5/15/2026 - Why Local AI Models and Hardware Specs Matter for the Future of Personal Computing

Transcript

1. 0:01[music] [music] [music]
2. 0:21[music] [music] >> Good evening, good people. Wait, you
3. 0:37can't see me. Hold. Hold, please. No. >> [laughter] >> Double producing. What's happening? Happy Friday night date night,
4. 0:48everybody. Um welcome to the AI learning lab. It is uh Friday. It's Friday, May 15th. I hope you're all doing well.
5. 0:59Um Friday nights are AMA nights, ask me anything nights, or as Silver Fox like to call them, American Medical
6. 1:05Association nights. So, I will answer either questions uh about AI or anything about how to run your medical practice.
7. 1:15Happy to do that, too. Um you know, what codes you enter to get reimbursed by insurance. One of my favorite topics. Go through
8. 1:25the codes. We can talk about code entry. We can talk about patient records. Um if there's any specific patients you want
9. 1:32to talk about and their maladies, this is a HIPAA-free environment. We can just talk about it. Nah, it's just Joe I'm just joking.
10. 1:42I take HIPAA quite seriously. >> [laughter] >> Um welcome, welcome, welcome. So, if you've got uh questions about um
11. 1:51about AI, pop them in the in the in the comments below. Um we've got some teed up that have been asked over the course of the week. There
12. 1:59were a couple of good questions that were asked uh in the in the pre-show uh for the uh for the evening. So, welcome, welcome, welcome. Um,
13. 2:10I'm trying to think if there's anything anything big and burning I want to start with. Couple of interesting things. I didn't
14. 2:18get a chance to play with it today, but but OpenAI this week um made four four decently They're not major announcements.
15. 2:29Four announcements. Um, one of them is that they're starting a deployment company, which was then quickly followed
16. 2:35up by Anthropic and Google saying they were doing the same thing. So, a services company. Um, one was they were launching a
17. 2:43security program for white hat hackers to get access to the models before the black hat hackers get access to it. Um, and then yesterday
18. 2:58OpenAI added Codex inside the ChatGPT app on your phone, meaning um you can um you can run your Codex Codex instance
19. 3:13that's on your desktop from your phone. So, just like OpenClaw. Uh and then today um OpenAI announced OpenAI ChatGPT for financial services.
20. 3:26So, you can connect all of your personal finances to ChatGPT. >> [laughter] >> Now, if if if if you're security-conscious,
21. 3:38you might be thinking like, uh is that such a good idea? Uh good thinking. Probably not. Um, I happen to think we live in a
22. 3:47post-privacy world where everyone knows anything every way anyway, so I will probably do it. Um, I was one to run off and get a uh
23. 3:58>> [laughter] >> a uh a world ID. I got my retina scan scanned um three two three two and a half years ago.
24. 4:09Um, so I'm willing to be your guinea pig. Yes, producer Brandon. Well, Kyle, go on. Um, if anyone has any comments, just pop
25. 4:20them in the comments below and uh go ahead and throw a question up on screen there, producer Brandon. Just checking my stuff here.
26. 4:40>> [clears throat] >> Okay, so this was a question that was just asked in the pre-show. Do I think that the
27. 4:56open-source models have gotten good enough to compete with the closed-source models? It's it Compete with is a really interesting
28. 5:05question right there. Um [snorts] the open-source models have gotten good enough that you can run something like
29. 5:13OpenClaw on a on a decently meaty um computer, which just when I when I say a meaty computer for um for purposes of AI inference, think a 32
30. 5:26GB or higher machine. So, 32, 64, 128 GB of unified memory on a Mac or VRAM on a on a PC on your on your GPU. Um
31. 5:41the Qwen models and the new Gemma models from uh from Google are quite good. Um, they're fast enough, they're powerful
32. 5:49enough that you can really run open claw and you can do full on inference on them. Um, if you're expecting them to
33. 5:56compete with, you know, Opus 4.6 running at high with high reasoning and high time spent on compute, it will not even be close to that.
34. 6:07>> [snorts] >> Um, but that's really reserved for kind of high-end coding and things like that. If you want to do just basic
35. 6:14I don't know, writing copy and and writing emails and doing basic business stuff and doing basic math and doing
36. 6:21basic coding, um, they are pretty remarkable. Now, I, let me share my screen. Just this afternoon created an infographic that is shared in
37. 6:36the AI Salon. So, if you're in the AI Salon, which if you're watching this, you are, um, you're going to hop over over here
38. 6:45on the left-hand side. Let me just make sure that you all can see this. Yes. Okay. Um, on the left-hand side, if you go to
39. 6:56the community feed, should be near the top here. Yeah, there's this post right here called running local models. And if you open this up,
40. 7:07>> [clears throat] >> I've got, um, the the left-hand column is about Macintosh computers, um, and how much,
41. 7:20you know, RAM you might want to look at and what you can run on it. In the middle section, here are the different, um,
42. 7:27local models that are that are good and interesting. Over here on the right is PC specs. Um, this little middle section here says if
43. 7:39you've got this amount of RAM, here's the kind of things you can run. Um and then down at the bottom, this talks
44. 7:45about multi-modality and and what these things can uh can do from from multi-modal perspective. Um is right there.
45. 7:57Hey Kyle. Sound like you did good. Yes. Yes. Uh they are clamoring and telling me that people might be having a hard time
46. 8:06finding the chat because it's hidden behind the AI Learning Lab logo in the upper right-hand corner of our display.
47. 8:12So if you're looking for where to chat your questions in, just hover over, you should see the view chat button on the
48. 8:19right-hand side, or just tag me anywhere in this line and I'll get it posted. But if you're looking for the actual chat,
49. 8:24if you'd like to chat with Kyle live, you can do that. Operators are standing by. And my screen share just went away again, didn't it?
