Feb 22, 2024 - AI Revolution: Exploring the Future of Generative Technology and Creativity

Transcript

1. 0:01[Music]
2. 0:16looking at the world telescope wish I wish I [Music]
3. 0:35looking at this blue telescope down a moon struck a road [Music]
4. 0:57tonight thank you Danielle thank you Robert [Music]
5. 1:28Rossy [Music] I didn't mean to cause you on a
6. 1:43[Music] sorrow didn't mean to cause you in a [Music] pain only one want time to see you
7. 1:58laughing thank you robt Ry only want to see a laugh in that bule rain purle rain purle
8. 2:12rain Purple Rain Purple [Music] Rain Purple Rain Purple [Music]
9. 2:28Rain want to see [Music]
10. 2:47in that's actually in that purple rain thing there's remember the opening of Twin [Music] Peaks huh that's the Twin Peaks opening
11. 3:01[Music]
12. 3:16[Music] [Laughter] [Music] welcome everybody my name is Kyle Shannon this is the AI learning lab the lead singer there is
13. 3:31[Music] Champ he likes to open up the shows [Music]
14. 3:57here well you done don't me you better I felt it tried to beat you after the M fell out through the cracks now I'm trying to get back before
15. 4:11the cool R out I'll be giving in my business nothing going to stop on the v w again my turn to win some or some but I
16. 4:26won't take no more no more it I'm [Music]
17. 4:49your all right enough mediocre [Music] singing I know I probably shouldn't call champ mediocre right I think if I put champ down people are
18. 5:03going to come after me we love champ why would you put him down he's a lovely singer he's clear he's clearly got classical
19. 5:15training I'll tell you one thing I do know champ he sings from the diaphragm like a good like a good singer should deep all right so
20. 5:31welcome to the AI learning lab if you have questions about AI pop them in the comments below they're okay I said it in January I said it a
21. 5:43lot in January shit's going to get weird in 20124 and then January was not weird it was just like yeah it's like chat GPT still kind
22. 5:53of sucks and Google hasn't launched their [ __ ] and then February came and it was still kind of nothing and like Google launched their
23. 6:05Gemini Ultra and it was like a piece of [ __ ] was like yeah whatever it's all right it's okay not great not what they
24. 6:13said and then in the past week holy [ __ ] you want to see something [ __ ] up well it's not [ __ ] up to you cuz
25. 6:24you're normal people but when you when you got the neuro Divergence when you're neuros spicy my own [Music]
26. 6:35self before I went live tonight I wrote down that many things this many things happened this week that I think are kind of batshit crazy and
27. 6:50big some are bigger than others and they're not really in order but I figure I figure we'll like we'll look at some of this [ __ ] tonight
28. 7:01so we had Sora right the open AI video model like oh [ __ ] I guess these things are doing video now like making like really
29. 7:13good nearly can't tell it from reality video I guess we're there now now it's not launched yet so we're not there there but we're
30. 7:26there 11 Labs then said oh we've got a um oh there's another one here wait let me put another one here then 11 Labs did the text to sound
31. 7:43effects thing where they put the audio behind the open AI Sora video which was really interesting then today uh 11 Labs has
32. 7:56been accepted into Disney's accelerator which is really interesting because if they're doing voice synthesis what does Disney have lots of
33. 8:06lots of intellectual property would say Disney characters and 11 Labs is also behind um oh someone said no for
34. 8:18what 11 Labs cloning your voice yeah exactly 11 Labs is also the voice Gemini is a train wreck I know it really is oh
35. 8:27they also you know how I was bitching about you can't get Gemini if you're if you have a work group account now you can so
36. 8:37either for $20 or $30 you can add Gemini to the thing you're already paying for and then at some point they say
37. 8:46we're going to give you access to Gemini [Laughter] Ultra so 11 Labs is now part of the Disney Disney accelerator like and 11
38. 9:02Labs is the one that sits behind Heen right the H Heen video translation stuff that's that's 11 Labs sitting behind that
39. 9:13then just because we [ __ ] all over Google let's unshit on them a week after they launch Gemini Ultra they announce and give limited
40. 9:28access to a thing called Gemini Pro 1.5 which I guess is their middle model but it's the 1.5 version of it and it's
41. 9:36got a million token context window which you're like I don't know what that means it means you can upload like six
42. 9:44complete white papers and it just will understand them all or you can up it's multimodal you can upload an hour of
43. 9:53video Ethan mullock today said he did a screen recording of himself reading a bunch of articles and uploaded that
44. 10:00video to to Gemini and it just understood all of what he did There's he he also uh uploaded him working and Gemini understood what he
45. 10:13was doing and suggested improvements like Ethan mullik so if you don't follow Ethan mullik he's a Wharton
46. 10:21business school Professor you should follow him he writes like I don't know once or twice a day for the past
47. 10:30three four days he's been writing like seven eight 10 times a day all about Gemini 1.5 and how a million token context
48. 10:39window changes everything everything's different this is a guy that's been paying attention to this [ __ ] every single day for
49. 10:49years he's like shit's different Deco here which is a tool that we've played with in here before that makes quick quick images it's using SD
50. 10:59XL turbo and they added a thing today called a reference photo which I'll show you later where you just upload a
51. 11:06picture of yourself and it's like instant AI of you it's like holy [ __ ] um Mid journey is apparently in talks with x.com
52. 11:21about getting getting hitched million million tokens is how many pot how many Harry Potter books I think it's probably all of
53. 11:35them I think it's probably all of them Nvidia announced their quarterly earnings and they dramatically exceeded the
54. 11:49market when did we get oh wait when did this start uh 8:30 I just started I'm just I'm just reading through a list of
55. 11:56[ __ ] that happened I haven't really started yet um the vision The Vision Pro goggles like people are just continue to lose
56. 12:04their mind over them like that that wasn't just a blip of Apple Fanboys going this is good there's been like a
57. 12:10sustained two we holy [ __ ] Vision Pro changes things and then you had Mark Zuckerberg in there going you know I
58. 12:18think that uh I think that the uh the the Oculus Rift is uh is every bit as good as the Vision Pro and it's 17th the
59. 12:27priz and you know I knew it would be better value I just didn't think it would be a better product but but it is
60. 12:33our product's so much better cuz it's made with plastic and there's has some sort of see-through glass and you know
61. 12:39it's it's just plastic so much better so that happened then grock not that grock not Elon Musk grock the grock with a Q not with a
62. 12:52K grock is a chip manufacturer and they manufacture some AI specific chips and you can go to gro.com gq.com
63. 13:03and go type a prompt in there go do it right now 500 tokens a second it's like 350 words a second it just goes like it
64. 13:13just like vomits out answers at you like bang you're like oh okay so not only are these tools going to be able to answer
65. 13:22anything they're going to be able to answer them like that and then right before the last thing I just remembered is uh
66. 13:30reddit reddit is announcing their IPO so holy [ __ ] and it was it was um Matt wolf wrote today he goes wasn't it last March
67. 13:43when like all of the companies launched all of the [ __ ] that blew all of our minds I have a feeling we're about to enter funny
68. 13:53season end of the first quarter shit's getting good but shit's getting weird and it's it's getting chaotic it's
69. 14:05getting really chaotic it going get more chaotic have you embraced the Jank have you embraced the Jank everything in AI right now is janky
70. 14:19learn to love it because it's not going to be here that long you can make money with we're going to look back at 2023
71. 14:25and go HH it was so sweet when the AI was slow and made mistakes and the the images had had personality cuz they were
72. 14:38janky it's going to get weird people it's going to get weird no I know I know it's 2024 20 I'm just saying 2023 was like normal 2023
73. 14:52the last normal year 2024 it started starting to accelerate so we'll go we'll go mid mid February is when the uh when the Rockets
74. 15:06kicked in my feet are cold my feet are cold Listen People my feet are cold I got to go deal with that okay okay I'll
75. 15:17be back I'm just going to go put on shoy shoes and then we'll be back and we'll do something big news today Google took down its own image
76. 15:23generator yeah I saw that did they actually do that also AT&T went down today and they're trying to figure out if it was a
77. 15:30hack cuz apparently well I can't say it on here some people have infiltrated our infrastructure more than the government
78. 15:39suspected let me get shoes hang on I'll be back talk about
79. 15:57yourselves
80. 16:27e I I'm afraid my rabbit will be [ __ ] by the time I get it I know me too my rabbit the rabbit phase one is supposed to come
81. 16:39out mid-march which probably means mid April their discord's been really slow too I actually bitched about it I went
82. 16:50in their Discord and I'm like does anyone that works at rabbit is anyone still posting in here like someone from rabbit was like oh
83. 16:57we're still here we're still [Laughter] here man all right what are we gonna do what do you
84. 17:12guys want to do what questions do you have let me let me let me see what's going on let me see what's going on in
85. 17:17here are we full of trolls tonight are we full of full of lovers we got empathy tonight oh we got hate and rage and
86. 17:27bitterness which one is it What's it gonna be tonight embrace the Jank um is the Google is the Google thing true did Google really pull down
