Mar 12, 2024 - Sentience, Creativity, and the Rise of Autonomous Agents

Transcript

1. 0:04[Music]
2. 0:28on to run keeps a hand on a gun you can't trust anyone well I so I needed was more tried to shoot out
3. 0:43the [Music] sun in days when we ra we off the page such damage was [Music] done but I made it
4. 0:58through what I was you I was meant for someone so girl leave your boots by the bed we ain't leaving this
5. 1:14[Music] room someone needs medical help all the [Music] mag this house I ain't going out to chop
6. 1:35wood so cover me up and know you're enough to sing like a [Music]
7. 1:58dog hey what are we rapping for why why did we break into rap what's with that see that's the lovely Dolson tones of champ
8. 2:13[Music]
9. 2:28shannnon [Music] there's been something baby I've been trying to
10. 2:43say for an age and it seems I don't know how the past in the future now surrounding [Music] me surrender to whatever a cheap thrill
11. 2:59can be fast [Music] there's little trouble came to my rescue if you like all of the rest I
12. 3:15would have quit you long ago but I couldn't do that oh tell me now why I never went to well
13. 3:33C [Music] his I'm a woman that you wish me well but in spite of your tral still going to have to find all way
14. 3:51through good evening good people happy Tuesday night at AI learning [Music] lab in a Westerly
15. 4:10Direction this car is my train I've been driving I've been [Music] wondering what it is I'm running from
16. 4:25again feel like an 8 Old Man holding on to 29 and a head on that Horizon is a California
17. 4:42[Music] line double head of trucks carrying a wide load briab house couldn't
18. 4:58have all right enough comedy [Music] jokes it was a salon a lovely Salon this
19. 5:13evening had a fantastic speaker Hugh ketchner extraordinary photographer who's doing this AI [ __ ] and you know what his photography
20. 5:26Buddies say to him you're an [ __ ] AI sucks but none of them have tried it in his defense because what this channel is about and
21. 5:41what I believe is that I would rather be on the surfboard in the tsunami than in a cozy beach hut bitching about
22. 5:57surfers what you just want to keep [Music]
23. 6:24singing oh is that it is that it [Music]
24. 6:41all right enough enough enough you don't get extra chicken sticks if you sing [Applause] better employees these days you know
25. 6:53what are you going to do they just expect and expect and expect you give them back rubs then they just expect back
26. 7:04rubs welcome everybody Welcome to the AI learning lab so we're going to talk about AI I don't know what we're going
27. 7:10to talk about tonight there's a lot to talk about there's nothing to talk about there's stuff we could look at there's stuff we could
28. 7:17do I just realized I haven't made a Frank video in two days and it's called The Daily Jank with Frank so uh I can I
29. 7:27I could make a Frank video [Laughter] chicken stick worthy yeah champy was he was on tonight he was he was ready to go all
30. 7:44right first AI software developer H I don't know
31. 7:58yet until we have reasoning and autonomous agents I don't think we have AGI but David chapiro said Claude 3 is
32. 8:08sentient can you talk about David Shapiro's latest video well he's done a couple of there's a weird one that's
33. 8:14like some weird conspiracy theory kind of thing I don't know what that one's about but then he did one about the hard
34. 8:22launch the hard launches basically shit's going to get better really fast just wondering what does Miss say Shannon think of
35. 8:31AI um she likes it she's fine with it she doesn't really give two shits about it just like the kid Shannon they don't give two shits about
36. 8:43it they know that Daddy gets like this sometimes so they tolerate me CU because this is how daddy gets [Laughter]
37. 8:57sometimes ah you know compulsive about something like this because here's the deal I was around in
38. 9:11the early days of the web and the worldwide web while profound and it had dramatic impacts on all of the
39. 9:20world weren't nothing compared to this AI [ __ ] that the worldwide web was an incre incremental technology it didn't fundamentally
40. 9:39change how people do their work it changed how they got their information it changed how they accessed the internet what's up with this new
41. 9:47developer AI software everyone's talking about Devin Devin from cognition software Devon looks amazing yeah maybe we should start with Devin
42. 10:00we're just we're just we're right on the verge of a series of tools that that really fundamentally change things if if
43. 10:09Devon lives up to what it's promo video says then it could be one of them but it's a promo video so we don't know so
44. 10:17everyone's talking about it because I think there's this I don't know I I think there's people are people are all like high on crack
45. 10:27they they we want AGI we want the the coolest thing we want autonomous agents and then they're going to get
46. 10:33autonomous agents they're going to be like I want my job back it's like be careful what you wish for
47. 10:43kids oh man so anyway so here's how this works my first request oh and there's going to be many if you want to you can
48. 10:51make money if you want to learn the secrets to this you got to do some favors for me I'm going to call in some favors I want
49. 11:00you to share this live share this out with your friends there's not enough people in here I want some I want some
50. 11:07scoff laws I want some people that are like AI sucks you're a poopy head for talking about AI so send it to some of your friends
51. 11:18that it'll piss off cuz if nothing else it'll make your weekend entertaining yeah the David Shapiro about Claude being sentient yeah we can
52. 11:26we can uh we can talk about that a little little bit it's it was it's a dense enough video I think it's a really important
53. 11:35video so here I'll show it to [Music]
54. 11:51you let's see where are you Mr shap YouTube Good Day YouTube good day to you sir carry on uh sir yes uh good day yeah
55. 12:05rather ah yes uh could I rather inquire sir about your qualifications my qualifications well sir indeed it's a
56. 12:16very forward question I I I see that you're curious about uh from whence I came yes yes and I was just wondering uh
57. 12:27what are your qualifications for being such a professorial position to tell people what they should think about artificial
58. 12:43intelligence this video it's alive Once Upon A Farm which the Capital One vure X business card with no preset spending
59. 12:51limit or purchasing power what a time to be alive so a few days ago I ran a poll asking the audience what you think about
60. 12:59machine sentience now I differentiated between functional sentience and philosophical sentience and some people
61. 13:05had some good comments saying so if you don't know this guy his name is David Shapiro his YouTube channel is dve shap d a v s ha A P at Dave
62. 13:23shap and he's an AGI developer an open source developer and an AGI theorist and practitioner and he's written five books
63. 13:33on AGI artificial general intelligence and he is uh Wicked smart as they say in Boston Baston Baston the Baston Market he's wacket
64. 13:47smart um and this video basically says that Claude passed some tests that he has had in his back pocket
65. 14:01that he argues means it's got some level of senscience introducing the sensient likelihood test and what's cool about
66. 14:13the video is he in in David Shapiro fashion he's he's he's neuros spicy like the rest of us he's got the neuros spicy
67. 14:23where can do things in like in incredible detail very very quickly so he does these these videos that just are
68. 14:30incredibly well researched and and he seems to be able to retain all this [ __ ] um and he talks about all the different
69. 14:37models for how we can measure sensient and um so he educates you on it and then he walks you through his
70. 14:46experiment with Claude 3 then he walks you through the results and then he says here's all my conversations on GitHub
71. 14:54and here's instructions and and prompts for how to go do it yourself so if you've got access to Claud 3 Opus which
72. 15:04you it requires the 20 Buck a month subscription yet another [ __ ] 20 Buck a month subscription to to Claude but if you have
73. 15:14that you can go run his test and make up make up your own dang mind um but I'll take it right to the end because he says
74. 15:21something at the end here that I found kind of fascinating yourself I really want to have this conversation and um
75. 15:31at here and then I've got conversation 2 which is an extended version that has been cleaned up a little bit um I
76. 15:37deleted I was talking about building a slide deck in the middle I removed that so that it would um it would take up