50. 8:32It did. I thought you took it away intentionally. No, but that's okay. But anyway, um yeah, so so this this graphic
51. 8:41um that I just created today is is in the community feed. It's like one post down. All right? Um Yeah, let me let me talk about um
52. 8:58why I think that I think it is I think it is worth everyone's time if if you're serious about AI, um it's worth your time to at least
53. 9:08start learning about what's possible with the local models, um and what that universe looks like because it seems
54. 9:22increasingly apparent to me that the ability to generate AI tokens, so tokens are the things that give you responses,
55. 9:34they're going to get increasingly expensive with the frontier models. They're going to get increasingly restricted from the frontier models.
56. 9:40They're going to save their high compute for their high customers. Um and they might be restricted in other ways. And and so
57. 9:50And and they're also likely going to get more expensive because right now they're being subsidized by investors. So every
58. 9:56time you hear Anthropic raise $30 on a 900 or 30 billion dollars on a 900 billion dollar valuation, that 30
59. 10:05billion dollars, a lot of that money is going to subsidize your ability to use Claude code cheaply. That piper will eventually stop being,
60. 10:17you know, subsidized and their rates are going to go up. So the cost of tokens is going to go up. So if you are working on something locally
61. 10:25and you're doing all of your inference on your local machine, it's just the cost of electricity that that thing's
62. 10:30going to cost you. So that's what I would I would uh would pay attention to. Andrea, let's see. This is a good question.
63. 10:40Question: On an old computer, could you use an open source model? It won't work. It won't update to W11 like Linux maybe.
64. 10:51Okay, so here's the deal. Um Brent Peterson, um who's been He's been writing a an AI agent memory system for the past 3
65. 11:06years. And he's finally got his thing dialed in and he's got 14 different agents, each of them running on a Raspberry Pi. Now,
66. 11:16th- this is one of the one of the really confusing things about things like Open Claw. For things like Open Claw, which think
67. 11:23of open an Open Claw agent, an instant instance of it running on a computer, kind of like your own employee. And And what what
68. 11:36that little computer does is it does all the routing and the manage like when you talk to it on Telegram and you send it
69. 11:43an instruction, it's sending all the instructions out. If he's doing things on a Raspberry Pi, >> [clears throat]
70. 11:49>> he's not doing any inference on the Raspberry Pi itself. He's going out to OpenAI or to Claude, to the to
71. 11:56the frontier model companies through an API and using, you know, those tokens. There are all sorts of different
72. 12:03services where you can you can send things from a from a an older crappy machine. Um the answer is yes, you can
73. 12:12run them. So, you can run Open Claude on an older crappy machine. You You can't run uh likely. You can't likely run a
74. 12:22powerful local large language model. Um but there may be some, you know, specialized models that you could run.
75. 12:30So, this really has to do with with GPU speed, GPU RAM more than anything. So, 32 GB of RAM is probably the the
76. 12:40limit at which you can run something decent. Um but it doesn't mean you can't run things like this. It just means you might need
77. 12:49to go find third-party, you know, something other than Anthropic or API or or OpenAI. But you can run their smaller
78. 12:57models. There's all sorts of things you can do to keep your token cost low. Um in fact, on OpenAI, you can use what's called OAuth. So,
79. 13:08basically you log in with your $20 a month subscription and you can have your Open Claude use the um GPT-5.5 model using your $20 a month
80. 13:22subscription. It's good. It's powerful. The problem with it is for 20 bucks a month, your open claw will probably burn
81. 13:28through all of your credits for the week in about half a day. Right? And so, that gets a little frustrating. So, then you have to kind
82. 13:36of dial down to the to to the cheaper, less expensive models and things like that. But anyway, okay. Beautiful.
83. 13:45I'm excited for the Okay, I'm in Orange Beach, Alabama. Cool. Orange Beach sounds lovely. Hope it is. Attending
84. 13:53grandniece's graduation. Oh, this is Mary. Hey. I don't want to lose my irregular status. Oh, you're amazing.
85. 13:59Thank you for checking in. It's good to see you. Um I'm excited for the AMD machine. Wait, where did that go?
86. 14:07Todd Waller, I'm excited for the AMD machine that's coming. Yeah, there's a there's a new machine coming from AMD.
87. 14:13So, right now there's the AMD machine that's coming. Uh Nvidia has the uh the Spark, the DGX Spark, I think it's called.
88. 14:22The the Mac Studios or the MacBook Pros um are really good architecture for this stuff. Um those are the ones that are all
89. 14:30dedicated. They all will run you in the $4,000 range, 4 to 5 grand. If you want to step up to Mac Studios and do things
90. 14:39like, you know, 512 GB of RAM, you're probably in the $10,000 range. If you're sitting there thinking like,
91. 14:48wait, $4,000 for a computer or $10,000 for a computer in an era where we can get a decent computer for 1,200 bucks,
92. 14:55why would I do that? Again, because AI tokens um are very valuable things, and you can you can think of it less like I'm buying
93. 15:07a machine and more like I'm I'm buying a potential employee that I can train up to do stuff for me that works 24/7. That's assuming you
94. 15:18know what you want to do with it. Okay. Um bada bada bada bada bada bada bada. The image post from my blog today has a
95. 15:28sign that says office hours with Kyle on the wall. I'm so good. That's awesome, Kathy. Thank you so much. I appreciate
96. 15:35that. Yeah, it was a good it was a good office hours today, I thought. Um. Oh, $2,000 for the AMD. That's not bad.
97. 15:44How many gigs of RAM in that thing, Todd? Is that 32 gigs of RAM for two grand? I assume they have higher than that.
98. 15:52Two grand for 128 GB of unified memory. Yeah, so >> [laughter] >> So, here's the thing. Wow, that's amazing.
99. 16:03So, a couple of things to be paying attention to, things like that. That AMD machine, two grand for 128 gigs.
100. 16:10That's very cool. Pay attention to Apple. Um, Tim Cook is stepping down. The guy that's stepping in is coming
101. 16:16from their hardware division. Apple, who's been made fun of by me and by everyone for not being in the AI game, Apple
102. 16:26is likely going to step in and and win on the hardware side of AI, where they're going to make machines that are incredibly powerful for AI
103. 16:36inference, and people will want to buy these more expensive Mac machines. I would I would imagine that that in the fall,
104. 16:45Mac will probably launch a machine that's probably in that neighborhood. Two grand, 2,500 bucks for 120 GB probably as a base.