87. 17:41their image generator let me go to let me go to Gemini Ultra because I have access to ultra
88. 17:57Gemini yeah see yeah hello Kyle how can I help you today make a picture of janky AI
89. 18:16[Music] technology doing something I got twirling Stars it did it all right let me go to non
90. 18:32Ultra let me go to Just Gemini little old Gemini and this isn't 1.5 by the way whoever told me that last night was
91. 18:40lying lying liing a cheat uh make an image of janky AI
92. 18:57technology I'm still learning to create images so I can't help you with that yet I'm sorry Dave I'm afraid I can't do
93. 19:06that all right if we're good at chat GPT do we really need things like jasper. a that is an excellent question Teresa um technically
94. 19:19no Jasper AI um copy AI tools like that are basically just pre-written prompts sitting on top of chat GPT or on
95. 19:32top of GPT 4 or three whatever they're using probably four if they're doing copy um you could technically go do all that [ __ ] yourself
96. 19:43right just learn learn to prompt the the problem with sites like Jasper and copy AI is that they start
97. 19:52out with like oh you know we should do 10 things like okay we'll do a short paragraph blog post and then we'll do a
98. 19:58tweet and then we'll do this and then we'll do that and then before they know it they've got 8,000 buttons and options
99. 20:04and things that are all prompts and sub prompts and things like that and it's like it takes you as long to learn the
100. 20:12interface as it would just to go to chat GPT and ask it to write you that [ __ ] so you're absolutely correct ah LinkedIn
101. 20:18how do I get to your office hours okay so tomorrow if you want to hang out in a in a more like just hangy out session I do
102. 20:28a thing called AI office hours at 11:00 a.m. mountain is that right yeah 10 a.m. Pacific 1 p.m. Eastern on
103. 20:40LinkedIn and so how you get to it is you go to my you go to my profile so I'm at Kyle Shannon on
104. 20:56LinkedIn and then are there
105. 21:11events huh resources show all five resources I don't know where events are on LinkedIn
106. 21:27like when I go to my profile hang on so if I just go to LinkedIn just look up AI office hours
107. 21:42and Kyle Shannon you should find it and I don't have tomorrow's up I'll put it up in the morning but but you should be able to
108. 21:53see so here uh oh here's the event link where can I put this I could put this oh you know where I'll put this I'll put this
109. 22:10in the Irregulars channel in the AI salon and then let's see I'll put it in the chat all right so
110. 22:27tomorrow AI uh office hours on LinkedIn 11 a.m.
111. 22:43MST and then I'll put the uh the that thing there you go so it's there there listen there are much more
112. 22:57professional ways to do that like I could have that thing up for more than an hour in the morning on a Friday but and welcome to Chad
113. 23:09ad I got organizational skills can I sell you some organizational skills you can make money with Chi I'll teach you
114. 23:17how to organize a chat chippy you just give me uh uh $49.95 402 79.95 for the first two hour one hour two hours
115. 23:31yeah yeah you understand what I'm saying all right a friend called and said he finally started using chat after months
116. 23:40of me prompting no pun intended he's hooked every [ __ ] skeptic I know every [ __ ] skeptic I know with the
117. 23:50exception of Eric marier he's sticking to his guns cuz he's using chat GPT and he still hates it I still am not
118. 23:56convinced he's using it right though he's bitter about it and I like that I appreciate that in him at least he's
119. 24:03[ __ ] using it right um every skeptic I know when they actually use this [ __ ] they're like oh that's better than I thought it would
120. 24:13[Laughter] be hey Kyle you know um I know you've been talking about this AI stuff for for a little bit lately I are you on Tik Tok
121. 24:25anyway so listen I was on have you used the chat GPT yeah yeah so I went on the chat GPT and I thought it was just like
122. 24:34Google right and then I um I did like a search and it like it said I don't have that latest information it was like old
123. 24:42and I was like well that's kind of useless and and then I asked it to make me a business plan and so anyway so I
124. 24:48just launched a new business I didn't know I could do that every [ __ ] one why do they always talk like that
125. 25:03too I'm a skeptic and so this is how I talk this is just how I talk I don't know what's going on with my mouth this
126. 25:11is just how God gave it to me you know bless him what is with those people figure I'd give chat jbt a try
127. 25:26and uh hey did you know that it makes pictures no but you know what's funny is I I put in uh I put in a it's a the John
128. 25:36Deere lawn mower the push mower the the green one and I I said wouldn't it be hilarious if it if it had a picture of a
129. 25:44a a John Deere uh push mower on a on a lawn and sure enough it had a picture right in there like I don't how many
130. 25:52pictures are in there it's a lot right that was that was that was really something what was the AI that came out today that
131. 26:01said to that you upload a photo to turn it into AI it's on your list Deco here so let's go play with that we'll get we'll get back to the
132. 26:10Skeptics because they by the way if you're in here right now right now I I see you if you're in here and you're like if you're one of
133. 26:21these sitting on the sidelines like well I don't know if you know this but it hallucinates you know it's the world's
134. 26:29greatest plagiarism engine if you're if you're sitting on the sidelines and you haven't [ __ ] tried it go right
135. 26:37now is he always this aggressive he seems Mighty aggressive go right now to chat. open.com just do it oh yeah I'll get
136. 26:49through it tomorrow no now now now now no do it now chat open.com and just start playing with it
137. 27:04while you're listening to me ramble and show off this Deco here [ __ ] go ask it to write you a poem ask it to write you
138. 27:09a song ask it to write you a business plan ask it to come up with a new pizza recipe ask it to write some code for you just go watch what it
139. 27:19does and then and then when you're like oh that's cute push past the cute and take something you know like I don't
140. 27:27know maybe you're really into I don't know uh uh 15th century literature ask it about some obscure [ __ ] from your literature degree and
141. 27:38watch what it does push it push it push it go now go all right all right I'm not putting up with
142. 27:54it I'm not putting up with You Nero well SK ICS who are sitting on the sideline if you want to be against AI
143. 28:02after having tried it I'm happy to have that conversation but if you're just over there listening to what you heard on the New York [ __ ]
144. 28:11times all right sorry you asked Jim right all right that had nothing to do with Jim by the way all right where are we going
145. 28:23we're going to go to app. deoh here.com uh uh uh uh r. dcoh here.com and we're going to sign in with
146. 28:39the Google single sign on Google made it so easy for us to all just sign on they're definitely not keeping our data
147. 28:47of all the apps we sign on with definitely not they take privacy really seriously okay so how we going to do this
148. 29:04uh okay so this is it so if you haven't seen it before you can just go in here and you can type in a large come on
149. 29:22semi truck in pink polka dots zooming down
150. 29:38the highway in New Jersey at dusk all right so if you haven't seen
151. 29:55anything built on top of sdxl Turbo that's what this is sdxl Turbo just like makes [ __ ] in real time as you're typing
152. 30:04it it's Bonkers it's crazy and then there's like a whole bunch of pictures here and then you can take any one of
153. 30:10those pictures I don't know why I'm talking like this all of a sudden GE williker Mr G Willer Mr McGill
154. 30:22Cy I don't think I've had enough water today the Ye Old vocal cords are season up on [Laughter] me all right I can take this thing and I
155. 30:35can click this little button right here and just say movie and it's wait oh yeah it's queued up okay so it's going to
156. 30:43make a movie out of that image so that's the lower thing down here okay so that's all cool right but they've added this little button right
157. 30:55here called use reference person and so oh I don't know let's take someone like o Taylor Swift and so it's now got the reference
158. 31:12in there and so I can say um woman sitting outside uh oops woman sitting
159. 31:26outside a chrome Diner in the
160. 31:411950s I mean holy [ __ ] I mean holy [ __ ] now okay couple of things that I've
161. 31:55discovered that are weird about it I can't say I can't say like closeup or like far away shot like we can't see like her
162. 32:08standing in front of the diner we can only see from like the shoulders up cuz that's what the reference photo is
163. 32:16but holy good Lord right and now let me change out that reference photo for who's that Brie Larson yeah so now it's Brie Larsson in
164. 32:29front of Jan 50s diner that actually kind of looks like her and then you're like oh well they
165. 32:44got all those pictures in there already that's how it can do that Kyle yeah I don't know if you know this but they
166. 32:50they put they put like you know they create a whole database of pictures and then they're just pulling it up there
167. 32:55and they're copy and pasting her face in front front of it no you Dum dumb it's not like that so we'll go get I I actually paid for a $9
168. 33:10um dreams I paid for a $9 a month subscription for that stupid thing so I could show y all this I invest in you people I do it I do
169. 33:25it I do it all right that's not what I was looking for dang it um contains dreams I should have Kyle Shannon dreams in here somewhere
170. 33:40[Music] no Kyle Shannon [Music]
171. 33:56dreams oh there look we'll do serious me all right there's serious me all
172. 34:15right let's say um let's say um okay that's just rude rude that is
173. 34:29just rude that okay we'll do this oh I got it I got it I got it okay um Sears carpet
174. 34:45salesmen from the 1970s