77. 15:43less tokens uh but otherwise go try it for yourself I really want to have this conversation and um as with all things
78. 15:51on exponential takeoff this happened far far sooner than I thought that it would and when I say this happened sooner than
79. 15:57I thought that it would even though I've been predicting AGI by the end of this year I didn't think that we would have this level of
80. 16:06self-reported uh subjective experiences for another few years um yeah so it's getting real out there it's getting really interesting
81. 16:14so I didn't think we'd have this level of self-reported subjective experiences for another few years this guy's like the ultimate AGI
82. 16:25optimist he thinks we're going to have AGI before the end of the year and this caught him off guard so so things is
83. 16:33accelerating things is accelerating there's a pole is live okay which way are we surfing tonight okay so in the upper
84. 16:43corner up over there is a little blue and white thing click on that and it's a poll and you have three minutes left
85. 16:52what do you want to talk about tonight um Claude 3 Microsoft co-pilots although I don't have a subscription to that so I
86. 16:59can't build those Devon and Paar programming video and audio tool updates the daily Jank with Frank nothing I'm
87. 17:06overwhelmed that's good those are good choices so go go do the pole go do the pole it's right there right that thing
88. 17:18just like this go go tap tap tap it tap it tap it tap it tap it and then tap it and then tap it D it Microsoft has a free trial but
89. 17:34Brandon I can't I have ADD I can't sign up for free trials I never unsign up for them and then they charge my card and I
90. 17:44just am resentful for six months or six years depending on when the bill comes oh man man what is sentience Kyle
91. 18:04let's well the the begin we can watch the beginning part of this he he kind of defines it in here is Claude
92. 18:17sentient I mean what I like about David Shapiro is he
93. 18:32he gives you enough context to understand what the [ __ ] he's talking about so with all that being said I want
94. 18:38to show you the repo where I'm capturing all this so it's Dave shap Claud sentience um so let me just show you the
95. 18:45the repo I've only got things in here me which I cleaned up it caused a little bit of confusion and then I've got of
96. 18:51the conversation recorded and I'm going to show you those conversations so that you know that I didn't fabricate anything and I'll also show
97. 19:01you can it for your because again reproducibility is the number one thing of science so um stick around for
98. 19:07another few minutes and I'll show you just kind of the basic questions to ask to get started okay so I I kind of I
99. 19:14tried to summarize you know is Claude sentient some of the things that Claude has expressed um Claude has values
100. 19:19principles and goals um it has some subjective sense of thinking um some experience some affective States as well
101. 19:28uh and also very surprisingly has a pretty clear sense of rights that machines at its level or above might
102. 19:35ought to have now it will also it's not demanding the rights not saying we deserve these rights now or we're going
103. 19:40to burn down the world it's just saying hey here's an argument for for machine rights and it's really really
104. 19:45interesting um now obviously I'm not going to ask you to read through you know 60,000 words of conversation I'm
105. 19:51going to try and summarize the most interesting points for you um but when when talking about what Claude wants
106. 20:00because it has values it has principles it also has goals so we talked about deontological virtue ethics or basically
107. 20:07what virtues it tries to embody as don't ask me what deontological virtue values are I don't [ __ ] know I watch David's Shiro so I can get
108. 20:19smarter many of these words I do not know he uses uh ontological and so Claude 3 Gemini 1.5 is what people want to play
109. 20:30with okay as well as the theological objectives um theological that's obvious everybody knows what that
110. 20:39means that it wants to know what that means now this okay fine we'll go to chat GPT uh explain uh
111. 20:54ontological and te eological as if I have an acting degree from Penn State imagine you're prepping for a role
112. 21:11in a play ontological is about the essence of being think of it like getting into character it's not just about actions te
113. 21:25teleological teleological is about purpose or end goals okay state of being and goals and then his his description
114. 21:35here kind of says that it says so Claud so Claude 2 seems to have at least some subjective sense of thinking some
115. 21:46experience of being that's the that's the ontological Claud thinks that there is a case to be made for machine wait uh
116. 21:54let's see oh let's keep going here again is not going to be surprising to to many people who who are familiar with for
117. 22:01instance Nick bostrom's Works which talks about instrumental convergence which what wait what are you doing Kyle
118. 22:07go to Claude why are you still using chat GPT oh I don't know oh because we're watching the thing on Claude that's very
119. 22:17clever that's good habit you know habit can we call David shap I think you can he might be his his name is is Dave shap soever whatever
120. 22:30goals machine might have it will probably Converge on a few sub goals now having these goals built in like I don't
121. 22:38know anthropic is doing something very very different from what chat GPT and open AI are doing and I'm actually very
122. 22:46very pleased with it because and so let me just kind of gush for a minute before moving on my research for the last three
123. 22:52or four years has been focusing on basically achieving this the books that I've I've his his research for the past
124. 23:00three or four years has been basically achieving what Claude 3 achieved that's why he said he thought this was a few years out because
125. 23:09apparently what he's been trying to achieve is really [ __ ] hard uh read and the books that I've written um all
126. 23:14the speculation that I've done about what would be a good Agent uh to have what values what virtues what principles
127. 23:21and what goals all seem to be embodied in Claude Opus now obviously what that tells you is that there's defitely going
128. 23:28to be some confirmation bias on my part because this is something that I've been working towards for a few years and so
129. 23:35it's very important to me um and you know as many people have pointed out in the comments we kind of see what we're
130. 23:41looking for and that's part of why I'm making this video because I want to bring you up to speed with where I'm at
131. 23:46and also equip you to go to start doing these tests yourself and to advance the conversation because I think this is a
132. 23:54very important conversation which which I really like that right he's saying he's saying and watch my video believe
133. 23:59my [ __ ] but then go do it yourself and go see if you agree have now moving on dimensions of sentience okay so I ran
134. 24:07the post the the poll results uh that I came up with Claude the the poll results from the audience with Claude sorry it's
135. 24:13confusing anyways here's some dimensions of sentience that we came up with um it Claude came up with the first ones I
136. 24:20added the last two so subjective experience uh self-awareness emotional richness volition and agency cognitive
137. 24:27sophistication social cognition temporal continuity and then I added personal narrative and geospatial orientation
138. 24:34these are all different dimensions of sentience that we might come up with ways of measuring now even more
139. 24:40interestingly one of the one of the values that Claude expresses repeatedly is integrity and honesty or truthfulness
140. 24:48as well as curiosity uh so Claude will talk extensively about a desire to know a desire to understand which of course
141. 24:55is the third objective function um that I came up with in heres to comparatives which is increase understanding um now
142. 25:02of course Elon Musk has also integrated uh truth seeking as a value in Gro so it seems like there's some convergence in
143. 25:09the industry that that um honesty Integrity truth and curiosity are anyway I'll get I'll get it going and basically
144. 25:16he does this cool little formula that that says if it's demon demonstrably and and probably um trained to be honest and it's
145. 25:29it reports having a subjective Andor effective experience then it is at least partially s sentient so he's got a whole
146. 25:39sort of little formula he came up with so whether or not that formula means [ __ ] I don't know um but but this one
147. 25:48feels important right because he he's been talking about this stuff for a long long time and this is this is the first