105. 16:56Um, maybe even pre-installed with some of the AI stuff, or who knows. Um. Cool. So, if you have questions, pop
106. 17:04them in the comments here. I'm happy to talk about anything. Um, on office hours today, I talked about um, Max Q,
107. 17:13uh, which is the moment um, a rocket is attempting to leave the atmosphere. It's it's sort of hitting the speed of sound,
108. 17:22hitting the atmosphere like a brick wall, and threatening to disintegrate or bounce off the atmosphere and fall back to Earth.
109. 17:30And I think that's where we are with AI. So, I'm happy to talk about that if you want me to dig in. Um, let's see. Mary Carroll, it may be
110. 17:38fun to make fun of Apple, but they hit an all-time high of a $300 a share today. No, I'm telling you, Apple
111. 17:46Apple coming in as as an AI hardware company, essentially, um, is is a pretty it's a pretty killer move.
112. 17:57And and cuz what they will do is they will probably make So, there's a thing called LM Studio that you can use on
113. 18:03Macs or PCs that's kind of like ChatGPT, but it'll it'll show you what are the local large language models you can run on your device.
114. 18:13Pretty slick. Apple will do their own version of that software, right? So, Apple doesn't have to have good AI if if they make it really
115. 18:22easy for you to use local models that run really incredibly on their hardware. And that's that's the stuff I see coming. Producer Brandon,
116. 18:34will data centers eventually become obsolete as we improve smaller models? No. Um, think of Okay. Let me go show you something. I
117. 18:49want to go find something. Hang on a sec. No share. Yeah, I know. I'm not sharing right now. I'm just I'm finding something first. Um,
118. 19:21>> Okay.
119. 19:38Okay. So, I think you've seen this chart before. I've showed this chart before. So, all of these gray dots are people
120. 19:48who've never used AI. This all the all these dots represent all of the people on Earth. Each dot represents 3.2 million people.
121. 19:55So, this is I don't know whatever it is. So, 84% down from 87 about 3 weeks ago, 84% have never used AI. The green dots are free
122. 20:09free chat GPT users. So, they're not using a paid account. So, the yellow dots are paid $20 a month for AI, 0.3%. 0.3%
123. 20:23are using the paid version. So, 99 99. What's the other one? 2.04. 99.6% of people are using free free AI or not not AI at
124. 20:39all. So, the implication of that is that as AI capabilities get better and more people discover them and AI starts to
125. 20:52take over work and more people start using it at work, more people start using it at home, more people start using it everywhere, the
126. 20:59demand for for AI inference, the the ability to use these things is going to go up dramatically. Now, Apple and and
127. 21:10AMD and Nvidia and all of these companies that are going to make specialized hardware to to to push inference to what they call the edge,
128. 21:20so to your local computer is going to help with that. But, the hunger for for training AI and for using AI is going to skyrocket.
129. 21:35So, I don't see data centers going away anytime soon. What I do see is data centers on the ocean. So, I saw a proposal today uh of a
130. 21:46company that is doing floating data centers where the wave action generates the electricity and the ocean water cools cools the things. Um
131. 21:57Elon Musk and a couple of other companies are talking about putting data centers in space. Um there's likely
132. 22:03going to be data centers in the middle of the desert with massive solar farms. Um there's likely going to be data
133. 22:09centers next to hydroelectric dams or substations where there's spare electricity. There's like there's all sorts of plans going on.
134. 22:19Um I don't see them going away at all. And and you know, this was this was asked about on office hours today.
135. 22:29Why do we need all these data centers? Well, AI So, if if you think think steam engines and then think electrical grid
136. 22:42and then think interstate highway and then think internet and then think AI. So, all of those major technological
137. 22:54developments, steam engine, right? Industrial Revolution, electrical grid powered the Industrial revolution and brought
138. 23:04electricity to the household. Um computers, internet, um AI is going to be infrastructure level um
139. 23:18capacity for every country. So, if you don't have enough compute to do all the stuff, to run all the robots, and run all the cars, and
140. 23:28you know, the fleet of robo taxis that Tesla's going to put out and everyone's going to put out, um your country will not be able to compete
141. 23:38if your town hasn't invested in local data centers, right? Or, you know, if if your town right now is saying [ __ ] the data
142. 23:48centers, get them out of here, and and you successfully just get them out of there, then they're out of there.
143. 23:55An- another thing another thing that you could do with your town is you could say, "Hey, we're willing to talk about having a
144. 24:02data center here, but we want to know what's in it for us. How about this? You build your damn data center, but you give us a wing of it
145. 24:11that is compute for all of our citizens in this town. And we're going to give free AI access 24/7, you know, up to whatever it is,
146. 24:22a billion tokens a day to every citizen in our thing. We And we want you to provide that as a tax for you running it here."
147. 24:29Like that could be a thing that happens. So, having access to high-powered inference will literally be power. Like like electrical power,
148. 24:40right? If you live without out electrical power, you can't do as much stuff. If you're going to live without
149. 24:45AI compute, you're not going to be able to do as much stuff. So, um yes, local inference will help. I think most
150. 24:55inference for individuals will migrate out to the edge. Like we will have phones within the next 3 years that are so powerful and then
151. 25:04models that are so good that are that are quantized for phones that everyone's individual device like this will will
152. 25:11likely be doing things as powerful as what I'm doing on a 64 gig MacBook Pro right now. Like that'll be coming.
153. 25:18Um [snorts] but the high-end compute stuff that I think the the appetite for that is insatiable. Todd Waller.
