I don't even know if misinformation is right like
481. 1:24:50just unverifiable information that that there's probably going to there's going to be a demand in 2025 for
482. 1:24:59some sort of tokenization some sort of way to prove that I as Kyle Shannon actually am live and that the videos on my channel
483. 1:25:10were actually from me and that they actually haven't been modified like that's all solvable with with
484. 1:25:17tokenization um but there's no real need for it right now because this stuff isn't widespread enough but I think I
485. 1:25:23think as we get closer to November this shit's going to get Bonkers I don't even want to think about it I know generated information
486. 1:25:35government intervention they install puzzle pieces outside the White House cryptocurrency wait cryptocurrency one coin does
487. 1:25:47something like that yeah and and uh you know Sam alman's uh World coin and World ID does that I still keep threatening to to
488. 1:26:01create a world ID section in the AI Salon Discord by the way if you don't know what the AI Salon is go check out that the
489. 1:26:11salon. that's going to take you to a link tree the first link there says the AI Salon Community click on
490. 1:26:18that go to the AI salon now tomorrow so okay let me give you a little rant about Community money with ch everybody wants
491. 1:26:34to know how to make money with chat gbt and there's lots of ways you can make money with it but we're about to enter a three-year phase of
492. 1:26:52weirdness where there's it's going to be a major transition have you ever watched video of an iceberg flip and like the blue ice comes
493. 1:27:03up we're about to like the the the iceberg right now the work iceberg is about to flip and right
494. 1:27:17now the highest valued positions in in our global economy frankly are specialty
495. 1:27:32positions Specialists people that took four years or eight years or 12 years to get their degree and they learned one discipline very very
496. 1:27:47deeply those deep disciplines over the next three years are going to be increasingly done by AI the Tactical execution of those
497. 1:27:57Specialties is going to increasingly be done by AI um and so there's going to be a transition from hiring people because of
498. 1:28:11what they can do their skills their highly valued skills in their profession of choice to who they are and how much we trust them so right
499. 1:28:25now if I want someone to do Kick-Ass Motion Graphics and the only person I know that does kickass Mo kickass Motion
500. 1:28:34Graphics is Charles and Charles was a big [ __ ] douche but he's really good at Motion Graphics well I gotta [ __ ] work with
501. 1:28:43Charles because yeah he's a dick and he overcharges and he's always late and he's not reliable and I [ __ ] hate him
502. 1:28:50but he's the only one I know that can do that well what happens to Charles when all of a sudden anyone with any kind of aesthetic
503. 1:29:00eye can do Motion Graphics as good or better as than Charles well then I don't [ __ ] hire Charles because he's a
504. 1:29:10douche I hire Kenny fredman who I know is AI literate I know has a good work ethic and is really [ __ ] talented because I like
505. 1:29:22Kenny so the reason that you need to get your ass into a community like the AI Salon there's there's couple first one
506. 1:29:31is if you want to get AI literate as evidence tonight with just what I talked about with what's happening in video over the past
507. 1:29:43week you can't possibly learn this [ __ ] you cannot like you literally if you did this 247 and you you had an IQ of 165
508. 1:29:56I think Einstein was 160 so you IQ of 165 you did this 247 you couldn't keep up with this [ __ ] so being in a
509. 1:30:04community allows you to be around people who this person Geeks out on this little piece of it and this person Geeks out on
510. 1:30:11this piece and if you're trying to ask figure out what's going on with this piece you ask someone and they're like
511. 1:30:17oh I figured that out last week they say liberal arts Majors may become high demand at least I'm wishing un unraced
512. 1:30:26Sun they're they're not just saying that the COO of Black Rock now if you're a liberal arts major you might not know
513. 1:30:35what black rock is but they [ __ ] own everything the COO of Black Rock three weeks ago said that he's now starting to hire liberal arts
514. 1:30:48analysts not math not science not statistics liberal arts because horizontal thinking is going to be improved so anyway you join a group
515. 1:31:00like this so you can learn you join a group like this so you can start to build [ __ ] so as you learn
516. 1:31:08these tools you start to build projects you can build them with other people you get to know other people they get to know
517. 1:31:16you and then the final piece of the puzzle is you now have a play place where you can lead and start teaching people what
518. 1:31:28you're learning so as you've learned and as you build you're learning things and if you learn out loud and do it in a