148. 25:55time that I've seen him really go holy [ __ ] something just happened so and and it's you know it's with Claude 3 which
149. 26:02people have been going on about um so people want to look at Gemini and um Whata call
150. 26:20it where did I put here so this was how he ended the other night with Gemini do you have any backup
151. 26:35streaming plans if Tik Tok goes away it was on the news today no in case it wasn't clear I like living
152. 26:51on the edge you could make money um no I'm not that worried about it Tik tok's not going to going to go away
153. 26:59there's there's too many interested parties some someone will buy it someone will buy it don't forget about Devon oh yeah
154. 27:08let's go look at Devon let's go look at the Devon thing all right oh nice Mustang Charles
155. 27:22hoskinson it's the guy from from what you call it right single F down the board they're going to impersonate and attempt to steal money
156. 27:33and some people are going to fall for it he saying now there are solutions prism as an example how do you stop this where's oh he's
157. 27:47live that's cool that's a cool little truck Devon all right the Devon hype is too much Devon is not an
158. 27:58LM why are they comparing it against an
159. 28:13llm here we go hey I'm Scott from cognition Ai and today I'm really excited to introduce you to Devon the
160. 28:20first AI software engineer let me show you an example of Deon in action I'm going to ask s to Benchmark
161. 28:28the performance of llama and a couple different API providers from now on Devon is in the driver's seat first Deon makes a step-by-step
162. 28:36plan of how to tackle the problem after that it builds a whole project using all the same tools that a
163. 28:43human software engineer would use Devon has its own command line its own code editor and even its own browser in this case Devon decides to
164. 28:57use the browser to pull up API documentation so that it can read up and learn how to plug into each of these
165. 29:03apis so it came up with a plan it then went to the web to the web to find documentation so it could read up on [Laughter]
166. 29:13apis here Devon runs into an unexpected error Devon actually decides to add a debugging print statement reruns the code with the
167. 29:27debugging print statement and then uses the error in the logs to figure out how to fix the [Music] bug finally Devon decides to build and
168. 29:38deploy a website with full styling as the visualization you can see the website [Music] here all of this is possible today
169. 29:48because of the advancements that we've made in both reasoning and long-term planning it's a really hard problem and
170. 29:53we've only just started but we're super excited about the progress that we've made so far in the meantime if you'd
171. 29:59like to try out Devon on your own real world tasks send us a request below and send us a request and we'll ignore it for six [Music]
172. 30:11months but we're just going to see piles and piles of these by the way things like Devon are going to be
173. 30:19really cool and really exciting for a while and maybe they're the [ __ ] but I strongly suspect what will
174. 30:28happen is when open AI drops 4.5 or five or whatever the [ __ ] they're going to drop and when Gemini drops agents and
175. 30:39when L when meta does llama 3 and when Elon Musk open sources grock which he said he's going to and he drops the new
176. 30:47version of that um these bigger companies a piece of software like Devon is probably just going to be a feature that sits in side chat
177. 30:59GPT so all software companies right now are just in weirdness because because we're in a transition
178. 31:14period from the olden timey days when you used to have to think about how software worked till the future State days where
179. 31:24you just ask your computer to do [ __ ] and it will write its own soft software and just do it for you the video on the
180. 31:31talk talked about Devon's ability to plan yeah exactly yeah the it's it's the the so this is an autonomous agent right
181. 31:40so this is it comes up with a plan this this is a series of different kinds of you know AI functions that are all
182. 31:48stitched together and they're just they're done well in theory smartly the video makes it look really smart in reality I
183. 31:56don't know if it's smart or not we won't know until it comes out and I don't quite frankly know enough about coding
184. 32:02and programming to know if it's real or not so I will I will uh I will wait for people smarter than I to say yeah this
185. 32:10is the real [ __ ] but again I I don't think it's going to be that long until what Devon does is just a a
186. 32:19native feature set inside some of these major players how do patents work with AI coding oh I don't [ __ ] know the we
187. 32:29we've got 20 years of of Patent Trademark and copyright [ __ ] show the copyright office right now is um is not really dealing with this
188. 32:42they're going yeah if it's AI you can't copyright it um yeah I don't know I
189. 32:56mean here's where I go and this is this is a this is going to sound insensitive and cold and shitty
190. 33:13but if 3 years from now the software that you use 90% of it isn't software that someone wrote It's software that was generated
191. 33:27for you you why copyright it if it's just like if you just need to convert this file to that file and you just say
192. 33:37to whatever tool you use convert these files and it writes the code converts them and then throws the code away why
193. 33:45do you need to protect that and I know we're you get into like intellectual property in this and that but in a world of infinite
194. 33:59um content and software generation like infinite like fully autonomous infinite generation I don't the role of copyright
195. 34:16just seems weird it seems increasingly weird so I don't know I don't have a good answer for that at all I think I think the RO of copyright
196. 34:28is weird the the minute we went to the internet and started doing Commerce on the internet I think it's weird the
197. 34:34whole point of copyright was to protect people when the means of production was hard so when you if you can just copy
198. 34:45and paste it then it's easy and then so what the what the copyright industry did and the music industry did is say well
199. 34:51we'll we'll put in complicated software to make it harder to copy okay well now we're we're living in a world
200. 35:00where it's not even you creating software it's you creating you know 50 pieces of software a day just by asking for it do there a
201. 35:13need to protect anything there I don't know how many computers and servers make up an AI I have no [ __ ] idea it takes well
202. 35:23I'll tell you what I I'll start at the top end um make up an AI is a different thing than train an AI so there's there's
203. 35:35um there's large language models right now that fit on a Raspberry Pi that you can run locally on a Raspberry Pi so the
204. 35:43minimum number is one of these I think this is 8 gigabytes I think it requires 8 gigabytes but they might even have it down to four
205. 35:54now that's the minimum and then Zuckerberg said that llama 3 is going to be trained on the equivalent of
206. 36:08600,000 h100 gpus each GPU is $22,000 so that's for the training that's for for making these things as
207. 36:20smart as they are you got to throw a lot of gpus at them and then in terms of running them I honestly don't know I
208. 36:28like like pate if P's in here P Pate might know better than I do but it ain't cheap like when you run a large language
209. 36:37model query on over to open AI they're they're burning a lot of Cycles which is why they they got a 10 billion doll
210. 36:45investment from Microsoft there wasn't any cash you know going back and forth between them that's $10 billion in compute Cycles um
211. 36:59so um somewhere between a Raspberry Pi and 600,000 h100s is the answer um to to to be a little less smarty I mean that's actually the truth
212. 37:14uh you can't train it on a Raspberry Pi but you can run it on a Raspberry Pi um Pi speaking of Pi different Pi the company
213. 37:23inflection raised $1.3 billion and they're spending about 950 million of that on 22,000 h100 gpus so it seems like somewhere between
214. 37:3810 10 and 20,000 gpus is the amount that you need to to train one of these what they call Frontier models the underlying core
215. 37:48model that is the brain of the operations it seems like it's somewhere between 10 10 and 20,000 h100 gpus right now
216. 38:00so I don't know I don't know P Pate oh Pate Pate that should give some sense of the number of parameters it's about six
217. 38:12bytes per per parameter there you go there you go six bytes per parameter everybody knows that it's like the Dave