154. 25:25Starting to think about solar panels to run my local models, charge electric cars and robots. Yeah. And there's there's a uh
155. 25:33there's a company right now that partnered with Nvidia and they're talking about putting essentially $150,000
156. 25:40worth of GPUs bolted onto the side of your house um that uses the spare capacity of electricity that is available to your
157. 25:49house. So every house uses one or two or three kilowatt hours? I don't know. Or kilowatts? I don't know what it is. But
158. 26:01they have capacity to for that to spike to like 48. So there's all this extra capacity. Um [snorts] and uh
159. 26:11Span is the name of the company that that that's going to put these things on your on your house. Sam the man, now that ChatGPT is
160. 26:20starting to connect things like personal finance accounts through Plaid, we Where should everyday users draw the line
161. 26:27between convenience and privacy? That's a great question, Sam the man. Um if any of us realistically think that we have any privacy
162. 26:42after 25 years of Google doing what Google does and all of the large data companies triangulating all of the data,
163. 26:50everyone knows everything. Everyone knows everything about you. They actually already know all the stuff that you're charging.
164. 26:58If you've been really diligent and not ever put yourself online and are confident that Google doesn't
165. 27:08have your [ __ ] and Facebook doesn't have your [ __ ] then maybe you'd be concerned about it. Um I don't know. Like I
166. 27:19When I was building websites in the '90s, one of the things that I got asked a lot is, "Kyle, you don't think that
167. 27:28no one's going to put their credit card on the internet, right? Cuz they're just going to steal all their money."
168. 27:35Right? And then people figured out systems to put it on the internet and you know, if they're connecting if
169. 27:41they're connecting Open AI through Plaid, um I'm going to assume I I'm assume I'm assuming it's not just going to be a sieve
170. 27:51that anyone can access my bank account, cuz they can't now, right? Um so, I assume that they're not going to be so
171. 27:58where I sit is I don't worry about it because I don't I don't have any illusions that I have any privacy right now anyway.
172. 28:07I assume that everyone knows every that all these companies know everything anyway. So, but but if you're not comfortable with it, don't
173. 28:18do it. Right? Keep using Excel and keep using whatever you use and do your finances locally. But, if you if you put your
174. 28:25finances online, right? If you use QuickBooks or if you use Rocket Money or if you use any of that
175. 28:33[ __ ] that's accessing your bank account, and one of those companies get hacked because Claude Mythos is released to the public and a
176. 28:42bunch of hackers just [ __ ] take down all the all the financial services companies like you're as much at risk in the in the
177. 28:51current infrastructure as you are with this AI stuff. So. >> [sighs] >> Kelly travels, Codex is now coming to the chat GPT mobile app
178. 29:03where you can monitor and approve coding work from your phone. Is this the beginning of truly useful AI agents? Are
179. 29:09we still in the human babysitter mode? So Kelly, this is this is exactly what I talked about on on office hours today. Like literally this.
180. 29:19Um there were four people on office hours today that talked about reaching a kind of a new level of AI
181. 29:34with their agents. One was Brent Peterson, one was Gareth um one was Vicky and one was Cindy [ __ ] And
182. 29:48all four of them talked about their agents doing actual valuable work for them that was valuable enough that
183. 29:57they're using less hours and less stressful hours to do their work. Now the requirement today
184. 30:12of getting to that point is you have to grind the hours in. You have to And Rick McCauley too. You have to grind the
185. 30:19hours in and grind the hours in and grind the hours in and and it's stressful and you can't keep up with it
186. 30:24and you just feel like an idiot and you just keep going and you grind and you grind and you grind. And at some point
187. 30:29you start to kind of get it and get what these agents are and what they do and you start to figure out that oh, I need
188. 30:35to actually understand what I want to be able to tell this agent what to do and then I have to manage it to make
189. 30:43sure that it's doing that, right? We are in an era right now that I created this graphic of a rocket taking
190. 30:53off talking about max Q, the the point at which in an ascent where it's at its maximum stress where it's just about to break into orbit.
191. 31:02I feel like that's where we are with AI right now that as remarkable as the AI tools are they still require essentially all of
192. 31:12our attention to be able to do work that's good enough to put in the world. I think we're very, very close to that
193. 31:19not being the case. I think we're very, very close to the AI being good enough that we can just hit a button and say I want you to do this
194. 31:27and it will go off and start doing that. And we can literally go have a sandwich and sit on a bench in a park and
195. 31:37get a notification on our phone, "Hey, your shit's done. All right, let me go look at it. Uh eh that's close. Um
196. 31:47can you format so that it looks like eh and then it looks like the table on page four isn't the data that we talked about.
197. 31:55Make it more like this. Click. And then go off. Then in, you know, 15 minutes later, "Eh you're done."
198. 32:06So, I think we're I think we're close. I think we're very close. I think I think that begins this year. And then I think it gets
199. 32:17relatively trivial in 2027. Todd Wheeler, I'm so tired of accepting cookies on every website. We need a
200. 32:24Chrome plug-in that just accepts them. This is one of those things. This is the thing about [ __ ] regulation right now, right? Like
201. 32:34the reason we've got all those cookie [ __ ] pop-up bot things is because at some point someone got pissed off and said, "Hey, Google stole
202. 32:44all my data." And then these cookie companies are tracking all my [ __ ] and they're not supposed to be triangulating that data
203. 32:51with this data and then and they went to their congressmen and people and they said, "Hey, they're stealing our [ __ ]
204. 32:56and they're tracking our and they and they don't tell us and no nobody knows what [clears throat] they're doing." And so
205. 33:05the lawmaker said, "Oh, no. Our citizens if we don't do this, I might not get elected again." And so they passed a law
206. 33:14that said, "Every website that uses cookies has to pop up THAT STUPID [ __ ] THING." NOBODY READS IT. Hasn't changed behavior.
207. 33:27Maybe six people out of 300 million in the United States actually took the time to read it and go, "Oh, these are these
208. 33:33are very draconian things. I'm going to reject their ability to track me and I'm going to read every agreement on every
209. 33:41website and I will reject it." Cookie Monster Frank. If you accept too many cookies, you gain weight. Um that's going to happen with AI, by
210. 33:52the way. There's just going to be some [ __ ] law passed somewhere that makes every AI model every time you you mention the word red meat,
211. 34:02we're going to get a dialogue box and says, "Do you understand that red meat, you know, can cause cancer in whatever. Uh yes, I know.