519. 1:31:36generous way and say hey here's what I'm working on here's what I'm learning here's where I'm stuck here's where I
520. 1:31:42have a great success here's where I failed miserably other people are going to be taking that in and learning from
521. 1:31:48that that's what this community is all about now tomorrow night is the thir Tuesday of the month so we have a meeting on the first and third
522. 1:31:59Tuesday of the month tomorrow night is is that meeting the first Tuesday of the month is what we call AI
523. 1:32:08Salon presents and that's where we bring in a speaker which by the way big announcement here the first Tuesday in um
524. 1:32:19August we have someone from LTX Studios coming to speak at the salon so a bunch of people in here got LTX Studio licenses um so someone from
525. 1:32:30LTX is coming to speaking but the third Tuesday of the month is called the AI Salon meet and greet if you're new to this stuff or
526. 1:32:40even if you're not but if you're new to this stuff get your ass to the meeting tomorrow because we we you know we talk
527. 1:32:48a little bit about the salon what we're about and then we um give every everyone like a minute to talk about who they are what they're up
528. 1:32:58to with AI and then something that they need from the community and something that they can offer the community and I don't give a [ __ ] if
529. 1:33:10you've never done AI in your life you've got something to offer to this community all right so go there and then I want to
530. 1:33:19show you one other thing I was talking about that little journey of like learning and then building and then leading and
531. 1:33:33teaching so we just came up with a brand new tagline for the salon And the tagline is play
532. 1:33:47first play first play first what do you mean don't don't I have to learn AI first and like no no because you can't learn
533. 1:33:59generative AI right now it's moving too fast so the first thing to do is play just play just play just go [ __ ] make
534. 1:34:07videos go [ __ ] make songs go make poems go make recipes doesn't [ __ ] matter play play play play Go build a
535. 1:34:17GPT and then you should read our our vision statement it's really good but this journey is what the salon's all about play first
536. 1:34:29mindfully create generously lead butow what does that mean am I doing this right now play first mindfully create
537. 1:34:46generously lead I'm now the chat GPT trainer of my colleagues in our new oh just said something unbelievable I just got some sort of
538. 1:34:59violation I don't [ __ ] know what it was I don't what what did I say I'm I'm so done with AI word word
539. 1:35:09monitoring oh good lord okay um in our AI pilot oh wait I'm now the chat GPT trainer of my colleagues in our AI pilot
540. 1:35:18so cuno cuno this is one of the things cuno very early on so she very early on played she very early on built and then
541. 1:35:27one of the things we talk about a lot in here she did this consistently is she just kept talking about Ai and she was
542. 1:35:34doing it like a year ago and what were people saying to her when she talked about it then uh yeah Chris could you
543. 1:35:39just not talk about that like I don't want to talk about the AI like whatever with your artificial intelligence you
544. 1:35:46weirdo I just I uh yeah whatever it's banned here so yeah I'm not going to go ahead and learn that and she just kept
545. 1:35:53talking about it and kept talking about it and then guess what happened when someone higher up said hey anybody know
546. 1:35:59anything about this here AI that that same annoying person was like well Chris knows a little bit about it she never
547. 1:36:05stops talking about it and then that higher up was like uh yeah hey Chris could we could you come what what's Ai and she told him like holy
548. 1:36:16[ __ ] we should put together a little group all right so she's on the group now she's teaching chat GPT to that
549. 1:36:25group The Pilot group there you go did she do that because she has a degree in mathematics from Stanford well
550. 1:36:34she might but I don't know but I don't think so you don't have to be a technological genius to be good at AI you just have to
551. 1:36:45be curious and so this this idea of of this journey to literacy play first joyfully exploring generative AI
552. 1:36:57gaining the knowledge and inspiration to mindfully create surprising ourselves in powerfully meeting challenges unleashing
553. 1:37:05us to generously lead others enriching lives redefining what's possible and unlocking a new era of human
554. 1:37:15achievement that's what the AI salon's about so get your ass in there you know read this stuff read what we're about
555. 1:37:22and then the first thing you should do go to introduce yourself it's the first link in the community section introduce
556. 1:37:28yourself go here and introduce yourself use your real picture use your real name and get your ass in this community
557. 1:37:38but Kyle I don't know anything about AI everyone will make fun of me no it's not a [ __ ] high school click I'm working my ass off to try to