175. 35:03[Laughter] oh hey Tom lungren here can I interest you some in some in some high pile listen I'll tell you what the new
176. 35:18thing's going to be the new thing's going to be tiled carpets yeah you don't buy it in rolls anymore you buy it in
177. 35:25squares and I'll tell you what the new stuff it's got adhesive on the back anyway here's my card Tom lungren give me a
178. 35:38ring uh let's see um so let's see um JC Penny um JC Penney yeah salesman yeah salesman's good uh let's see the
179. 35:54salesman we'll say uh vice president from the 1970s what what was why was there a gun there in the
180. 36:09corporate headquarters all right let's see
181. 36:26um tattoo artist in grenwich Village New York City in the
182. 36:5680s
183. 37:10listen I'd like something Little Lacy La see across the chest so I just shaved I you know I shaved down to my nipples I
184. 37:17don't you know I don't like to use that term but I don't want to be forward anyway uh so I'm clean you know just
185. 37:23just down to sort of you know water level let's call it water level okay so anyway so what I'm thinking is I would
186. 37:29love just a little lace if I could have like a not so much a necklace just kind of like the like a like a you know
187. 37:36whatever you would call that right up through [Laughter] here let's see from the 80s uh with a
188. 37:51tribal tattoo on his face come on let's get some face tattoos going listen I was
189. 38:13want look at the hair you got to show your wife I think my wife has seen enough she's seen me with this hair
190. 38:32oh good lord all right let's see um let's say uh motor
191. 38:48cycle gang member in southern la in 1999 on a
192. 39:05big Harley yeah yeah we call him baby face you don't [ __ ] with baby face yeah he looks sweet
193. 39:20enough he all right he's all right I just wouldn't I wouldn't [ __ ] with him I mean look at his hair anyone that goes around with that you
194. 39:30know they're willing to take you on and they're going to [ __ ] take you down he's not going to risk messing that hair
195. 39:36up you know what I'm saying you don't [ __ ] with baby face should have invested in Invidia honest to God I like I was doing this I
196. 39:47started the AI Salon last [ __ ] January Nvidia was like what was it 100 bucks or 70 bucks now it's 760
197. 39:58$760 it's like I think I think nvidia's market cap just exceeded Canada's GDP or some [ __ ] like that it's like good
198. 40:13Lord oh man man man man man okay so anyway that's Deco here's oh let's let's look at the the video that it made for us so here's the pink
199. 40:26truck yeah see yeah the vi the these videos like the the uh the stability AI videos right now are kind of
200. 40:36shitty they're not great but I did a cool thing I took this picture this was one I did earlier and I threw I took the picture
201. 40:46so this was the picture that Deco here made out of my photo earlier and then I took it into chat Je
202. 41:07I took it in the chat [ __ ] and it made that image oh yeah [Applause]
203. 41:22yeah it's electric it's middleaged B how you doing I told it to give me salt and pepper hair and it made it like some rad
204. 41:36Punk thing Kyle you never told chat what the app was yeah I did Deco yeah app. deoh here. this this other app that we were
205. 41:56doing this one this is Deco here app a. deoh here d e c o h e r e. a Mr K what's happening look Mr K you
206. 42:13too can be a biker let's see pop art post
207. 42:30with blonde spiky hair no that [Laughter]
208. 42:51what yeah that's good that's I think that's appropriate [Laughter]
209. 43:10oh so this is one of the things on the list let me go back to the list it's crazy right and like like this is like what
210. 43:22what's what's freaking me out cuz one of the things when I first started doing Kyle Shannon dreams so like fall of 22 fall of
211. 43:3522 I started making these images and people were like you know how I just want to upload a picture of myself and
212. 43:41like make a thing like well it's not that easy and you got to make a model and you got to do this and you got to go
213. 43:47to dream booth and do that and and then there were these apps that came out and it's like well you got to pay the $9.95
214. 43:54a month or the1 1995 a month but then make sure you cancel cuz just and so now it's just like yeah you
215. 44:02just put your picture there and you just make yourself into some androgynous 80s pop star boom done now does it look exactly like me no
216. 44:15but it's enough to make me laugh like it's it's close enough it's in the hood crazy right let's turn Joe Rogan into an 80s pop
217. 44:31star shall we Hey Joe Rogan look you think you're so Macho look at you now with your pop background come on come
218. 44:50on he looks like dairo and taxi driver I'm a sunflower hey it's Joe Rogan sunflower
219. 45:04welcome to 80s pop oh yeah you go Joe Rogan you're so Butch love you Deco here has a few guard rails compared to the
220. 45:18other tools which is ref oh has few guard rails yeah and in fact you can turn them off the safety filter you can turn off
221. 45:27if you if you subscribe to it which is pretty smart this is actually pretty smart so so this tool by the way Deco
222. 45:34here is just an application it's like a website someone built on top of stable diffusion tools okay so just like Jasper
223. 45:42was sitting on top of chat GPT and I'm like it's not really worth playing with some of these tools built on top of
224. 45:47stable diffusion are pretty slick because they take all the complexity away right so it might be worth paying a
225. 45:52month or two on this just to play with it but you could go do all this [ __ ] in stable diffusion if you want to geek out
226. 45:58you can go do all this stuff for free um but anyway they charge you you can see that you can turn water marks off
227. 46:08and safety filters off only if you subscribe which I thought was pretty slick Hey Joe Rogan you're looking fresh hey thanks that's
228. 46:25awesome okay I have an idea let's go find a funny one now that's too real well that's not bad all right that's good so
229. 46:36I'm going to turn motion up to 10 and we're going to turn this into a video let's let's see how wacky it makes Joe
230. 46:46Rogan what did the New York Times say I don't know about what what's the name of the AI please tell us what the name of the AI is Kyle
231. 46:57I want to know what the name of the AI is is this the one is this the one is this it how do you get there it's pinned
232. 47:05it's pinned it's pinned it's not pinned now deoh here. a the hair is over 9,000 uh looks like Billy Idol Deco here my camping neighbors must
233. 47:21think I'm mental all they can hear is me laughing oh look we got a we got a Joe Rogan movie come on please not be boring that's pretty
234. 47:33[Laughter] cool oh you know what I'm gonna do let's download this uh download did I download
235. 47:49it yeah it did okay we're going to go to x and we're g to I love doing stupid [ __ ] like tagging Sam Alman which by the
236. 47:59way he never got back to me and he never rendered my movie Butthead so I'm GNA go at Joe Rogan at Joe
237. 48:14Rogan um hey man let's see you were so cool in the
238. 48:3280s [Laughter] man and then let's upload our little
239. 48:52movie all right so do me a favor go to my Twitter go to my ex which is it's at Kyle Shannon and all right I just tagged Joe
240. 49:02Rogan so go boost the [ __ ] out of that thing let's see if we can get Joe Rogan to to [Laughter]
241. 49:17respond oh Lordy all right that's some good fun right there I'll tell you what mm that's some good times you go out
242. 49:25there on the internet and you play play with some of them their AI tools and you make some funny picture like Joe Rogan
243. 49:30with a tie-dye background you put it out there on the X they call it X now they used to call it Twitter anyway you you
244. 49:37put it on x.com and then you just say to Joe Rogan Hey Joe Rogan look at you in the 1980s it's so funny that's you got any more
245. 49:55Twinkies did he take his medication [Music] today all right what do you want to do what do you want to do Chrome gets built-in AI rating
246. 50:10writing tool powered by Gemini really oh I think I saw that I was in something yeah it was like three little
247. 50:25options came up I think I saw that that little writing tool inside Chrome yeah this stuff's going to be everywhere Adobe brings oh
248. 50:33conversational AI to trillions of PDFs with a new AI assistant and reader and acrobat I haven't played with that yet
249. 50:42I'm I really want to play with the new um Gemini 1.5 but I assume that they're not releasing it because the million token
250. 50:52context window is probably expensive is is my guess Kyle didn't take his medications yeah but look I did do my
251. 51:12hair put my head in a blender did my hair can we write some classical piano music we could do that you could do it at
252. 51:28refusion but I think refusion now I don't know if refusion I don't you know what I'm not good enough at the music
253. 51:34tools to do it in a way can you explain stable diffusion and how it works I can I can explain how diffusion models work
254. 51:44Yes stable diffusion is just one of many tools that are using the diffusion method or model or whatever the [ __ ] they call it um
255. 51:57I have a slide that explains how that works if you want to know I think I'm GNA have to pull something down from yi yi
256. 52:13iCloud but let me do that and I'll yeah I'll I'll edim M you on how how these image generation tools work when
257. 52:22you okay so one of the things that is um is a misconception of AI right now is that it copy and pastes from The
258. 52:35Source material it does not the source material ceases to exist once it's trained and embedded um in the case of text
259. 52:50documents they all turn into these things called vectors which is just a long string of numbers numbers that
260. 52:57just live in Thousand dimensional mathematical space I'm not that sounds like I'm just making [ __ ]
261. 53:05up that's what it is you have a point in Thousand dimensional mathematical space and when you type a prompt the mathematics of the large
262. 53:17language model says based on that prompt there's a probability somewhere in this massive Vector space they call it this