212. 34:11Sorry I used the term red meat. I'm exhausted. [clears throat] But we'll have AI to be able to write around that. We can write our own auto
213. 34:21clicker thingies. Um Okay. So if you didn't catch office hours today, um you can see office hours in demand in in the uh AI Salon. If you go
214. 34:37over here to the left, there's there's previously recorded events. Um if you go to that uh to that uh QR code, that'll take you right into
215. 34:48that section. Um also the mastermind. If if you are a member of the AI Salon and you haven't joined the mastermind, the mastermind is
216. 34:59for people that are looking to level up their game and sort of hang out with other people that are upping their game.
217. 35:05Um we've got workshops. Brandon's doing a lovable workshop. Andy Scarantino's doing a um decoupling your identity from
218. 35:12your job workshop. Um Sid Hargrow and myself do the mastermind practice lab. Um there's a lot of stuff going on in
219. 35:20the mastermind that is is worth you paying attention to. Um we're at a really intense time with AI right now, like a really intense time
220. 35:30with AI. And uh being around people who are thinking thoughtfully and critically about how to use AI right now, uh
221. 35:40will serve you quite well. And so And so that's what's that's what's going on in there. Um if you're new here, I see I see we've
222. 35:49got some some new people in. Um tonight is an AMI AMA night. That's correct, Cam. So, ask away if you've got questions. Um
223. 35:59happy to answer them in any way I can. And this is the AI Learning Lab. It's Friday night date night, so I hope
224. 36:08you're having fun and you brought a date. And if not, I'm your date. All right. Not Jim. With AI tools changing
225. 36:15so fast every week, how should a beginner decide what is actually worth learning versus just headline noise? Okay. Great question.
226. 36:23>> [clears throat] >> Okay. >> [snorts] >> So, it depends where you are in the scale. So, so Kelly Camp um
227. 36:34has gotten hired by a couple of big companies to come in and do AI education courses for these cohorts of 25 people.
228. 36:42She has all of the people self-assess how good they are with AI. She says to a person, they're all saying they're four out of five
229. 36:53in terms of AI usage. And she said to a person, they're not even a one. So, so so I think some I think people who have
230. 37:03used ChatGPT once or twice self-assess themselves, well, I know what AI is. Um so depending on where you are on the
231. 37:12scale kind of changes the answer to this question. If you're truly a newcomer and truly a beginner, um it's really quite simple. Just start
232. 37:20playing around with with ChatGPT. The free version of ChatGPT, I I would strongly suggest if you're serious about
233. 37:28learning this stuff, spend the 20 bucks a month. Spend the 20 bucks a month and you will be in the point three percent 0.3%
234. 37:39of human beings on planet Earth pay the 20 bucks a month for for some AI system. That's staggering to me. So,
235. 37:48if you want to be if you want to be one of the top one-percenters, pay the 20 bucks a month. So, I would just start there.
236. 37:55Here's the Here's the second thing and I think this one's actually really important. Whatever you do
237. 38:07and know, and like what you do for work, your instinct will be, let me go use AI to do that stuff I know.
238. 38:17And it can do that and you'll be blown away and you'll be like, okay. And you'll be able to go find AI that can,
239. 38:23if you're a copywriter, write copy. If you're a strategist, it will do strategy. What I strongly make recommend and and
240. 38:31in the AI Salon, if you go to the welcome the welcome tab in AI Salon, you'll see a graphic there that's called the cycle
241. 38:38of AI readiness. At the very top of the cycle of AI readiness, play first. Play first. What I would strongly recommend is
242. 38:47feel free to go explore AI for the stuff you know. But the real power of AI comes when you discover what what's possible beyond
243. 38:58what you know. If you're really good at numbers and you're really shitty at art, go make art with AI. Just understand what's possible.
244. 39:08If you're like, "I've always wanted to write songs, but I don't I I can't play anything. I'm not good at music."
245. 39:13But you love music. Go to suno.ai, make music. If you're good at music and you really hate math, go to ChatGPT and understand the data
246. 39:24analytics it can do. It's insane. If you've always wanted to code, but you don't have the skills to do that, go to
247. 39:31lovable.dev and type in, "I want an app that does huh." And just watch what it does. Just go play. Without expectation.
248. 39:41Outside of your domain. Now, how do you pay attention to what to do? Hang out here. Hang out in the AI Salon. People are
249. 39:52talking about Go on X. Start following people that are trustworthy about AI. And then follow the people that those people follow.
250. 40:04So, start changing your algorithm on X because what happens on X is people are talking about this all the time.
251. 40:11And if you start following the right people, what will happen is this. You'll see a post that's like you know, Higgsfield changes the game
252. 40:19for videos. You know, Hollywood will never be the same and you're like, "Oh my god, I got to go play with
253. 40:23Higgsfield." Then you're like, "Wait a minute, this user has like 400 followers. And oh, it looks like
254. 40:30they're being paid by Higgsfield to say that. But then some other thing will happen and you're like you see people
255. 40:37talking about it and you're like, "Oh, that's actually a person I respect that seems excited about this. Oh, there's
256. 40:42another one. Oh, there's another one. Oh, there's another one." So, you start to see patterns of people that you know
257. 40:48know what's going on with AI talking about something over and over and over again, that's a thing to go pay attention to.
258. 40:56So, don't ignore the headlines and the pretty graphics and here's a video that was made in 30 seconds and it Hollywood's dead.
259. 41:04Most of that [ __ ] is just [ __ ] But when you start to see patterns, that's what to play with. All right?
260. 41:20Literally just had this happen with Higgsfield. The guy only had 5K views and I was like, "Eh, okay, maybe not the person to listen to."