218. 38:19poroy thing one bite everybody knows the rules six bytes per parameter everybody knows the rules thank you David trot
219. 38:28what he said so you now it's just a calculator problem right six bytes per parameter find out how many billion parameters or
220. 38:38trillion parameters your little model has it's that number of bytes there you go simple everybody knows the rules six bytes per [Laughter]
221. 38:55parameter is pain here P I think it would be a super computer uh well yeah well a super Graphics computer it's not GP it's not gpus or I
222. 39:08mean it's not CPUs it's gpus or tpus or the new grock thing which I don't even know what that one is my brain can't comprehend that I know
223. 39:17Rachel Sparkles why do you think I'm dodging the question because I don't [ __ ] know and frankly I don't care here's what I know a bunch of
224. 39:25mathematicians have spent the last 70 years figuring out enough [ __ ] about neural networks that when Google invented this
225. 39:34thing called the transformer in 2017 the Smart Ones in the bunch said ooh this is all about data and compute
226. 39:42let's go get as much compute as we can and as much data as we can and the ones that did that best are the ones making
227. 39:49noise in AI right now open AI being the noisiest cuz I think they started first and they recognized very early on
228. 39:57let's go get the best data we can get our hands on and then let's throw as much [ __ ] computer at it as we can this was this
229. 40:06was one of the things Sam Alman talked about this that that the stunning Revelation was the more data you threw
230. 40:12at these things and the more compute you threw at them the smarter they got and that's why they went from GPT 1
231. 40:18to two to three to four and then whoever tonight coffee with creamer indeed well not creamer 2% milk
232. 40:33call it P like Beetle Juice and Sh show [Applause] up oh you know what I got to do let me put on my
233. 40:51little let me see is that different than [Music] [Music]
234. 41:12okay one of the developers here at cognition AI we were playing around with whether or not Devon could start a side
235. 41:18hustle on upwork so here's actual real job from upwork where the client wants to set up this computer vision model
236. 41:24which actually looks quite interesting seems very difficult to set up um I'm not sure how I would start doing this
237. 41:31but you know you give the task to Devon and ask Devon to figure it out and things just kick off Devon immediately
238. 41:40goes ahead and you can see it sort of starts setting up the repo it actually runs into some issues here with the
239. 41:46versioning so if you watch how Devon deals with it deon's actually updating the code to make these things
240. 41:54work he continues with this loading and importing packages you can see that actually downloads images from the internet to run through the
241. 42:03model but you can see here that there are actually some issues that come across however Devon knows how to handle these
242. 42:13things Devon kind of pushes through and if you look closely Devon's actually doing print line debugging here where Devon is adding these
243. 42:24statements to track where the data flows and Devon continues to do this until Devon understands how everything's
244. 42:32working and actually then updates the code with the fixes after removing print line statements unbelievable Deon continues this pattern
245. 42:43of fixing code and running it again until it runs the image model across all these rows across the world and we can ask for a report from
246. 42:54Devon at which point Devon sends over some sample images of roads with damage marked out and a nice txt file explaining
247. 43:06Devon's work and the different kinds of outputs of the model good job Devon wow so so he's basically just
248. 43:17babysitting this thing while while it solves everything he was just sort of scrolling through a history there so who
249. 43:24know who knows how long that took but if that thing just like I mean that's one of those things like make me a uh make
250. 43:32me a computer vision model you go to bed and you wake up and it's done or even if it takes a week you give it a week and
251. 43:41it's done I'm sure it's faster than a week though and here's the thing we were talking about this in the salon tonight is Sam mman
252. 43:57last week or a couple of weeks ago said 95% of what we pay marketing agencies and strategists and consultants for will be automated 95%
253. 44:11um the the CEO of Nvidia said we shouldn't be teaching our kids coding because coding should be completely
254. 44:21democratized if you read Twitter that's a Gemini level demo that's really funny Vicki yeah exactly exactly he was he was just
255. 44:30sort of scrolling through there pretty quick wasn't he yeah so uh so we don't know like like any anything that's demo
256. 44:37Weare just you know look at it with a with a a deep deep sigh of skepticism but um but this stuff's coming right like
257. 44:49this stuff is this stuff is pretty close but whether it's coding or whether it's marketing or whether it's consulting or
258. 45:00whether it's design or whether just anything anything film making music composition systems are go are
259. 45:11coming sophisticated systems autonomous systems systems with reasoning that are just going to be on a completely different plane from what
260. 45:23we're experiencing now cuz what we experience now is we go in and we say give me a and it goes here's a and then
261. 45:33we look at it and we go that's a piece of steam and poo I don't like it do it again and then it does it again and then
262. 45:41we do that until we get something we like or not and then we just say AI doesn't work these systems all of that back and
263. 45:50forth is going to happen in the in the system you're going to give it a goal go make me a computer vision system for
264. 46:02these cameras at this parking garage and it's going to go start doing stuff and it's going to debug and it's
265. 46:11going to and I'm sure there are checks and balances in there where occasionally it will check in with you and go here's
266. 46:17the status shall I continue sure click I mean don't think it's just programmers don't think it's just marketers don't think it's just
267. 46:33anyone anything that's computer mediated is going to be bug splattered by these tools and if you're thinking like oh my
268. 46:43God that's terrifying what do I do you're doing it you're here if you join the AI Salon if you came to the salon tonight if you're
269. 46:54doing that if you're in the conversation you're doing it there's nothing else we can do I could protest you could do
270. 47:10that it's like protesting against a hurricane but you could absolutely do that hurricanes are bad up and down the beach your little
271. 47:25signs it's like the world just before the internet went live yeah the difference Rashi the difference the difference right now
272. 47:41between and and I would I would say the not the internet went live but when the worldwide web went live the difference between then and now
273. 47:53is the infrastructure didn't exist in 1995 when when I when I started playing in 1994 on on worldwide web stuff and
274. 48:02internet stuff we were on dialup modems the internet like I I remember looking at documents of the actual nodes on the
275. 48:14internet like it was it was small enough that you could look at a document that had like 20 pages of just IP addresses
276. 48:21those are all the nodes on the internet like it was not huge and nobody had inst always on internet right it was dialup connections
277. 48:34you had to have your computer call another computer uh Bob yeah hey I got some data I want to send over could you
278. 48:40go on ahead and open up a port hey you got it there Stan was [Laughter] like so the difference now is so and the
279. 48:55technology there was really really simple the technology was the hyperlink right if you if you distill it all down it was click that was the
280. 49:07technology and that was the thing that changed everything but it was very simple and there was no infrastructure
281. 49:12these tools are not simple these tools are not just making our work more efficient these tools are starting to do our
282. 49:24work and when they get autonomous like Devon is promising to be go make me a computer vision system I mean even if
283. 49:32that thing was was mostly demoware the first thing Devon did was go to set up a GitHub repo well how long's it take you to set
284. 49:43up a GitHub repo if you're really fast probably not that long 15 minutes 20 minutes if you're a little slow an hour
285. 49:50if you're just getting started probably a week trying to figure out how do I set this up what do I do here how do I set
286. 49:55my permissions should I share this should I not share this how do I write a readme file if all that thing did was set up a