263. 53:26thous dimensional space where one of those points is has the highest probability of being the right thing and
264. 53:35then that point in space is associated with a word and then it puts that word in and then it does it again and then it
265. 53:44does it again and again and again so when it spits out all that text it's generating it token what what
266. 53:51they call token by token but effectively word by word it's writing original content content the image tools are similarly funky and
267. 54:04weird and they're not copying and pasting so all right this is how diffusion images are made so see that pretty picture let's
268. 54:13say that pretty picture is in the training set so so when you use an image generation tool you've got two models that are
269. 54:24interacting you've got the large language model which understands the words and then you've got the diffusion
270. 54:29model which which has got some data about images but what happens is you would take this image for example and
271. 54:37you would tag it right so human beings or maybe even computer systems using AI would write a bunch of tags for this so
272. 54:44they would say I don't know anime you know cartoon illustration whatever those trees are cherry blossoms or whatever the heck
273. 54:54they are um Cliff Cottage House Rocks garden lawn you know bushes whatever uh Sky clouds Mist Haze right
274. 55:09so there'd be a whole bunch of tags and then this thing gets embedded or this thing gets noised so they they noise it so okay so so that's
275. 55:21so you're uploading a whole bunch of images into the data set that have been tagged and then the way these things work is
276. 55:33this you you have the large language model that understands the words and then you have the image model that understands
277. 55:41images the so see how that thing just went noisy where it was an image before and now it's been noised into just sort
278. 55:48of this random noisy image so the labeled images have noise added and are vectorized so that they're basically turned into number into
279. 55:58numbers and then sorry no um and then based on your prompt the large language model interprets and filters the relevant image vectors now
280. 56:12that doesn't mean [ __ ] but here's what happens so this is a very simplified Vector space this is a two-dimensional
281. 56:19Vector Space X and Y right and then these little clusters you have you know related little clusters of points in
282. 56:28Vector space that are tied to the tags of those images right and so when you type in your prompt which in this case is Cat on a
283. 56:39desk the large language model sort of goes through its Vector space and it says there's a bunch of these things
284. 56:45that are not relevant but these pink ones you know are are all sort of cats and these uh blue ones are all sort of
285. 56:55desks right they've been tagged and then associated with all of those points in Vector space are a bunch of these noised
286. 57:02images that you can't see anything on and then you know somehow mathematically they take all the [ __ ] that is high
287. 57:10probability of these noised images that it it knows nothing about them other than there was some association with
288. 57:17some word and there's some probability and it turns it into a like a a a like an average of all of these IM of all of these noisy images
289. 57:28here's this noisy image and then what the system does is it takes the same algorithm that turned them into noise in
290. 57:36the first place and it denoises this averaged image oops what did I do oh so it denises it and it gets a little
291. 57:49less noisy and a little less noisy and then in the end you have a cat on a desk so it has no idea what's in the
292. 57:59image so this is why early on because it wasn't trained right you get hands that were like this cuz when it averaged the
293. 58:07images together maybe it brought hands together in a weird way I don't know how they fixed that [ __ ] but
294. 58:14like it's all just math it's all points in Vector space and it's these these sort of you know messes of images that get
295. 58:23averaged into this one image and that's why you sometimes get things that look normal but not quite like if you ask for
296. 58:30a 70s muscle car it'll look sort of like a Mopar car and then it's kind of got a Chevy chevel kind of feel to it or or an
297. 58:39AMC Javelin kind of feel to it so that's how they work isn't that crazy and what's even crazier when you're doing something like Deco
298. 58:51here where's Deco here there Joe Joe's so cute let's get rid of Joe and we'll start over
299. 59:05here and we'll go um ancient teapot on a museum pedestal all that [ __ ] I was just
300. 59:21talking about is happening in real time with this model [Laughter] so so like there's just like [ __ ]
301. 59:32millions of images that have been noised and in real time it's averaging them and denoising them in like half a second or
302. 59:39I what however long it takes not long so like these are not pictures that exist but these are amalgamations of
303. 59:53some elements of pictures that exist but like the the the actual images cease to exist in the in the space in the
304. 1:00:02latent space or in the diffusion space I don't know what they call it with images isn't that crazy it's just nutty it just nutty it
305. 1:00:15just nutty now I don't know you know that the AI they put them in the they put it in there they got the they used the
306. 1:00:22GPU so so they used the GPU the GPU it stands for graphics PR setting unit in the in the in the gamers The Gamers The
307. 1:00:32Gamers used to buy the GPU to make the game make the game go faster they they wanted to 60 Herz and then they wanted
308. 1:00:39to to 90 Herz and then they wanted 120 H Hertz so they can put hurt on the enemy right and that's so they did GPU and
309. 1:00:47then all the other chip manufacturers are like yeah you can go Nvidia you just go make them gpus for all them snotting
310. 1:00:53those little teenagers that's good that's fine everything's fine you can go do that we're not going to make them why
311. 1:00:58don't you go make the gpus we'll make the CPUs in the RAM and then you make the gpus and then everybody good right
312. 1:01:05you good we're good and and then and then the crypto came crypto came and they were they were making blockchain
313. 1:01:11they were they were doing the the the Bitcoin and the nfts and they and they use the gpus and all the all the chip manufacturers like
314. 1:01:19yeah you just Nvidia go go make them them them chips for the blockchain that you that's good the gamer don't want it
315. 1:01:26anymore but the blockchain people do yeah just go make that now and then and then all of a sudden the AI came out and
316. 1:01:33then never was like yeah Nvidia just just go make them AI chips and and we're good we're not going to make them and then and then chat gbt
317. 1:01:43came out and 100 million people used it and all the other people were like holy [ __ ] we got to make gpus what do we do it was too
318. 1:01:54late Nvidia has the market cornered cu no one gave a [ __ ] about it apparently [ __ ]
319. 1:02:08crazy I don't know why my Billy Bob character there knew so much about chip manufacturing but you know stranger
320. 1:02:16things have happened the world's going to get weird people are going to have all sorts of knowledge you don't think
321. 1:02:20they should have access to and seen and seen thank you Lord digital Gods was that was that payment for my [Laughter]
322. 1:02:36performance all right let's go do some more show and tell how do you get over the criticism that everyone hates you
323. 1:02:43well that's not what she said how do you get over the criticism that training sets are mass theft I can never reconcile that um
324. 1:03:01it's not an easy answer if it were if it were actually sampling I I would
325. 1:03:15say um you know if if it were actually like like sampling in the 80s and 90s with hip-hop they were actually chopping
326. 1:03:24chunks out of song and putting them in other songs right so they're they're actually doing that that's not first of all that's not
327. 1:03:32how these tools work so part of it is just there's an assumption on the part of the artist whose Works have been used for
328. 1:03:44training that these tools work a way they don't now why it's gray and weird is you can prompt large language models and you can
329. 1:03:55prompt image generation tools to get really [ __ ] close to someone's work and you know turns out that Ye Old mid
330. 1:04:03Journey they don't have any [ __ ] Scruples at all and you know there was a document that was leaked from mid
331. 1:04:09Journey that said here's all the artists that we explicitly trained trained on their data and they put those names in
332. 1:04:14the model so that you say when you say in the style of some artist it looks just like that artist right because mid
333. 1:04:21journey is just basically run by a cowboy that's like [ __ ] it let's go woo W Adobe Firefly on the other hand was
334. 1:04:31not trained on copywritten images so one of your counters can be if you're really concerned about that go use Adobe
335. 1:04:40Firefly because it was only trained on open source or public domain images and images that adobe already bought the
336. 1:04:48rights to or owned the rights to and then Adobe is also allowing artists to opt out of being trained opt into being
337. 1:04:55trained trained and they're developing um attribution and compensation models I don't know where they are on that but
338. 1:05:01Adobe is at least being conscious of this mid journey is like [ __ ] it which is why mid journey is better than everything
339. 1:05:10else because they just said screw it now here's here's how I think about it from a conceptual standpoint the way
340. 1:05:23these models work it's actually a lot closer to if you go to college and all you do is study pop art you love pop art you
341. 1:05:33love Andy Warhol you love lickstein you love you you like all Keith Herring you like all the pop art you know artists of
342. 1:05:41the 80s and you study everything about them and just in your head are going images and and words and how they think