261. 41:28Yeah, Higgsfield like Higgsfield's got some decently powerful stuff, but it's it's kind of like um like Creia where
262. 41:36where it's a centralized location where you have access to a bunch of different models. And they charge a lot for it and they
263. 41:42just they they just market the snot out of it. Um and so like I like I've just gotten to the point with Higgsfield where I'm
264. 41:50like, "I don't really give a shit." Because their marketing is so sleazy that I'm just like if it really becomes something big a
265. 41:59year or two or three from now, fine. Whatever. Eh. Some people like it. It's fine. AI curious, OpenAI, Anthropic, and
266. 42:09Google and others seem to be racing toward enterprise AI and agentic tools for regular workers. Does that mean AI becomes a
267. 42:17helpful teammate, a productivity surveillance layer, or both? You mean for regular workers within the
268. 42:25companies that are going enterprise AI? Okay. Um It's going to be all over the map.
269. 42:43So, >> [sighs and gasps] >> yeah, like we're in such an exhausting time right now. Most big corporations have no idea what generative AI is.
270. 42:57I think at this point most CEOs are probably saying, "Use AI. Go use AI. Lay off. We need to lay off 15,000 people. Just tell them AI
271. 43:06did it. Go lay them off and figure this AI stuff out. Make us profitable." I think there's a lot of that going on right now.
272. 43:16As a result of that, I think you have a lot of IT groups leading the charge of a rollout that should not be led by the IT group.
273. 43:28Should IT be involved? Yes. Should they be leading it? No. But, Kyle, it's AI and it's technical. NO.
274. 43:40I'VE TOLD THE STORY BEFORE in here about when I was at agency.com how we won the Coca-Cola business. Coca-Cola was looking for someone to
275. 43:50build their websites. And when you went to coca-cola.com and coke.com, there were HTML pages that the IT group had published
276. 44:05that were like light yellow with tables and hyperlinks to annual reports. And then that was coca-cola.com and then
277. 44:16coke.com had a white background and tables and links to other documents. This is the world's most powerful valuable brand.
278. 44:32And coca-cola.com and coke.com had been created by the IT group and were shockingly embarrassingly bad. And so we went to Atlanta
279. 44:46with a slide in a PowerPoint deck that I was I was waxing poetic and waving my hands like one of those air dancers.
280. 45:03And we showed a screenshot of their website. And then a headline that said where are the polar bears? Cuz the polar bears
281. 45:12were the big part of the campaign back then. And the executives at Coke stopped the meeting. They left the room. They said hold. Hold.
282. 45:24Can you just We'll be back. They all left. They had never looked at their website. That's happening right now with AI. So
283. 45:36one of the things that's going to happen is if you work for one of those companies where senior management is delegating AI to the IT group
284. 45:48I would start looking for another job, personally. Because they're going to [ __ ] it up. And yeah, they're going to use it for things
285. 45:55like well, you know, what we could do with AI? We could surveil all of the employees and see who's productive or
286. 46:00not. We can measure productivity with a So yes, some of the companies are going to do shitty shitty shitty stuff.
287. 46:09Because they actually don't understand what it is. They don't understand what it makes possible. They're not thinking
288. 46:13about it critically. They're not thinking about the impact of the human beings who are going to have their jobs completely redefined by AI
289. 46:22and their sense of identity and value stripped away from them. You know, like but the company's more efficient now. Hope you enjoy your job.
290. 46:33So there's going to be a lot of that. The next tier is going to be companies where senior management actually kind of
291. 46:41has a clue and they recognize, oh this is a big deal. We got to take this seriously. And so they're going to be using it for
292. 46:47efficiencies, but they're also going to be doing some innovation. Maybe they're going to be looking for innovative employees.
293. 46:54They're basically going to keep doing what they've been doing, but they're going to now do it with AI
294. 46:59and they're going to get a little bit of a little bit uh innovative in in how they do what they do, but they're
295. 47:04basically going to be on the same same course. Those kind of companies will likely also understand that we got to take care of
296. 47:12our people and make sure that in the change management they have a voice in how their roles are evolving.
297. 47:26And then there's going to be a third kind of company. And I'm talking about when I'm When I'm When I'm talking about
298. 47:30here is relatively big companies. There's going to be a third kind of company and that's in my seven economy
299. 47:36slide, it's what I call economy five. And economy five is going to be really interesting and really challenging, but I think would be a
300. 47:45fascinating place to work. An economy five company is going to look at AI as a capability and they're going to say, holy [ __ ]
301. 47:56We can actually completely transform our business model, our operating model, our go-to-market strategy, our customer engagement stack,
302. 48:07how we engage with customers, who does what, what the roles in the company are. We're going to completely rethink
303. 48:14how we do what we do while we're running the company. I think those are going to be very rare. Um if you want an example of that, Jack
304. 48:25Dorsey's Block right now is doing that. They're the only company that I know that are full-on experimenting with their business model, their
305. 48:39employment model, their customer engagement model. They laid off 4,000 of their 10,000 employees, which you can be cynical about,
306. 48:49but if you look at what they followed it up with, they Jack Dorsey wrote an article, which is worth you reading, called um from hierarchy to
307. 48:58intelligence, where he gives you the history of middle management, starting with Roman soldiers, going all the way up to modern times,
308. 49:08and says, "Every large organization requires middle management. The problem with middle management is that humans are slow communicators,
309. 49:17so to get something from down at the bottom of the organization up to the top, there's 17 layers or 700 layers,
310. 49:26and they're inefficient at every step of the along the way." And he said, "We're going to get rid of middle management.
311. 49:32We're going to replace that communication routing to the AI. Let the AI handle that cuz it's intelligent enough to do it. Break
312. 49:41the company into small eight-person pods. Have three roles. Think of it like a ring of people that interface with customers, so you have
313. 49:52people interfacing with people and then AI doing the routing of who does [clears throat] what within the organization.
314. 50:01Is that going to work? I don't know. Jack Dorsey doesn't know. He ended that article by saying, "This is a grand
315. 50:07experiment. It will likely fail, but we will learn things along the way." That's a company I'd want to be a part of. So,
316. 50:22here's my guess. 60 60 to 70% of companies are going to do this horribly. [snorts] 25% of companies are going to do this
317. 50:39okay. 5% of companies are going to really try to revolutionize what they do. Those are existing companies. That's That's my guess.