558. 1:37:47make sure that the culture of this place is welcoming and empathetic and generous and [ __ ] smart and talented and it really
559. 1:38:02is if you go down beneath this little journey that I I talked about here's our values so the values are kind of a reflection of who's
560. 1:38:12already in the community curiosity we're eager to ask the obvious questions learn new things and experiment experiment experiment our
561. 1:38:21Relentless quest for adventure and exploration ignites our inner spark transforming curiosity into unexpected
562. 1:38:28discoveries and Newfound passions that's that's this group this group on a regular basis is like holy [ __ ] you're
563. 1:38:34not GNA believe what I figured out oh cool tell me amazing playfulness we value creativity joy and the freedom to experiment we first play
564. 1:38:48and transform learning into a captivating Journey we collaborate and mindfully create together when we can to
565. 1:38:55learn more make friends and develop new Circles of trust this new Circles of trust this is the future of work in my
566. 1:39:04opinion remember all work and no play is how we did things before the AI Salon ah look we're clever H empathy we're
567. 1:39:12respectful inclusive and kind we value diverse perspectives recognizing that our difference differences make us
568. 1:39:18stronger we accept that people might be afraid and angry and we work to understand them we actively practice
569. 1:39:25gratitude and acknowledge the efforts and successes of those around us humans matter imagine being in a group where that just
570. 1:39:39existed oh yeah good point Brandon this doesn't cost anything bravery it like it takes guts right now to step into I don't know what I'm
571. 1:39:51[ __ ] doing but I'm going to do it anyway Source Camp trying to figure out what she wanted to do next she's like [ __ ] it I'm
572. 1:40:00going to start an agency I'm going to start an AI agency she started a year ago nobody was talking about AI well
573. 1:40:05they were talking about it nobody was doing anything about it takes guts we have the courage to step into the unknown taking bold
574. 1:40:14actions in the face of fear and challenging ourselves to push the boundaries of what we thought possible
575. 1:40:20let's face it with every CH everything changing as fast as it is is we're all making this up as we go along might as
576. 1:40:26well be brave about it suck it up buttercup and then the last one generosity we are willing to share our resources feedback and wisdom with
577. 1:40:37others and support each other's growth and development there is so much happening and so much to learn we can't
578. 1:40:42possibly keep up with it all understand that even if something is obvious to you it might be a profound Revelation to
579. 1:40:48someone else Learn out Loud share what you know generously leave that's who this group is so you should be a part of this and
580. 1:41:02if you're a part of this you know just remind yourself if you haven't been playful if you haven't been curious if you haven't
581. 1:41:12been generously leading or learning out loud get back to that this is a practice this isn't a state we achieve
582. 1:41:19and we're done this is an ongoing rolling play first mindfully create generously lead just keep looping through
583. 1:41:29this cuno it's a lovely crew of humans it really is and so tomorrow night is a thing now here's a little bonus if
584. 1:41:37you're still here after listening to me ramble for all this time down at the bottom of all the oh by the way all
585. 1:41:44these different areas there's a there's a show Andel area there's like an ask for help area I maintained a list of
586. 1:41:51tools in here there's um guilds creative guilds business guilds Tech guilds and social impact guilds and then there's
587. 1:41:59clubs and we have a club just for us called AI Irregulars when chat GPT was down today everyone was in here going oh
588. 1:42:07my God chat gp's down what do I do what do I do what do I do why is it down is it coming back why is it down is it
589. 1:42:12coming back do you think they're giving us real-time voice what's happening but anyway oh happy Father's Day that's beautiful nice
590. 1:42:22Kate what's that deadlines for nominations is tomorrow for the women in AI Awards oh nice um so here's you know just all
591. 1:42:33sorts of things that you can if you want to make fun little images like Father's Day reunion that's awesome um start
592. 1:42:40using AI today it will be easier tomorrow nice um if you want to make images if you want to make songs and
593. 1:42:47share them in here with this group while these lives are still going on while they're live it's a lot of fun all right
594. 1:42:53right so that's AI salon so tomorrow night 5:00 p.m. to 700 p.m. mountain time is the AI Salon meet and greet get your ass
595. 1:43:04there say who you are say what you you need and say what you can offer and it's I don't know I I find the
596. 1:43:19meet and greets we used to do the meet and greets all the time and then the group got kind of too big and it like I I kind of felt like we