343. 1:05:49about things and their poetry and who were the Poets of the time and they were listening to Velvet Underground and you
344. 1:05:55got Lou re and Lori Anderson and all that shit's going in your head and then you you leave college and you're like I'm going to
345. 1:06:03make a masterpiece and the first piece of art you do looks just like Keith Herring or it looks just like Andy Warhol
346. 1:06:13well is that their work are you going to get sued for that well the Keith Herring estate sues the [ __ ] out of people that make fat marker
347. 1:06:26you know people that look like Keith Herring drew them so the Keith Herring estate you know they've put a flag in
348. 1:06:34their ground and like we'll sue you if it looks anything like it but you're you've been inspired by those guys and you're doing a kind of
349. 1:06:44work that that is you're the one coming up with the idea you're the one executing it and you might have been inspired by them so it's
350. 1:06:55more like like that and most of the times what's happening when I see people doing things in AI they're not going I
351. 1:07:02want something that looks like Andy Warhol and I want it like that now you can do that it's really [ __ ] lazy the
352. 1:07:08artists that I know that are doing this are like okay I want something that kind of has this vibe to it no I want it more
353. 1:07:14like this I want it more like that if you watch Corey Sandler her art she'll go through 30 40 70 100 different
354. 1:07:23variations of an image before she gets to one where she's like o that's what that's the feeling I'm going for and then she pulls that one
355. 1:07:33out and then the other place I go is so so that's kind of it conceptually and then the other place I go is
356. 1:07:44this machine learn the the p in GPT stands for pre-trained and what open AI did for chat gbt is they said let me go get as
357. 1:07:54much of the public internet as I can and let me train this thing on all that data out there so they're they're they're
358. 1:08:04effectively performing the function of a library or a museum they're collecting the output of humanity that was put in this particular
359. 1:08:16form called the internet and they're putting it in the digital equivalent of a museum right of a library so they said
360. 1:08:30okay as much of the internet as we could get we're going to get it we're going to jam it in here if you carry that to its logical
361. 1:08:42extension you take all of the output of humanity and you make it accessible to the rest of humanity there is a much
362. 1:08:53bigger thing at play here than individual contributors going you owe me a millionth of a penny every time an
363. 1:09:00image might slightly look like my myew glasses from my famous photo right I love your lives you make AI fun thank you Suzy AI ninja
364. 1:09:14Mastermind all it's Susie AI ninja Mastermind how can I help you welcome to the party kids so for me
365. 1:09:30the you should never train on someone else's art is kind of like saying you should never put [ __ ] in the library and you
366. 1:09:40should never put [ __ ] in museums because only the people that created them or have the rights to them
367. 1:09:47should ever be able to do anything with them there's also there's a really interesting response when when Sam Alman
368. 1:09:53was asked about the New York times lawsuit in Davos they're suing him because what they're claiming is the
369. 1:10:05the the data set was trained on New York Times data because they can do a prompt in a certain way that will pull up a
370. 1:10:14direct quote from a New York Times piece so Sam Alman had a couple couple of interesting points the first one was
371. 1:10:26we don't need the New York Times for training data like we've we've got enough and the analogy he used was if you want to get like a high
372. 1:10:34school biology education and you've read 2,000 books on biology does the 2001st book really make a difference no it
373. 1:10:48doesn't what he said was the reason they they were in negotiations with the New York Times to cut a publishing deal
374. 1:10:55was on the output side so that when somebody did a search on chat GPT and it came up with an answer like
375. 1:11:04they could actually reference New York Times data and say hey here's an actual reference for the New York Times click
376. 1:11:10here to go there they were trying to cut an attribution deal not on the training data but on the output side the other thing he said
377. 1:11:21is the New York Times publishes digital data and so when someone's got a subscription to the New York Times and
378. 1:11:31there's a really good article there sometimes they'll copy and paste that and stick it on LinkedIn or stick it in
379. 1:11:38a PDF or stick parts of it in their in their article so while open AI could easily say don't train on New York
380. 1:11:50times.com it's effectively impossible for them to say find every in of every mention of the New York Times at every
381. 1:11:59location on the internet for all of time could they do it well they could but you know they go goes back to
382. 1:12:09the library thing should no New York Times ever be in a library should you never be able to go back and look at
383. 1:12:15those why do libraries exist for the good of humanity so if you play out the open AI data training set to logical extension
384. 1:12:26you put all of the data from all of humanity in It Mary Mary thank you Kyle this is so common sense to me as well now
385. 1:12:39listen I don't think it's Crystal Clear but I think it's impossible it's an impossible solution
386. 1:12:54right here here's here's the solution as I see it right now everyone's chasing the outputs everyone's saying well if I put
387. 1:13:05out a song and you copi my song I'm going to sue you because that's copyright we're moving into a world of infinite infinite content
388. 1:13:21generation so how many millions of fans does TW Taylor Swift have 100 million a billion I let's say let's say a
389. 1:13:29billion so how many millions of fans does TW Taylor Swift have 100 million a billion I let's say let's say a
390. 1:13:38billion that talk about my dog named Sam and my sister named Lily and the Little Creek that we go swimming in and a song will start
391. 1:13:53playing that is Taylor Swift singing that song and at some point I promise you one of those billions of people that
392. 1:14:06writes one of those trillions of Taylor Swift songs that song is going to become a hit and then what happens does Taylor
393. 1:14:16Swift sue her fans or partner with them right so I think it's ridiculous even even in a three-year window to go after the outputs because
394. 1:14:33the outputs are gone like copyright as a mechanism was based on the fact that it used to be hard to replicate
395. 1:14:43content and then the internet came and digital music came and they sort of finagled some [ __ ] Byzantine laws about sample sizes and if
396. 1:14:54you use less than 11 seconds and it's common like all that [ __ ] to work around the fact that like it's just
397. 1:15:03infinitely easy to copy and distribute Digital Data well now we're moving to the next evolution of this which is it's
398. 1:15:13infinitely easy to generate original content that sounds like whatever the [ __ ] we want it to sound like
399. 1:15:20so here's where I think the solution goes and YouTube has already teased this so YouTube about a month ago put out a
400. 1:15:29thing I forget what they called it it was some sort of audio thing where if you want original music behind your YouTube
401. 1:15:39short they cut a deal with Charlie poth and Tay and you can go I have a video about me ranting about copyright issues could
402. 1:15:50you please have Charlie poth sing my background music and it makes a song written in the style of Charlie poth with the voice of
403. 1:16:00Charlie poth about your little topic and they let you put that song behind your your YouTube video and play it for everyone to
404. 1:16:11hear they're paying him on the generation Tay and Charlie poth have an original model that is licensed to YouTube on the
405. 1:16:23generation [ __ ] them let them make all they want just pay us every time they push the button every time you go make
406. 1:16:31me a Charlie poth song send me a nickel Charlie po song nickel Charlie poo song nickel and then if that song becomes a
407. 1:16:40hit [ __ ] it that was cool I got my nickels Sora red team memb said that it takes one hour to render one minute and
408. 1:16:49said 90 hours for a 90-minute movie yeah that'll get faster Tak one hour to render I figured that would be the case
409. 1:16:58because when Sam wman was saying hey you know send me your ideas I'll render your videos all of the videos that Sam Alman
410. 1:17:05was doing were 10 seconds long so if it takes an hour to do a uh to do a a 60c clip yeah so so it's 10 minutes for 10 seconds so it's it's a
411. 1:17:18sec it's it's a minute it's a minute per second right that's what it sounds like so that will get faster but that doesn't
412. 1:17:25surprise me at all like it doesn't surprise me at all and that's why that's also why I think we're probably not
413. 1:17:30going to see Sora for two three months maybe even more but they're putting it out there to let us know what's
414. 1:17:38coming and I don't think Sora is going to be the last one from open AI I think open AI is going to give us four or five
415. 1:17:44of these things I heard boxes wrestling and woke up I have my box out this is the world's biggest library and biggest
416. 1:17:57Museum I didn't go the whole way through it but I was yelling about uh about copyright so anyway that's my two cents
417. 1:18:07on it that's my that was that was my $187 on it and the other thing is this you can [ __ ] all you want you can whine all you
418. 1:18:20want you can pick it all you want the horse gun be out of the barn they shouldn't train on our data they already
419. 1:18:33did they shouldn't have access to all of our private data well that's Google and Facebook we gave it to them 20 years ago
420. 1:18:40so that ship's also already sailed they should pay me for my blog post okay how many blog posts do you think are out
421. 1:18:51there I if you took your your stupid little blog post let's say they weren't stupid let's say they were brilliant and