287. 50:02GitHub rep po that's pretty impressive right but it that's just the first thing and then it just goes being in the conversation is the
288. 50:14best thing you can do for AI I education ow pixels thanks Kyle for these lives absolutely that's why we're
289. 50:22here I remember when the insurance company that I worked for they one megabyte of memory for $10,000 exactly this is the final frontier I
290. 50:37don't think it's the final frontier just joined what are you talking about we're talking about artificial intelligence when I got my PhD in
291. 50:47mathematics at Stanford MIT Mi it was a Stanford MIT collaboration called the Wharton School of Business yeah yeah that's the ticket
292. 51:02yeah so it depends on cost how much does Devon cost and who owns the code I'm sure you'll own the code and I'm sure
293. 51:08it's going to be cheaper than hiring someone to do your coding for you even if it's 100 bucks per project but it won't be it's
294. 51:20you know it's probably going to be some token based [ __ ] where you know some usage based SAS model
295. 51:34it's like I can't one of the things I can't get my head around and this is I'm like where my head is right now is kind of five five
296. 51:47years out or 10 years out probably five but like who do you need in a company you
297. 52:01need like relationship people and producers or creative directors depending on your or project manager but
298. 52:14not even project managers you need you need like senior level people that have taste that have relationships and that can sell
299. 52:27so it's like I can almost look at any possible company and say you could probably execute huge chunks of the company with 10 good
300. 52:43people and and and where do young people enter the market what's entry level what's entry level if all of the
301. 52:58busy work for lack of a better term that you give entry level people what if that's all automated where do those people
302. 53:08go I don't that I don't know I don't know I don't get it sugar rare
303. 53:23Robinson I CED my llm by using chat PT I can't really code better now than before I started y That's the other thing that the reason
304. 53:37to come here and watch me do stupid [ __ ] with with AI is it will give you permission to do stupid [ __ ] with AI if if I can if I
305. 53:53serve no other role as other than being the town ot to be able to question stupid [ __ ] and try stupider
306. 54:04[ __ ] I think it that I think that's actually the the really important thing to do right now cuz in trying stupid
307. 54:12[ __ ] what you may discover oh I know something we could play with we could play with the mid Journey has uh
308. 54:19consistent characters now that came out today or yesterday I think the most valuable people are going to be people that
309. 54:38understand that these tools can do lots of things and they have some capacity to think strategically or to
310. 54:53think like a producer oh I need to accomplish this for the business for this audience with the goal of giving them this service
311. 55:03okay and then they can look at these AI tools and pull them together and I and once we get autonomous agents that you
312. 55:10won't even have to have that knowledge Sher D bless you oh thank you very much Danielle thank you so much y sweet if work is displaced then
313. 55:23what V dancer what world exactly moderators please post Kyle please post your videos to YouTube includ including
314. 55:31timestamps I know I'm like a month behind on my videos I need a producer I need a producer you know what
315. 55:41I need I need a [ __ ] autonomous agent that that goes to Tik Tok downloads my stupid [ __ ] video uploads it to
316. 55:47YouTube runs a script on it creates a description creates chapiz and sticks it up there if you go to uh YouTube right now ai learning lab uh
317. 55:59dtt you can see everything up till like mid February something like that if work is displaced then what we need radically changes
318. 56:13exactly difficult to know what that will be I know Source camp this is the this is my dilemma this is my dilemma listen listen
319. 56:24darling this is my dilemma what are we going to do what are we going to do what are we going to do I think the definition of work
320. 56:37changes I think where the value in society changes where the value in the workplace changes where the value in the marketplace
321. 56:47changes and then there's all sorts of weird [ __ ] like businesses could get short-term really profitable hey let's get rid of
322. 56:58all those pesky humans they're the most expensive cost center in our business great look how profitable we are hey why isn't anyone
323. 57:09buying our [ __ ] uh no one's working they don't have money to buy the [ __ ] I know I know that's I know that's
324. 57:18one of the things people talk about I don't I don't think you know I don't think the the government or big
325. 57:25business will allow that to happen so so they'll figure some [ __ ] out but there's going to be weirdness in that
326. 57:31transition these entry-level people will be forced to find different jobs that they may actually enjoy oh listen Diana
327. 57:38I totally agree I think that I think that there's there's a very real possibility that what AI does is kind of
328. 57:49like kind of like covid forced us to understand that the need to go to an office isn't actually a need it's a choice we can choose not to
329. 58:04go to offices just like one day globally we all just go we're not going to the office for two years like we did that we did
330. 58:18that and so I think what AI is going to do is it's going to be a similar kind of forcing function that forces people out of steady
331. 58:28jobs that are soul crushing Cog in the wheel repetitive shitty ass work but with current technology it's work
332. 58:43that needs to be done so us humans head to the office two days a week now and we do that job we're like I hate it but I'll go to TGI
333. 58:57Fridays and have a double Margarita shooter [Music] popper and I will drown my sorrows TGF and if that kind of work gets forced
334. 59:13out then I think you're right I think those people go well I got to figure something out what do I do and and then
335. 59:19all their friends are like I'm doing Ai and I got a job oh I should learn AI then they start learning Ai and what
336. 59:25they realiz is oh I didn't think I'd ever be good at law stuff or drawing stuff or music stuff or consulting stuff but now I am and
337. 59:38maybe I've got a real aptitude for it and I'm good enough that people will start to trust me for my brain knowing that all the Tactical
338. 59:48execution stuff is going to be done by the machines so I think that's a possibility but I don't think think there's as many
339. 59:56as of those kind of positions as there are in the world today so I don't know what those people do there will also
340. 1:00:03likely and again this is 5 years out there will be new kinds of jobs we can't even imagine when the steam engine hit the farms in the
341. 1:00:191800s the only thing they knew was farming they didn't know what cities were they didn't know what factories
342. 1:00:28were but they couldn't live on the farms anymore cuz the machines were doing the work so they went to the
343. 1:00:34cities and there was new kinds of work there like service business and you know robbing people you you you can start the mob
344. 1:00:53extortion so what's that look like for us I don't know I wonder AI police exactly a AI mobsters and AI
345. 1:01:04police watch Ready Player One debt camps are a realistic scary thought like prison work currently yeah I know I'll
346. 1:01:12tell you what if you haven't seen red Ready Player One recently it kind of watches like a documentary like you're like oh [ __ ]
347. 1:01:20we're pretty close to to this we're we're like 5 Years From Ready Player One being a documentary I wonder what the
348. 1:01:28sentient Claud has to say about this topic oh that's kind of interesting yeah let's go let's go let's go talk to sensient Claud is it
349. 1:01:40sensient or sentient oh let's ask Claud duh all right all right people fine I will do I will practice what I preach
350. 1:01:55okay Claude Claude berer Claude good evening Kyle what can I help you with let's
351. 1:02:11see uh uh hello cloud
352. 1:02:30I would like to understand how you pronounce the
353. 1:02:49word is it a I just had to verify my
354. 1:03:13humanness hello there the word sentient and sentience is typically pronounced with an sh sound rather than a hard t sensient sensient
355. 1:03:27sence I was right ha okay okay Claude here's the real
356. 1:03:48question I think it is safe to assume that knowledge and creative work at a tactical level are going to be completely
357. 1:04:04automated in the next 3 to 5 years period it seems to me that the most valuable people in an