597. 1:43:29really lost something so we split the meetings into two so the first meeting of the month is we have a speaker in and
598. 1:43:36then the second one is just the meet and greet and so there might be people who are regular Salon members that are like
599. 1:43:42oh I don't want to do the meet and greet I don't want to you know I think that's a shitty attitude I I think every
600. 1:43:48opportunity you have to connect with Community is going to serve you in the in the next 5 years any opportunity you have to get
601. 1:43:59trusted and to establish yourself as a contributor as a hard worker as someone with Integrity as someone who can get
602. 1:44:09[ __ ] done as someone who's inventive as someone who's creative the next time someone needs to hire someone like that it just might be
603. 1:44:19you not because you have the skills to do whatever the thing is they need done but because you have the skills to
604. 1:44:27use AI because they know you're curious and they trust who you are and that's why you're going to get hired this is a big Iceberg flipping
605. 1:44:40moment and the blue ice is coming up isn't that beautiful when those things flip and the ice that was underneath is is all that aquamarine
606. 1:44:50blue that aqua blue when they they flip that's what we're about to enter we're social people too uh is it
607. 1:44:59in Denver in real life it is not it's just online um we've got local we've got local events in the off Tuesdays um
608. 1:45:09they're in a handful of cities right now they're they're a little disorganized from from us from the central group um
609. 1:45:16but but we're going to we're going to get our [ __ ] together with those all right I think that's it any questions any
610. 1:45:26thoughts before we head off into the evening we are made or built to be social social animals who crave
611. 1:45:39Community you know what's funny Gavin I think that I think that AI is counterintuitively going to force us not force us
612. 1:45:53Inspire us to reconnect maybe force us to reconnect Force us to ask for help Force us to get in community and in a weird way make us
613. 1:46:08better people like this AI stuff like we haven't scratched the surface of what it's about to do like we just haven't and even if the if the
614. 1:46:21calls for AGI don't happen this year and don't happen next year but they happen in the next three years AGI is
615. 1:46:27artificial general intelligence that's when the AIS are good enough to do most of our jobs as well or better than we
616. 1:46:33can not just contribute to our jobs do our jobs as good or better than we can like that's coming not in our lifetimes like in the next three
617. 1:46:45years when that happens Everything Changes and do you want to just be sitting on the outside like uh hey does anyone know
618. 1:46:57what's going on with AI I just our whole company just well three people run it now you don't want to be that person
619. 1:47:10standing out there in the desert you want to be in a group you want to be in a group that is doing
620. 1:47:27it all right eventually we'll have ai groups in different countries and eventually we can have annual events yep absolutely
621. 1:47:36actually we're talking about that Drew um we we've got some uh some offers of support for the salon that um could be really good so and
622. 1:47:48we're talking about doing exactly that I've seen some Google AI classes by corsera any feedback on General class or
623. 1:47:56should I just keep playing um listen I I I think any opportunity to learn is good what I would say about any
624. 1:48:04any course right now that's written down um is going to be out of date now there's there's two kinds of courses so let me let me
625. 1:48:18clarify there's the AI that um has been around for 30 years there's all sorts of courses on that if you want to learn
626. 1:48:26about machine learning and tensor flow and working on tpus and and creating you got out champy hang
627. 1:48:45on if you want to learn about all the math side of this [ __ ] the engineering side of this [ __ ] there's plenty of
628. 1:48:50courses on that there's free ones from Stanford and from MIT and all that sort of stuff um go go take all those those
629. 1:48:57are probably all good solid even though things are moving there it moves a bit slower on the underpinning
630. 1:49:03side the generative AI stuff that's the stuff where if someone's you know someone's got their prompt Mastery
631. 1:49:11course um prompt Mastery is not going to be a thing that's going to be needed for a lot of these tools um it doesn't hurt
632. 1:49:19to know it so you can take the courses and they'll probably help you short term but things are just moving so fast I
633. 1:49:26would say don't rely entirely on any courses like you should be playing and you should here's the thing where I
634. 1:49:33think you where I think everyone this is why we ended up going with this idea of play first I think it is really
635. 1:49:43important that you play with AI tools that are out of your comfort zone so if you're really shitty with writing
636. 1:49:54start writing with AI if you're really shitty at images or really shitty at music go make that stuff or if you don't