422. 1:18:58you did a blog post a day for the last 10 years how many other blog posts are there out there that if if someone
423. 1:19:08prompted chat GPT in a way that somehow said something that sounded sort of like something you wrote in a blog post if we could figure that
424. 1:19:19out what percentage of a penny would be fair to give give to [Laughter] you I made six Sense on my Collective
425. 1:19:33writings over the past decade I don't know it's it's just I have deep deep empathy for artists and creators who are feeling like their
426. 1:19:50reality is being ripped out from under them because it is I don't think it serves them to just [ __ ] about it and cross their arms and
427. 1:20:05go [ __ ] AI because it's not going to make ai go away and what's going to happen is you're going to have an artist like
428. 1:20:17Grimes shortly after Spotify took down the the weekend and Drake song that was an AI song that that was a hit on
429. 1:20:28Spotify and then Drake and the weekend's Publishers said take that [ __ ] down right now the week after that or maybe even
430. 1:20:37the day after they took that down Grimes Elon musk's ex who's a musician said to her fans synthesize me all you want deep
431. 1:20:52fake me all you want if you make any money give me half she's going to make a
432. 1:21:06fortune do you do you know who's going to make money for the artists that are trying to sue their fans the
433. 1:21:14lawyers the IP attorneys right now are [ __ ] dancing they're yacht shopping they've got the big glossy yacht
434. 1:21:22cataloges they're like dud do I go with 120 M meter huh okay it's going to take it's going to take a crew of seven at a
435. 1:21:34minimum all right so my annual burn is going to be $20 million or do I go with the 150 30 million bucks a year to run my
436. 1:21:47yacht I think I'm going to 150 I think I'm going to 150 oh yeah sorry yes Sue the [ __ ] out of the fans absolutely yep
437. 1:21:55no we're right there with you just we we're writing up the we're drafting the lawsuit right now hang I just got one
438. 1:22:02thing yeah let's go with the 150 let's go with the 150 meter yeah fantastic that what oh I can only keep
439. 1:22:10it in certain ports it it it the draft is too deep that's all you know what it could we uh could we dredge the uh the
440. 1:22:20ported ABA we could fantastic all right let's let's go with the 150 any anyway yeah we'll be suing those fans just shortly there
441. 1:22:34Lars that's who's making money in this [ __ ] so my my advice my my point of view is if you're an artist get AI literate
442. 1:22:46figure this [ __ ] out learn how it works understand what are the mechanisms cuz I guarantee you there are going to
443. 1:22:53be some companies that come out that make it really easy for artists to make a model of themselves and monetize it in fact 11 Labs is one of
444. 1:23:07those look
445. 1:23:23payouts uh voices create let's see um 11 Labs um forget what they're
446. 1:23:39called um voice for pay the [ __ ] was it called when you share your voice in the in the voice Library so on 11 labs
447. 1:23:54payouts 11 labs when you share your voice in The Voice Library you can earn cash rewards every time it's used so in 11 Labs right now if
448. 1:24:05you upload a 30 minute or longer sample of your voice so this is targeted as at voice actors you can create a model or multiple
449. 1:24:17models let's say you do a funny voice like this and then you do a fantastic voice like this and then you do do normal Kyle voice I
450. 1:24:28could make those three models put them on 11 Labs Market them and then every time someone uses that model I get
451. 1:24:39paid that's What actors should be doing that's what artists should be doing and you know what's going to happen a lot of
452. 1:24:45artists are going to be sitting like this going AI is stealing from us and the other ones are going to go
453. 1:24:55huh this is scarier than [ __ ] but I'm going to learn it and I'm just going to act like it's just another in a series
454. 1:25:02of Technologies that's going to change the game and I'm going to be ahead of the curve and I'm going to go make my
455. 1:25:07model in 11 labs and then I'll go make one over on Apple and then I'll go make one over here and I'll go make one over
456. 1:25:12there and then meta's going to do one and and then at some point it's all going to shake out and there will be two
457. 1:25:18leaders and I'll be one of the Premier voices on one of those two sites that survives and I'll have a career still and the
458. 1:25:28people that are like this they're stealing from me are just going to be drunk and angry well I could have done that but you
459. 1:25:41know and again I have deep empathy for them but it's just it's not going away it doesn't matter what you think about
460. 1:25:48it it doesn't matter if you think it's morally not cool it's coming anyway it's a
461. 1:26:05tsunami you can run around the beach all you want arguing about whether or not tsunamis are good it's still coming
462. 1:26:17right so buy a [ __ ] surfboard and start paddling man that was a soap box and a half wasn't it so I don't know if that
463. 1:26:30answers your question about how you deal with people that say this [ __ ] stealing I understand why people think
464. 1:26:40that I I I don't disagree with some of their points and this [ __ ] ain't going away so figure it out get ahead of it
465. 1:26:54license your [ __ ] before someone steals your [ __ ] recording artists earn from the live performances now less from
466. 1:27:05recording so I've heard yeah that makes sense and I listen tonius I think the value of live performances goes up in a
467. 1:27:13world of infinite content being able to see your favorite artist get up there and kill it pretty cool I forgot that Charlie thing this could
468. 1:27:27be a cool essay yeah I think this yeah this would be a good essay I should I should grab the transcript of my rant and see if there's
469. 1:27:35any coherent thought in there New York Times thinks they're dealing with old school search engine they're Clueless yeah
470. 1:27:44exactly exactly well I think I think all the people that are making these lawsuit if you look at the language of the
471. 1:27:52lawsuits they they're they're basically based on the 80s and 90s music sampling technology Theory we found clips of our [ __ ] in
472. 1:28:06your [ __ ] therefore it's in your [ __ ] but it's not and so so we'll see ultimately it comes down to who's got the better lobbyists
473. 1:28:20right and also the other thing about the New York Times It's mighty convenient that after they did aund million deal with Google they
474. 1:28:29sued open AI right I'm sure I'm sure Google had nothing to say about about yeah no you probably I wouldn't I
475. 1:28:38wouldn't cut a deal with open AI I mean haven't they trained on all your data I mean I'd sue them if I were you anyway
476. 1:28:45you do you here's your $100 million see you next week [Laughter] Bob not that bitter or cynical or
477. 1:29:05anything I wonder how all these will continue develop after the war will start what what are you talking
478. 1:29:15about I want to create one with a lot strong Russian accent uh but you need a paid act before you can upload you need a paid
479. 1:29:32act before you can upload to 11 Labs oh paid account yeah you do you do that's the thing about the media the
480. 1:29:44media stuff the audio and the video stuff right now is pretty expensive I wonder how these will continue to
481. 1:29:50develop after the war will start I don't know what that means still don't really care it did in a way but small artists barely get any
482. 1:30:00opportunities for live shows yeah so I mean I have a degree in acting I started and ran a theater company I'd work for 3 months to put on
483. 1:30:14a show that 17 people would come to see if you're in the Arts and you have some expectation that you should be given a [ __ ] career tough
484. 1:30:28[ __ ] [ __ ] grind no one owes you [ __ ] anything like get on Tik Tok and get you know 500,000
485. 1:30:42followers and get them to come to your show if small artists can't get people to come to their show it's because they
486. 1:30:49don't have talent or they're not hustling sorry it's the entertainment business for every job there are 10,000
487. 1:30:59people that would do away with a family member to have a tenth of an ounce of a career it is it is not a career for the
488. 1:31:12faint-hearted and you know you don't go into that world because you expect shit's going to be given to you
489. 1:31:26like go live in New York for three years and and try to get your ass on a stage try to get yourself an agent I did it for two years and I'm
490. 1:31:39like [ __ ] it and started a theater company I we'll do our own shows I did that for four and a half
491. 1:31:50years I don't do do you think I ever had a complete completely sold out run maybe once in 4 and A2 years it's a brutal business and it's
492. 1:32:06just been made more brutal but here's the thing people with talent that get AI literate are going to be like [ __ ]
493. 1:32:22superheroes because if you can put on a live show and you can wield this technology and you can spin things up in
494. 1:32:32real time and you can do big performances and you've got talented people around you and you can Market it
495. 1:32:40and you can get venues and you can get asses in the seat with very little money talented people are going to be able to put on
496. 1:32:52unbelievable spec spectacles with original music every night and original video every night and Original Stories every night but it
497. 1:33:03still kind of has a theme so I don't have a lot of sympathy for but you know small artists have tough times selling out tickets [ __ ]
498. 1:33:15get out there and grind you know you you came you came to the wrong out of work actor for for you know no one owes anyone's
499. 1:33:30[ __ ] I'm as empathetic as as the next guy I'm I'm I think I'm I index high on the empathy empathy scale and I I also have very little
500. 1:33:44tolerance for just expecting [ __ ] to be handed to you especially in the Arts especially in the Arts go watch Corey Sandler just she's
501. 1:33:58[ __ ] grinding right now right she's doing her Pottery she's doing cooking singing now she's doing gpts she's just