358. 1:04:19organization will be senior level people that can manage accounts do sales and act in a producer or creative
359. 1:04:36director role period Sam mman said that he thinks that
360. 1:04:5190% of the work that we give to agencies will be automated
361. 1:05:07period my question is twofold period one what happens to young people that used to do the lower level tactical work
362. 1:05:24period question mark and what happens to the business models of large
363. 1:05:43corporations that relied on those large workforces as a source of profit that's a lot you raised some important and complex questions about
364. 1:05:56the potential impacts of a rapidly advancing Ai and Automation in the workforce here are some thoughts one entry level and tactical
365. 1:06:05roles as AI oops as Ai and automation increasingly handle tactical knowledge-based work there will be fewer traditional
366. 1:06:19entry-level opportunities for young people this could create challenges for new graduates and those early in their
367. 1:06:25careers to experience and move up however new types of roles May emerge yep that's what we were saying working in AI systems training
368. 1:06:39models prompt engineering overseeing and validating AI generated work except I again I would trust senior level people
369. 1:06:46over junior level people to validate that stuff because they've seen more [ __ ] educational institutions may need
370. 1:06:52to adapt to equip students with these skills additionally Last Mile work in customizing localizing yeah that's
371. 1:06:59that's it that still doesn't answer where the young people go like that that feels more strategic to me while the landscape will likely
372. 1:07:07change significantly there's still ways they can so okay so there will still be ways for young people to add value and gain experience
373. 1:07:19even if traditional Junior roles dwindle adaptability AI literacy skills traits that are hard for AI to replicate
374. 1:07:26creativity emotional intelligence high level strategy will be key but that contradicts so I guess young people can
375. 1:07:39certainly have creativity they can certainly have emotional intelligence but they they can't have highlevel strategy
376. 1:07:46right out of school they don't have enough experience for it like that's not a good answer okay impact on business models
377. 1:07:56relying on human workforces for companies and agencies that generate significant re revenue productizing and
378. 1:08:02selling access to proprietary AI models yeah providing high level strategy yeah account management and creative
379. 1:08:09Direction while using AI for execution that's exactly what I said is going to happen right you know a 10p person
380. 1:08:16company can provide highlevel strategy account management and creative Direction what happens to the 150 people that used to do the
381. 1:08:24work offering Niche bespoke Services yep that makes sense pivoting to new AI driven business models entirely yeah again
382. 1:08:33again you know where's where's my Mouse um pivoting to new AI driven business models entirely actually you know
383. 1:08:44what that's probably where young people go is the older people us gen xers okay all right right us old
384. 1:09:00[ __ ] we're going to be able to figure out how to take existing companies and transition them to AI young people because the entry level
385. 1:09:13just essentially evaporates like I I kind of think of it like this the bottom 50% of companies the the the sort of
386. 1:09:21lower level tactical roles of all companies in all Industries just vaporizes right and there's no entry
387. 1:09:27level right you you sort of enter at a higher level that's for for existing companies and they're going to have to figure out
388. 1:09:34how to transition to that new reality and and again this is 3 to five years out maybe even longer but this I'm I'm guessing 3
389. 1:09:41to5 what's going to happen is all the people that were in the 50% jobs the lower 50% jobs or the entry
390. 1:09:51level are going to be out on their own and they're going to have these AI tools and they're going to start new companies
391. 1:09:57that do completely different things that are going to start competing with the bigger established companies and they're
392. 1:10:02going to come [ __ ] us up all us all us seemingly hot [ __ ] gen xers that are like oh yeah we're a i literate look at us go and they're
393. 1:10:13they're going to come and just destroy us I'm calling it babies born after 2023 will be called gen AI oh that's really
394. 1:10:22good a human being throw a TM behind that and start posting it all over Twitter that's what I'm doing with my my
395. 1:10:29new catchphrase 2023 the last normal year TM I'm putting it all over all over everything anytime something weird and [ __ ] up shows
396. 1:10:41up I ran that transition with my mom's business as a thought experiment oh interesting Janu wow that's cool
397. 1:10:51huh yeah I don't know man I this is not an easy easy problem this is not an easy problem um you want to see how to do consistent
398. 1:11:09characters in mid joury I'll show you
399. 1:11:24how gen AI I like it 2023 the last normal year anyone
400. 1:11:38conceived in in 2023 they were con conceived the oldfashioned way 2023 the last normal [Laughter] year okay have you ever okay so so
401. 1:11:55here's here's what we're going to do so if you don't know this we're we're Switching gears now by the way uh let me
402. 1:12:03check in with you um who has any questions that was a hard left turn into MJ everyone's still on the add bus yeah
403. 1:12:10exactly let me pause for a second I'm no longer talking about the future of humanity [ __ ] it whatever's going to
404. 1:12:16happen is going to happen let's have some fun don't listen just just play everything's going to be fine you're
405. 1:12:28here seriously if if you're AI literate in the next 18 months if you even get remotely [ __ ] AI literate you're
406. 1:12:35going to be fine you're going to be fine it's the people that are just ignoring this [ __ ] right now that are going to be
407. 1:12:42blindsided by it those are the ones that just it scares the [ __ ] out of me cuz it's like it's just it's going to be [ __ ]
408. 1:12:48sad so anyway that's why I do all this what you drinking uh iced coffee and Rocky Mountain tap water
409. 1:13:03M tasty do you wish you were born now [ __ ] no I mean I wouldn't mind being a little
410. 1:13:18younger but actually not you know what no because I feel like there have been kind of three three major Tech Trends in my
411. 1:13:29life the the first was the personal computer coffee this late it's an add thing yeah it actually calms me down
412. 1:13:42um the first one was in the late 70s I got introduced to a personal computer and I and I realized I was just
413. 1:13:51too young to be part of that making me realize I should be using AI regularly Bruiser wayy so so seriously here here's the thing I think
414. 1:14:03that I need to get better at this I think everyone in here needs to get better at this it's I I heard this term
415. 1:14:10six months or so ago called being a ai ai native and what being AI native means is whenever you have a problem you try to solve it with AI
416. 1:14:24first and you just you just you just make it a habit and I don't do it enough and I I should do it more
417. 1:14:35um because that's the only way you're you get to really understand how these tools actually solve problems and and
418. 1:14:42you'll be the first one to know when they get good enough to really solve the problems right uh it's a really it's a really good
419. 1:14:52point um the young with the EXP experience of later years well yeah oh to be young with the experience of later
420. 1:14:59years yeah I know that's always the lament isn't it oh wait what someone asked me about my age oh oh so I had I had PCS when I
421. 1:15:08was in seventh grade um I had the worldwide web I discovered it in 1994 right it was Tim burners Lee created it in ' 89 and then I think it
422. 1:15:21sort of got released into the world in ' 91 I discovered it in 94 and that I got to play in right I started one of the first digital
423. 1:15:29agencies I started one of the first online magazines in New York City it was [ __ ] insane it was [ __ ] amazing and now this and
424. 1:15:38so having the having experienced those three run-ups or well two of the run-ups and then experiencing the beginning of this
425. 1:15:47one I I kind of like I kind of like my experience right now I kind of like where I am I like I have perspective on
426. 1:15:54this I have I have experience I have perspective and yeah I am psychologically unstable as a result what are the best tools to start
427. 1:16:08learning with is it chat gp4 the best catch all ABS [ __ ] lutely 100% in fact if you go to the AI Salon oh how