318. 50:5070% will just [ __ ] it up. They'll [ __ ] it up for a decade. 25% will figure out some version of it working. 5% will really transform
319. 51:01their businesses, and of the 5% that really transform their businesses, a third will succeed, two thirds will flame out.
320. 51:11But I But I'd much rather be in that [ __ ] experiment than in one of the 60% companies. And then you're going to have all the
321. 51:20people that got laid off now have access to AI. So, they can all team up and go make small little nimble companies powered by
322. 51:29AI that can 10x their personal impact. So, if you have 10 employees that are 10x'd, that's effectively a 100-person company
323. 51:40going back and competing with the company that laid them off. Would love to work in one of those companies. So, that's what I see going
324. 51:47on. All right. Any thoughts on Figure 1 package sorting robot live stream? Um yeah. >> [laughter] >> So, if you haven't seen it yet, is it
325. 51:58still going? Let me see if it's still going. Um Figure live stream.
326. 52:14Okay. Yeah. Yeah, well. Oh, wait.
327. 52:40Okay. So, 57 hours 20 minutes. So, that's Jim. So, there's three robots. So, I'm I'm guessing they can work for 8
328. 52:51hours on a charge is my guess. These two guys are charging or maybe it's 4 hours. It doesn't really matter.
329. 52:57But, they're they're rotating they're rotating three robots. This is Jim. Um They've sorted 72,000 packages. It looks
330. 53:05like the sort is they grab a package and flip the label down. That seems to be what they learned to do. Um
331. 53:15And this is a live stream. There's 22,000 people watching this. Now, um where this started out, where that thing started out, hang on,
332. 53:30let me stop it cuz it's annoying. Um where that thing started out was it was they they said, "We're going to stream live for 8 hours
333. 53:42of this robot sorting packages. And so for the first 4 hours people on Twitter were like, "Oh, no. That's they're being
334. 53:53uh whatever, uh managed by humans that are running them and you can see them adjusting their AR goggles and this and that." Um
335. 54:02people do not want to accept that that is coming. So they they got through 8 hours. I assume why they said an 8-hour thing was
336. 54:14because they were confident they could get to an 8-hour thing. I don't think that they were necessarily
337. 54:21confident that they were going to just let it keep going. But they got to 8 hours and they said, "Well, why don't we just let it keep
338. 54:27going?" And so they're now at 57 hours and I think they're now at the point where they're just going to let those three
339. 54:32robots keep sorting those packages 24/7 until one of them breaks. There are there are hilarious moments
340. 54:45where there There's a moment, I don't know if you see it. The the robot is sort of doing its thing. He's flipping packages and then at some
341. 54:53point the robot just goes it like grabs imaginary motorcycle handles >> [laughter] >> and um and is like daydreaming about riding a
342. 55:09motorcycle for like 30 seconds just like >> [laughter] >> and then just goes back to sorting packages. So apparently these things can
343. 55:18daydream just like humans can. Um All of the humanoid robot the major humanoid robot companies are on their V3
344. 55:31the their their V3 of their of their um their devices their their robots. Um and V3 are the ones that are
345. 55:41manufacturing ready and they're refined and a lot of the work is being done on the hands with touch sensors and things like that. Um
346. 55:51I think we're probably a year and a half out from those things being fairly commonplace in situations like sorting packages and
347. 56:05probably 3 years out from those things being commonplace in the world. Like it's not going to be weird to be in
348. 56:12a grocery store and have a robot walking by you shopping for [ __ ] that you're shopping for. It's odd to me that they design robots
349. 56:20in a human form, so inefficient for to me for for most jobs. But here's the thing, Kelly. I So the answer over time is yes. The only
350. 56:29reason that a kitchen is shaped the way it is is because human beings need to be in it. Our entire infrastructure for the most part
351. 56:41is designed for humans. So the reason to start with humanoid robots is that they can just go into existing situations and function.
352. 56:54But what will likely happen is that that infrastructure will start to die out and start to be replaced by more
353. 57:02efficient forms, right? So So you have some sort of kitchen where there's no counters, like everything is prepared at
354. 57:10floor level. So you just have, you know, it it effectively becomes a warehouse with cooktops on it. And you'll have very efficient things
355. 57:18that go around and clean clean up after things and then other things that, you know, prepare food and then other things
356. 57:24that cook it and other things that package it and other things that ship it. And it'll all be designed incredibly efficiency
357. 57:33um from the ground up. But But we've got a world right now that's not designed for that. And so so there will be probably a 20-year
358. 57:42transition. Um So like an interesting an interesting robot company to look at is a company designing robots
359. 57:54for that future, right? The humanoid robots are transitional technology. And then, you know, the next evolution will be
360. 58:05spaces and environments that are designed by AI, built by humanoid robots, and then populated by specialized robots, right? Um
361. 58:17All right, how we doing on time? Great. Let's see. Um let's go take a look. My My screen share doesn't last that long, but let me
362. 58:26let me just pop it pop it on here for a second. Um I just want to get give you a quick little I know you're in the in the AI
363. 58:33Salon. We're all in the AI Salon right now. But if you're new here, um when you come into the AI Salon, you
364. 58:41start with the the welcome screen. And here's the I mentioned this earlier, the cycle of AI readiness. Um so this is
365. 58:48what this came from is how people in the salon that we're doing really interesting work, how they worked anyway.
366. 58:54Um and so we talk a lot about that and there's some welcome stuff there. Um the AI Learning Lab Live is right there.
367. 59:00There's information about the mastermind here. Um if you want to understand our values and what we care about in the
368. 59:05community, um how we want to treat one another, how we want to show up, that's in our values. Then we've got an events and connect
369. 59:13section. So the community feed and the community chat or you can just chat with people. You can talk about anything.
370. 59:18It's a free for all. Um that's where um I put up a post uh today about running local models and the hardware and the models you can run. So,
371. 59:28that's there. There's a play and create section. We talk about play first. And so, if you want to share share uh show
372. 59:35things off, go to look what I made. There's challenges and competitions, which you can find out other uh
373. 59:40challenges and competitions or put in your own. You can ask for help. There's a tribute to to Serena, who was a uh
374. 59:47really amazing woman who who was an irregular from day one on on this very uh channel on the AI Learning Lab live.