502. 1:34:06cranking cranking cranking cranking if you've ever done that you know in a year goes by and then 3 years go by and then a decade goes
503. 1:34:22by and you're still going you've essentially won but when you're in the middle of it it doesn't feel like you're winning
504. 1:34:32you're like I just need that [ __ ] break so I'm going to keep going because you're called to do it all right anyway we're supposed to be
505. 1:34:42talking about AI what did I put them is that water or or Schmutz I think it's Schmutz it's like so schmutzy I thought it was a
506. 1:34:59button and ticket prices oh my don't excellent points Kyle you're the best thank you Vegas Moxy Moxy thank you for
507. 1:35:07being a subscriber also I missed the rant it's still going I've been ranting now for two hours or an hour and a
508. 1:35:15half someone asked me about copyright Joker and I just I haven't really stopped Oh Lordy bring it Kyle oh yeah who reads
509. 1:35:31the New York Times love your passion Kyle thank you Edgar what is the new Sora a what is this new Sora AI technology I'll show it
510. 1:35:39to you oops not sure what I posted it was a good rant he's still going H let's go look at what this new
511. 1:35:51Sora [ __ ] is so I'll show you this I'll show you the Sora [ __ ] and then I'll show you the 11 Labs
512. 1:35:58response to it which I thought was super cool and I'll also show you how the Sora stuff works because the sorus stuff is really
513. 1:36:07significant it's really significant it's so funny too because it's like when when open AI puts out something like
514. 1:36:17Sora there's two kind of naysayers right there's the naysayers that are like the oh we're we're all assumed everyone's
515. 1:36:24going to die there's that that's to be expected that's kind of boring but then there's the technology naysayers that
516. 1:36:30are like it's not that big a deal I'm like well if it's not that big a [ __ ] deal where's your version of it you
517. 1:36:38know the [ __ ] Chief dude from meta came out like three days after Sora that's not really that SI technology is
518. 1:36:47not really that significant you know it's um and they're just taking some mathematics and they're applying it in
519. 1:36:53in a in a both a parallel and a serialized fashion over time and you know they claim to be simulating reality but you know from from a true
520. 1:37:01perspective they're they're not actually modeling a 3D World and they're not actually simulating the physics of the
521. 1:37:06world they they're just approximating it in a diffusion kind of model and you shut the [ __ ] up Jesus go look at the video where's
522. 1:37:20yours if it's that trivial show us yours so anyway let's go look at it I'm [ __ ] Saucy tonight I'm Saucy a [ __ ] had it
523. 1:37:37right good we [Laughter] good all right so if you want to see this Sora stuff go to open.com SLS
524. 1:37:52o and then you get [Laughter] [Music] this all right so there's our fancy little
525. 1:38:11website okay before I show you any cool video other than that background there which is pretty cool that's sort of video this first
526. 1:38:25sentence for me is a huge [ __ ] hint because the second sentence talks about video right this is a creating video from text right
527. 1:38:40is what the product is and their first sentence has nothing to do with video what does it have to do with we're
528. 1:38:49teaching AI to understand and simulate the physical world in motion with the goal of training models that
529. 1:38:55help people solve problems that require real world interaction like I have to when I first read that I read that sentence like 10
530. 1:39:12times wait you're what we're teaching the AI to understand and simulate the physical world in motion
531. 1:39:22with the goal of training models to help people solve problems that require real world inter inter what what does that
532. 1:39:31mean does that mean they're going into like Unreal Engine 5 creating worlds that have physics on it and training
533. 1:39:40like like just letting those worlds interact and exist with some sort of characters in there and training the AI
534. 1:39:47on that are they are they creating simulated physics world with material like what what does that even
535. 1:39:55mean with the goal of training models to help people solve problems that require real world interaction so what that says to me
536. 1:40:03is this thing this video [ __ ] is the it's like it's like the tippity tip of the iceberg of of whatever they're
537. 1:40:13doing so then it says introducing Sora our text to video model Sora can generate videos up to a minute long
538. 1:40:21while maintaining visual quality and adherence to to the user's prompt and then this is the this is kind of the
539. 1:40:26hero clip that they they keep showing this is a 59 second and even if this took an hour to
540. 1:40:43render this is pretty something look at there there's some weirdness in her feet but most of the other stuff here is pretty coherent and
541. 1:40:52for me the thing that Impressions me the most is all of the signs in the background stay and all of the
542. 1:41:00people are as they started out and then that edit that cut to the closeup of her that's not an edit the Sora model that's part of this 592
543. 1:41:13generation it cuts it cuts its own scenes together as part of the generation so that's that one that one I think's a piece of crap
544. 1:41:30this one's amazing again this is a single prompt with all these scenes that generated all these
545. 1:41:52scenes I mean good Lord right prompt a movie trailer featuring The Adventures of a 30-year-old Spaceman
546. 1:42:06wearing a red wool knitted motorcycle helmet look he's got a red so it was funny someone said oh it's not realistic
547. 1:42:16his his helmet is knitted it's in the prompt it's in the prompt he's got a red knitted motorcycle helmet why cuz I don't know someone
548. 1:42:27prompted it that way look at this this drone shot of Big Sir look at the look at the lighthouse look how it's
549. 1:42:38not moving and it's not warping and morphing notice how the detail in the cliffs is still there the detail in the waves is still
550. 1:42:49there so even if this [ __ ] is rendering at 60 Seconds per second you know that's a 10-minute shot rather than hiring a drone pilot to go
551. 1:43:04out fly around there or or paying the stockhouse to do it and you know the shit's going to get faster if you don't think it's going to get
552. 1:43:16faster if you if you don't think it's going to go faster go to here we'll go here watch this gro.com gq.com have you seen
553. 1:43:31this watch this um let's see write me three sonets about three people who each love
554. 1:43:50the other to look how fast that went Sonet one the engineer in numbers and in Street the lover of concrete and the true with
555. 1:44:05every beam and girder does reveal the strength and beauty of designs so new yet in his heart a poet's Soul resides a
556. 1:44:14tender loving artist he may be so this is this is not this is just using mix drill's 8X 7B model 524 tokens a second 524 so like
557. 1:44:27350 words a second to generate the response it's not copy and paste it it's generating it like that so this [ __ ] that's too slow right
558. 1:44:38now it's not going to be that slow for long anyway so Sora what else is there's there's another one that I think is super
559. 1:44:48impressive this one's super impressive so the The Prompt here is um closeup of a 24-year-old woman's eye in maresh during magic
560. 1:45:05hour and then okay so you're like okay that's cool but then wait right there the reflection in her Iris
561. 1:45:14or in her yeah in her iris is maresh at sunset ah holy [ __ ] what does this thing know
562. 1:45:31well they're understanding and simulating reality or physical motion this is this is the one that blew me
563. 1:45:42away did you see that wait I'll wait for it to Loop right when the dark building goes by outside
564. 1:45:56the train you see the reflection of the woman filming this video The Prompt here is Reflections in the window of a train
565. 1:46:06traveling through the Tokyo suburbs and there she
566. 1:46:21is that's [ __ ] B Bonkers so that's Sora so then 11 Labs 11 Labs two days after that came
567. 1:46:38out drop this thing so what 11 Labs said was gee isn't there something wrong with this super sexy video like say it's
568. 1:46:51quiet so they said we have a new product coming out you might want to get on the wait list it's called text to sound
569. 1:46:59effects or text to audio or text to whatever it's called but basically they use a text to audio generation model to create audio for all
570. 1:47:10the Sora videos or you know this this edit of them
571. 1:47:27that one's amazing so's
572. 1:47:44that in a place beyond imagination where the Horizon kisses the heavens one man dares to Journey where few have vent
573. 1:47:53ured armed with nothing but his wit and an unyielding Spirit he seeks the answers to Mysteries that lie beyond the
574. 1:48:21stars all of these sound effects were generated by 11 Labs without modification well it's not that
575. 1:48:37impressive you know from a technical perspective all they're doing there is they're taking this uh algorithm that came out in in
576. 1:48:452020 is the paper is is the is the they take the string algorithm and then yeah yep exactly it's not that
577. 1:48:58impressive go watch a 4 second Runway ml video and then go watch these and I don't give a [ __ ] if it took an hour to
578. 1:49:04render these and you can render a Runway ml video really quickly these are like usable commercially usable right Runway mLs are not unless
579. 1:49:15you're making some sort of you know Twisted anamorphic an you know morphing sort of thing um so let's go look at how these things
580. 1:49:26work or how this system works because it's [ __ ] Bonkers and I don't give a [ __ ] if if they're taking two relatively
581. 1:49:34simple mathematical Concepts and smashing them together they're the ones that did it first and no one else has got anything
582. 1:49:44close to this people are [ __ ] all over open a well they don't they're doing this and they're doing that and
583. 1:49:49they're doing this well they're their [ __ ] GPT is still better than your [ __ ] chat gbt and honest to God I swear to God as