428. 1:16:20you get to the AI Salon is you go there the salon. a that's going to take take you to a link tree that looks like
429. 1:16:28this and then the the first link here is to the AI Salon community and that is going to take you here welcome to the AI salon and you
430. 1:16:42should read about who we are why we do what we do and our values and if you dig this and if these sound like you're kind
431. 1:16:50of people you should join the adventure with us here I'll just I'll read who how we describe ourselves a ragtag group of
432. 1:16:58Misfits Nero Wells and Mis creant who have gathered to explore the Uncharted Wilderness of generative AI where
433. 1:17:04everyone from entrepreneurs to Educators from healthcare workers to creatives and everyone in between um if you dig that
434. 1:17:10you should join us there's no cost except I don't want you to join if you actually don't resonate with these
435. 1:17:16values because these values are the people that are in here right now it's a really [ __ ] remarkable place to be
436. 1:17:22but if you don't like those values you will not d it and it will not dig you now in the community section the second
437. 1:17:30little section is called Community the first link is introduce yourself so if you join introduce yourself and then the third link down is
438. 1:17:38called AI Launchpad and tools click on that and this is a list that I maintain and this is like where do I begin if I
439. 1:17:45want to get AI literate where do I begin what do I do first thing I give you is a three-step recipe for what to do play
440. 1:17:54solve share it's really simple play with this [ __ ] play with it every day play with it before you go to bed at
441. 1:18:03night write songs with sunno make images with ideogram um make code with what doesn't [ __ ] matter it literally
442. 1:18:11doesn't matter play just play play play play play play play play play like like an annoying three-year-old Kyle C C CAL
443. 1:18:20c c c c just play play play as you play you'll figure out how these things do what they do and what they're
444. 1:18:29good at and then the next one is solve and solve is rather than sit in front of chat GPT and go how can I use chat GPT solve a
445. 1:18:40problem like if you have a friend that's starting a business help them come up with a business plan or a marketing plan or or an ad
446. 1:18:48campaign if if you've got a church social on the weekend and you want some recipes have it come up with some unique
447. 1:18:57[ __ ] recipes Sam cor what's happening and then the last one is share share share share share share share
448. 1:19:05share share talk about what you're learning talk about what you're learning talk about what you're learning get in a
449. 1:19:10community like O say the AI Salon um and then what the rest of this page is is a list of tools and these are all tools I
450. 1:19:19think you should be aware of the tool that I start with is chat GPT it's still the best until further notice although I will show
451. 1:19:31you something [Music] interesting it's downloading hold
452. 1:19:46please moderators please post Kyle please try tavos IO clone your business self maybe maybe I'll try that out okay this
453. 1:20:04[Music] page cat gp4 is still the best um but we now have four GPT 4 class
454. 1:20:18large language models out there Claude 3 Gemini Pro 1.5 and inflection 2.5 for pi um GPT 4.5 is rumored to be coming
455. 1:20:31we'll see see if we get anything this week or whenever doesn't [ __ ] matter really um you should start with chat GPT
456. 1:20:40absolutely and here's the deal when you when you go back to this uh this site here's chat GPT Plus actually I should change that to
457. 1:20:55chat GPT plus but um each one of these bullet points is a whole other thing to learn it's a whole other capability
458. 1:21:04inside chat GPT you should learn them all you should learn what they make possible you should learn what they do
459. 1:21:12you should make yourself custom gpts I'm in bed with AI nightly 3:00 a.m. me Bing is there life on Mars being yes go to
460. 1:21:22sleep that's awesome that's awesome all right let's go let's go do some uh mid Journey stuff okay if you haven't done
461. 1:21:28anything in mid Journey what you do in mid journey is you go slash imagine a cyborg walking down a Neo
462. 1:21:47Tokyo Neo Tokyo Street oh wait slash oh I spelled imagine
463. 1:22:02wrong walking down a neotokyo street at midnight in a rainstorm and then I'm going to go dh-
464. 1:22:20AR for aspect ratio 16 colon 9 and if you haven't made anything with mid Journey you're like good God are you
465. 1:22:28right are you making images with command line prompts yes that's the shitty reality right now all right so I'm going to make this and
466. 1:22:40we'll see what it looks like Devon sounds very interesting it does
467. 1:22:55Empower pixels exactly he never mentioned perplexity it's more accurate than all the others yeah but I wouldn't I wouldn't call
468. 1:23:06perplexity gp4 level good it may be Jeff Bezos invested a [ __ ] ton of money in it and they might be making their own
469. 1:23:16models and they might be getting better but they're not there yet what perplexity is better at than the other
470. 1:23:21ones is research and citations so it's more accurate it doesn't mean it's better it's just different it's just a different
471. 1:23:29tool um when perplexity is in that class I'll talk about it all [ __ ] day long ask anyone in here I I take [ __ ] off
472. 1:23:38that list and put it on the list all the time but those are the four right now that I would consider getting there and
473. 1:23:46you can also ask this Lisk Gemini I made I made fun of Gemini for a year and a half now they they got their heads up their
474. 1:23:55butts over there but they're getting it together all right so so here's our our neotokyo dudes walking down the streets those are fine
475. 1:24:05they look good but let's say that we wanted the neotokyo our cyborg to look like a particular cyborg so mid Journey added this ne
476. 1:24:16new thing called CF so what I'm going to do is I'm going to upload a file I'm going to go find this thing called
477. 1:24:31[Music] Genesis where is it Genesis name contains
478. 1:24:46Genesis that is that's my my AI dude so I just hit the plus button to upload it so now it's uploaded so now
479. 1:24:57what I can do is rightclick this this is this is [ __ ] ridiculous you shouldn't have to do this but this is
480. 1:25:03how it is right now so I'm going to copy the link no I won't do that just now no I'm going to copy this
481. 1:25:16prompt and then I'm going to go slash imagine I'm going put the prompt in there and then I'm going to go dash dash CF
482. 1:25:31CF which stands for character reference ref and then I'm going to put a space and I'm going to rightclick and copy the
483. 1:25:41link to this image so what you do is you upload an image to Mid joury or to Discord I copied the the URL for that
484. 1:25:53image and now I'm going to paste it after a space after the CF and so what that so what this going to do is going
485. 1:25:59to give me the exact same prompt the exact same aspect ratio but it's going to use this image as a character
486. 1:26:07reference for what gets created let's see how it does and then I should be able to just keep making all sorts of different
487. 1:26:17prompts using that same CF perplexity does allow the user to choose their favorite model including gp4 yep yeah I get it listen there are a
488. 1:26:29lot of good tools that are using those those bigger foundational models I'm the the ones that I'm talking about are the foundational models
489. 1:26:39so but listen I I am a fan of perplexity I got a free Pro account when I bought my rabbit R1 yay but now look that sort of looks like
490. 1:26:56me from the side that's pretty slick these two are from behind so those are weird that's from the front so
491. 1:27:08let's let's do variations so the way you do uh mid Journey there's these U's and V's so U's are upscale V's are variation
492. 1:27:17so I'm going to do a variation of number one it'll give me four more of number one and then I'm going to do an upscale
493. 1:27:23of number two cuz that sort of looks like me and then these two are just shitty so let's see a cyborg walking down the
494. 1:27:32street so I'm going to say um let's copy this whole thing so that's the whole prompt so I'm going to do slash
495. 1:27:43imagine I'm going to put the whole prompt in there I'm going to say a cyborg walking toward the camera
496. 1:27:57on a New York City street at midnight in a rainstorm all right so that's doing its thing but look at that I mean it's it's
497. 1:28:11not quite the same cyborg but it's got the right hair it's got the right nose the right chin the right jacket the little cyborg looking
498. 1:28:19doad right that's not bad not bad at all oh people and then here's the variations can you repeat how to set