637. 1:50:01even even if you don't enjoy it uh M you know uh making images is really frustrating to me and I hate it okay go
638. 1:50:09start making images with mid Journey or dolly or ideogram start understanding that the [ __ ] you thought you were bad at
639. 1:50:24you're now actually really good at if you can sort of think of the idea and put it into the tool the tool is
640. 1:50:32going to generate something that's going to surprise the [ __ ] out of you you should have as many of those little
641. 1:50:38Revelations those Kevin McAllister moments in as many tool types of tools as you can and then in doing that what
642. 1:50:47you may discover is oh [ __ ] I really like this whole area over here that I never thought I would be good at but I
643. 1:50:53love it and I'm going to focus all my attention there you won't discover that if you're just like I'm going to take a
644. 1:50:58course on prompt engineering because I want to be a prompt engineer want to get a job again specialization is going to be
645. 1:51:08devalued generalization is going to be valued over the next 3 to five years right now it looks like this Specialists make more money
646. 1:51:21generalists are laughed at get out theater boy we don't take theatricals here this model is about to go and all these Harvard mbas are going
647. 1:51:35to be like ah hey there theater boy could you get me a job how do you do that oh I just look across all these
648. 1:51:42things and I'm like a conductor at an orchestra I never learned how to do that could you help me [Laughter]
649. 1:51:54I put my entire workplace and employee phone list and chat gpt's memory it's convenient oh that's super cool love that
650. 1:52:03nice my recent gtai classes were boring yeah and Murphy's in the house all right I gotta get out of here people anyway
651. 1:52:18um you know what to do like if you want to support the channel [ __ ] subscribe to [ __ ] or follow it or just keep
652. 1:52:24showing up here um thanks to regulars for all the the love and the Taps and all the stuff you do every night thank
653. 1:52:31you mods really appreciate what you do um Brandon and ion chy who came up and shared their stuff thanks for thanks for
654. 1:52:40popping up on stage that was cool those were good fun fun conversations um all right let's see who's in here tonight
655. 1:52:48call it a portfolio career or generalist equals highly sought after yeah yeah that's really interesting portfolio
656. 1:52:54career I like that Silver Fox thank you very much kind words appreciate that Danielle good night Tobias steo peace
657. 1:53:01out to you too Diana's in the house Papa JJ uh lunatick in the house Wolfman Clint thank you as always Lord digital Gods good day to you sir
658. 1:53:14good day a cricket a cricket a crickit a crik a crick KK you gotta I've talked to the janky master your name is janky you gotta fix
659. 1:53:30that [ __ ] name with some capital letters or something I can't read it it's the letters are too small and there's too many there's too many
660. 1:53:37possibilities phonetical sayings uh Brian Whitney good to see you an Murphy peace out Joey Brandon Mr It joker
661. 1:53:47cuno uh Jim Ross from Three Mile Three Mile Storage if you need to to manage your storage facility talk to Three Mile
662. 1:53:56Jim Ross is a blast at this [ __ ] he's really good at it Tom's in here um Drew learning AI coming in calling in
663. 1:54:07from Down Under along with steveo those two wacky kids they're always in here Drew is usually in here when he's mowing his
664. 1:54:15lawn he's like I don't know what you're doing in America I'm mowing my lawn right now I I think Drew M Lawns for a
665. 1:54:24living it's all he seems to [Laughter] do I just knocked my I just uh swiped up on my uh on my Tik Tok all right
666. 1:54:41everybody uh I'm gonna get out of here again tomorrow night 5 to7 mountain is the AI Salon meet and greet get your ass
667. 1:54:49in there um introduce yourself get the details for the meeting in the AI Salon um so introduce yourself in the website
668. 1:54:57but then come to the meeting tomorrow and introduce yourself start being part of community it's the future of work like I
669. 1:55:07know this now like I just this is this is not I I used to think like well here's a theory no like this is you have to be in trusted
670. 1:55:20circles have to be 700 p.m. I don't know what time zone you're in Mr it 5:00 p.m. 5:00 p.m. to 700 p.m. mountain time so 700 p.m.
671. 1:55:34Eastern 6 Central 5 Mountain 4 Pacific ah and then whatever in Europe and then other side of the world you're on your
672. 1:55:43[ __ ] own I don't know I had two I had two calls tonight one to Japan and one to Thailand I'm like right before right
673. 1:55:55before I went live all right I'm out of here everybody peace out uh AI learning lab late tomorrow possibly um yeah
674. 1:56:03probably I'll probably I'll probably start at 8:30 tomorrow night all right and but I'll put out a reminder all
675. 1:56:09right peace everybody have a good night hey we got 69 people in here huh know what I mean no all right