584. 1:49:59soon as someone comes out that's better than open AI I'll shut the [ __ ] up about him but ain't no one caught him
585. 1:50:11yet and when they do [ __ ] like that like I think they did this because I think they knew that something was coming from
586. 1:50:16Google that was going to take the little Shine off their their marble so cuckoo all right here's what we're
587. 1:50:28going to do learn more in our technical report I want I'm not going to go deep into this
588. 1:50:43but okay here's video This is how video works if you don't know how video works it's a stack of images right thing called Persistence of
589. 1:50:56vision the sort of minimum Persistence of vision is 24 frames a second which is what movies are generally rendered at and that's
590. 1:51:06why movies tend to have that kind of warm analog kind of feel cuz it's like right at the edge of where you know they
591. 1:51:13don't flicker it just looks like motion to us and then videos 30 frames a second video games are 60 or 120 frames a second
592. 1:51:26but our our eyes above 20 I think it's 22 can't can't really see the individual frames we just see the motion so
593. 1:51:37what and and so if you've been on here for a while earlier I showed how diffusion models work right so it
594. 1:51:46amalgamates a bunch of images that have been noised into a single noisy image and then it Den noises that image it has
595. 1:51:54no idea what's in the image none which is why you get weird hands and weird feets and a nose on the side of the head
596. 1:52:04and an extra leg is because it just smashed some [ __ ] together denoised it and what it gets it gets it's completely unaware of of what it
597. 1:52:15is what open AI is doing here is they're in the coding process when they when they encode it on video when
598. 1:52:26they're teaching it about video they're taking each frame they're doing visual encoding of it and they're breaking it
599. 1:52:33into what they call Patches so the front of this grid here is a single frame that they've broken into patches and they do
600. 1:52:41that for every frame so they've got this kind of volumetric understanding they've got this understanding of all of the different
601. 1:52:50items sub elements of the image over time and then if you take any one of those little patches the way they're render the way
602. 1:53:03they're maintaining continuity is that they bring a single patch in and and they have like a 100 frames so like three seconds of video
603. 1:53:18and I don't I assume they're going front and back or maybe they're just going back and I don't know it doesn't [ __ ]
604. 1:53:23matter but let let's say down here is like the wheel of a sports car in that little patch they can look at the wheel
605. 1:53:31of that sports car over time and maintain continuity across it so they're doing micro diffusion models but they're checking them over
606. 1:53:40time so they're they're improving continuity over time for a little patch so that's how they're maintaining so what it means is this
607. 1:53:49system understands this world it understands that it's a reef with a fish in it and a scuba diver in the
608. 1:53:57distance and it understands what they are over time and because they're training the model to understand the physical
609. 1:54:07world it actually understands the relationship between a scuba diver and a fish in a reef holy [ __ ] [ __ ] Batman
610. 1:54:23this has nothing to do with video although it does it makes really cool video every time I learn about how they
611. 1:54:36do this [ __ ] it it it's like who's figuring this [ __ ] out well here's the deal people have been figuring this [ __ ]
612. 1:54:43out for the past three decades it's just now coming to fruition and and as overhyped as it is this stuff
613. 1:54:56is coming to fruition right now because open AI released this thing called chat GPT on November 30th 2022 and 100
614. 1:55:06million people started using it within 6 weeks Word of Mouth only and that changed everything because the focus of every
615. 1:55:18single tech company starting at the beginning of 2023 until today today is 100% on generative AI not just the AI that's been sitting
616. 1:55:29in our machines for 15 20 years generative AI the stuff we get to play with they're all focused on it so what
617. 1:55:37we're starting to witness right now all the all that list that I read at the at the beginning of the live is a year of development of people
618. 1:55:46dumping tons of money and tons of focus on these tools that's why they're all of a sudden starting to get better and
619. 1:55:52better and better all of a sudden it's like how is it possible this is all just happening right now because it's been
620. 1:55:59building for years and then starting last year all this Focus like laser focus went on generative AI so all the development's turning this
621. 1:56:09way all the money's turning this way all the research is turning this way and you've got someone like open AI
622. 1:56:16that you know they've been at it for for five or six years in a very focused way they put out a thing that they didn't think was a big
623. 1:56:26deal they had drinks the night before they launched chat GPT and said ah maybe we'll get a blog post or two and a couple of thousand
624. 1:56:35comments 100 million users in six weeks fastest adoption of technology in history with no marketing crazy
625. 1:56:51right B conkers all right I got to go people what else did I have on here on my little list my list 11 Labs yeah the Google pro 1.5 stuff I
626. 1:57:06don't have access to it yet but like the people that are talking about it like Ethan mik go if you're not
627. 1:57:12following Ethan mik go follow Ethan mik follow him on Twitter follow him on LinkedIn and go to his blog one useful
628. 1:57:22tricky Point how many people will make special spectacles in the future will be lost in the crowd many that's art true Kyle grind grind
629. 1:57:31grind yep just you know and you know Corey you're grinding is you know I say grind grind like a Gary vaynerchuk thing
630. 1:57:42that's not it it's like you have a passion for what you do right and it's not about the commercial success
631. 1:57:50necessarily like be nice to have I think as artists we all dream of having it and then it's art sometimes it resonates sometimes it
632. 1:58:03doesn't Tyler Perry calls on entertainment industry and government to cral AI before everyone's out of business
633. 1:58:10exactly the robots are going to kill us I'm never going to use it I'm a writer and I will never use AI you're not going to use the the most
634. 1:58:22powerful writing tool in the history of humanity really Skippy really you should try it it's pretty [ __ ] miraculous I've
635. 1:58:34written seven screenplays I'm kind of partial to the word I think AI gives me [ __ ] superpowers you know how I know that cuz
636. 1:58:48I used it I didn't just [ __ ] about it so [ __ ] bitter [Laughter]
637. 1:59:06tonight oh Lordy
638. 1:59:21again I'm extremely late but I have a good excuse I was making a GPT one useful thing yeah one useful thing. org is Ethan mullock's
639. 1:59:33blog just go look at what he's writing about Gemini yeah Gemini 1.5 Pro like Ethan mullik like he's he's
640. 1:59:43really even keeled and he's like here's some research that shows this that and the other and you know he's very much
641. 1:59:48he's very much a a uh an AI Optimist and you should be using AI but he's he's thinking about it very pragmatically go
642. 1:59:58look at what he's writing about the million token context window in Gemini 1.5 he's like this is different shit's
643. 2:00:06different feels different something's different this is different if Ethan mullock's saying something's changing something's [ __ ]
644. 2:00:15changing so anyway all right I got to go people thank you to all of you thank you to the Irregulars thank you to the mods
645. 2:00:23I know this was a non-standard evening but but I had I had to rant I had to rant I just had to do it people
646. 2:00:33um if you want to support the channel subscribe to the lives buy one of the video series in the corner or the video
647. 2:00:40series in the corner follow my channel um go to go to there the salon. a check out this community this is a community I
648. 2:00:54started the week after chat GPT launched it doesn't cost anything but go read about who we are and what our values are
649. 2:01:02and if you want to hang out with people that are up to [ __ ] in AI this is the group and by up to [ __ ] I don't mean
650. 2:01:10like they figured it out this is a group that understands that no one's got it figured out and they're just exploring
651. 2:01:16and they're playing and they're collaborating and they're supporting one another and they're not being [ __ ]
652. 2:01:21to one another and in fact when anyone in here is slightly assholish it's like this weird thing cuz no one in here's like that so if you
653. 2:01:35can put down your [ __ ] hat and connect with people and you want to learn some [ __ ] this is a remarkable
654. 2:01:43group non-standard is the standard all right everybody I am out of here Amelio's wife great to see you Vegas
655. 2:01:50Moxy thank you as all always good good good Danielle thank you investor friendly agent peace out laughing Thunder thank you thank you thank
656. 2:02:03you Brian Whitney sherid D thank you y'all are so swell you're so swell I just love you up Lord digital Gods good
657. 2:02:14evening uh LinkedIn office hours tomorrow 11:00 a.m. mountain time go to my LinkedIn Channel @ Kyle Shannon on
658. 2:02:21Linkin in and just check in the morning I'll have uh the the link there or you can go to the Irregulars Channel at the
659. 2:02:31salon um and I have the link the the link to the Google meet for my AI office hours to M tomorrow against the norm
660. 2:02:39been watching you every night thank you very much keep coming back keep coming back it it does get better no it doesn't this is pretty much
661. 2:02:47it Cory Sandler Pottery good night TK good to see you Joker a human being good to see you Debbie uh all right I think I got most of the
662. 2:02:59people if I missed you I love you anyway you know that Mr it good to see you peace out people have a good night bye