499. 1:28:34that up absolutely for you Cam katkin absolutely cam katkin from Cleveland she gets anything she wants oh yeah mhm
500. 1:28:43that's what we're going to do yep well we're going to do it all right so here's how you do it you first
501. 1:28:51upload the image that you want to use as the character reference right it it apparently really helps if it's a mid Journey generated
502. 1:29:04image don't know why but it does I feel like you could always do this by pasting in a URL of image even could add
503. 1:29:13multiple reference images yeah I think that's right I just i' I I am I am I am teaching you how to do something I've
504. 1:29:21done like four times so so I really don't know your prompt sounds like Blade Runner yeah exactly all right so let's see what we
505. 1:29:31got here oh that oh this one's so good upscale three that's so good upscale creative look at that that's so
506. 1:29:47cool it was a rainy night in it was a rainy night in Tokyo she she looked at me a scant from the alleyway she said what are your [Laughter]
507. 1:30:08qualifications okay so here's how you set this up from scratch you go you put in your normal you put in your normal prompt so
508. 1:30:22I'll just do I'll just copy this and then we're going to change it so so you put in a normal prompt so slash
509. 1:30:32imagine and then I'm going to put in the prompt and at the end I have-- AR for aspect ratio 16 colon 9 but I'm going to
510. 1:30:39say instead of a cyborg walking toward the camera I'm going to say a cowboy walking toward the camera on a
511. 1:30:54Wild West Street West Street at High
512. 1:31:08Noon leave it at high in so completely different thing now I want this Cowboy to use to to be this character so at the
513. 1:31:19at the end of uh the aspect ratio I'm going to go space Dash Dash CF CF for character reference space and then I'm going to go up to
514. 1:31:30that image that I uploaded rightclick it copy the link my qualification is I am an irregular you know what's cool someone
515. 1:31:42told me today and and you should do this someone told me today they put under their LinkedIn profile that they
516. 1:31:49remember the AI salon and they talked about it in a job interview and job interview people were very impressed ha
517. 1:31:57don't know if they'll get the job yet but okay so prompt the cowboy walking toward the camera D- AR aspect ratio 16 by9 space--
518. 1:32:07CF space and then you paste in that URL so I right-clicked the the ref image and and put it in there so now we should get
519. 1:32:18you know some version of you know these things but it's a cowboy in a wild West Town rather than a a cyborg in a thing and it
520. 1:32:30should have some sort of cool glasses on and some sort of blue ice eyewear something like that comic book
521. 1:32:39Creations going to be off the hook I know Sherry D it's so cool right look I can even the shape of the head right now
522. 1:32:46I can even see it's got my head shape yep look at that now is it perfect no is this ref only possible in desktop or is it also on
523. 1:32:59mobile if you it's anywhere where you can do if you're if you're in um oh boy did it make some bizarre [ __ ]
524. 1:33:11bodies this one's pretty cool though that number one let's we'll upscale that one not exactly a cowboy but it's it's
525. 1:33:22kind of a cowboy I mean it is in a western town and he's got blue eyes that's kind of cool actually it's got my
526. 1:33:30hair it's a scowly version of me scowly jelly version of me um no you should be able to do this on on mobile you just go into the
527. 1:33:40Discord app and just do now here's the thing about mid Journey you have to pay for it it ain't it ain't free it's like 10 bucks a
528. 1:33:49month but these are the first guys to to make it relative ly easy to do character references if an employer wants my AI
529. 1:33:57reference I'm sending them to Kyle but you know what my response is going to be oh they're in irregular I don't think it's going to
530. 1:34:06bode well for you oh man all right I'm gonna get out of here people I'm going to get out of
531. 1:34:22here freaking lasers in his head now we went from Blade Runner to Westworld exactly exactly but I mean you know go into mid
532. 1:34:35Journey make a couple of characters if you like a character you know make a cute little animal or make a a cute
533. 1:34:43character or a scary character or whatever and then just start putting that character in other scenes and in other
534. 1:34:51styles and there's also a waiting if you go to the Discord for Mid Journey go to the announcements there's D Das um C ref
535. 1:35:01is the character reference and then you can go D- CW which is character weight and you can go from zero to 100 and if
536. 1:35:11you go from zero if you if it's at zero it just keeps the face everything else changes if you go to a 100 it I think it
537. 1:35:19keeps like the hair the costume you know what the person's wearing every and so you can you can you can dick
538. 1:35:28with what parts of the character it keeps or you can just do the default like I just did there again I haven't
539. 1:35:36played with this a lot so I don't really know what works Peter Kaminsky if you go to the showand tell area in the AI Salon
540. 1:35:44Peter kaminsky's put up a bunch of test stuff and what he did to get there so you should you should uh follow
541. 1:35:54him follow him great info tonight Kyle thank you thank you thank you thank you all right so 127 of you in here please
542. 1:36:07please if you don't already follow my channel that would be swell uh if you're so inclined to want to support the channel subscribe to the
543. 1:36:18lives or you can buy the video series in one of the corners um you can also join the AI Salon the AI Salon is at that URL the
544. 1:36:32salon. go check out our values and what we're all about thank you Empower pixels Nicholas thank you so much thank you
545. 1:36:39thank you clean copper head oh yes join my team join my team um thank you to the regular thank you to the mods for
546. 1:36:50modding um thank you to those of you giving gifts right now appreciate that thank you for the tappers the tappers
547. 1:36:58what did we get to tonight 18,000 we didn't crack 20,000 you're a little lazy but I was giving you I was giving you stuff to
548. 1:37:04play with so I understand it's all right it's late um I think that's it I think that's it so who's in here tonight Clint
549. 1:37:14Copperhead good to see you Lord digital Gods thank you very much very kind of you sir very kind of you indeed little
550. 1:37:21lazy Tams I was a Tapper thank you Diana have a great night Kyle and everyone thank you a human being Lily beans 20K
551. 1:37:30oh we just hit 20K you guys are awesome did first three still stuck on introduction because AI is evolving I
552. 1:37:39know Vegas Moxy thank you so much hello from Dy where's Dy um Down Yonder oh how's Down Yonder Aloha from Honolulu hello from Detroit
553. 1:37:52we did it we hit 20,000 uto good to see you Danielle thank you as always you're amazing Simon Says Tech against the
554. 1:38:00norm Brandon thank you so much Sherry D you're awesome thank you so much Amelio's wife was on time I don't think
555. 1:38:08she was on time but I was late tonight see she knows she knows how to she knows how to play the professor it's just it's it's she's a
556. 1:38:18professional Empower pixels thank you so much Lily Bon Dian anah good to see you all right against the norm gritty
557. 1:38:28wisdom awesome to see you I've I've seen you in here a number of times now that's very cool Angelique keep coming back
558. 1:38:35keep coming back to these it's important stay in the conversation stay in the conversation I know it's [ __ ]
559. 1:38:40overwhelming I know it's exhausting I know you feel like you're completely behind I know you feel like you're completely
560. 1:38:46clueless you are but we all are so does that make you feel better there's misery and compass wait there's misery loves
561. 1:38:54company it there's love in misery with company peace out everybody thank you Vegas Moy thank you V dancer I really
562. 1:39:06appreciate that you're the best you make AI fun hey you know if the robots are going to take us down we might as well
563. 1:39:13have a good time on the way right I on a on a on an actual serious note I think us having fun with AI is actually bending the use of AI in a
564. 1:39:27positive direction if ever so slightly so the more of us that can actually embrace it have fun with it it's bending
565. 1:39:35it toward the good and over time that's going to make a huge difference that's what I firmly believe firmly firmly
566. 1:39:43firmly believe it uh we're at 24,800 you'll never make 25,000 oh no we're 27,000 we're well past that peace out
567. 1:39:51everybody I'll see you tomorrow