375. 59:54Um there's a learn and grow section, and then there's the mastermind. And we've got the mastermind practice lab, the
376. 1:00:00great repurpose, we've got AI prototyping in there, tool talk, things like that. Um in learn and grow, we've
377. 1:00:07got AI for business, we've got AI uh life hacks uh that Brandon and uh Brandon and Claire run. Uh you can watch
378. 1:00:14past meetings there. And then we've got advocacy and governance, and then partnership and promotion. So, there's
379. 1:00:20all sorts of things to check out here. So, if you haven't dug around, uh you should do that. just go to the events
380. 1:00:27tab at the top, and there's lots and lots of things coming up. So, go check out what's coming up. And uh yeah, that's about it. So,
381. 1:00:36um please join us. Oh, and it looks like my share worked for nearly all of that time. Danica S. There's so much technical babble. Yes,
382. 1:00:46it's flipping horrible. What AI might actually change how normal people use AI. Great question. Um So, all of the stuff you're hearing
383. 1:01:00right now about agents and even the stuff I'm talking about, local models and all the all that stuff. That's that's not going to that's not
384. 1:01:14going to hit the masses. Um Uh >> [sighs] >> I think a couple of things need to
385. 1:01:29happen. So, right now the opportunity in AI right now is if you're a little bit geeky, if you're willing to put in a little bit
386. 1:01:41of time, if you're willing to learn what ChatGPT can do and and what what its image model can do and what you can do
387. 1:01:47with programming with it and and and just be willing to explore a little bit, you'll find that everyday people can use
388. 1:01:58ChatGPT or Claude or, you know, any one of these models that the Gemini models from from Google, they can use them to do remarkable things.
389. 1:02:12And like really remarkable things. Like like 10x their output. Like run a single person company and and execute as if
390. 1:02:20they had a full-on marketing department, a full-on finance department, a full-on this, a full-on that.
391. 1:02:26The people that can do that already have a tremendous amount of agency, they have a tremendous amount of entrepreneurial drive,
392. 1:02:34they have a tremendous amount of curiosity, they have a tremendous amount of adaptability, willingness to play.
393. 1:02:40So, if you've got all those attributes, you can do all the amazing stuff with AI right now. It's maddening, you can't keep up with
394. 1:02:48it, it's jankier than crap. But, what's going to happen is this. The problem with AI today is that it is
395. 1:03:02it's remarkable enough to be remarkable and have people go, "Oh my god, I've completely transformed my life with this thing." But, it's
396. 1:03:11it's not good enough that you don't have to baby sit it all the time. Like right now AI, you have to babysit it all the [ __ ] time. And if
397. 1:03:21you start something like Open Claw or these agents or Codex with ChatGPT where it can run 24/7, it's a nightmare. You've got to babysit
398. 1:03:31your stupid flippant It's just it's a nightmare. So, if you don't have the appetite to babysit shitty technology, it's a little early.
399. 1:03:43But I think starting in 2026, I think toward the end of this year, the AI models are going to get so good
400. 1:03:51that the only thing you'll need to know is your goal. Huh, I kind of want to start a business about yeah.
401. 1:04:01And then you say to something that will likely be initiated on a phone. You probably won't have to log into some
402. 1:04:09big fancy interface and learn a bunch of different tools. You'll probably have, you know, some app that you downloaded that you make an
403. 1:04:17account on. This could be from ChatGPT, could be from Anthropic, could be from Apple, could be from any number of places.
404. 1:04:25Sitting behind it are 27 different models and an orchestrator agent and all sorts of [ __ ] like that that end users will never see and will
405. 1:04:33never know about and will never have to care about. And then you'll just go, uh I kind of, you know, I was watching
406. 1:04:42The Sopranos and they were into the import-export business and they seemed to make a lot of money. That and garbage.
407. 1:04:48So, I want to do an import-export business. And then your little thing will decide, I'm going to go spin up a
408. 1:04:54researcher and a business strategist and a and a market analyst and I'm going to go find what sector of the business of
409. 1:05:03the import-export world is interesting right now that we can make some money on. And then it'll go do that work and
410. 1:05:08it'll come back to you and go, here's Here's six options. Oh, I can do furry animals, fur furry plushies. I can do plushy I can do
411. 1:05:19a plushy business. Yeah, you can do a plushy business. Okay, let me start a plushy business. And then it'll spin up some agents to
412. 1:05:26go, you know, it'll know who you are cuz you've been talking with it for a while. So, based on who you are, it'll come up
413. 1:05:32with a brand that it'll think you like and it probably you probably will. It'll come up with the brand. It'll come
414. 1:05:38up with the products. It will negotiate with the factories in whatever country you're going to manufacture in and you
415. 1:05:45could probably say, "I want this to be ethically sourced. I'm willing to spend a little bit more and make a little bit
416. 1:05:51less profit, you know, if we're not using child labor." Right? [ __ ] like that. It'll go find all that. It'll negotiate with the with the
417. 1:05:59vendors. It'll figuring out It'll figure out the shipping. It'll figure out the marketing strategy. It'll figure out what sites to put you
418. 1:06:06on. It'll build your own website. Like, when it can do that, that's when everyone will come. Right now, you can do what I just described
419. 1:06:23with agents that will do it in this fully automated way, but it's going to take you 3 months, 6 months, whatever it is, of
420. 1:06:32literally slamming your head against your computer keyboard 3, 4, 10 times a day, bloody eyed, sleep deprived, stressed out
421. 1:06:45for the Shangri-La of AI. >> [laughter] >> Valerie Cox, I work with a lot of analytics. I think current AI technology
422. 1:06:55is almost fantastic with working through difficult data structures. Almost. Exactly. 3 years ago, it was like not even close
423. 1:07:06to to being good. It was impressive until you tried to use it. Now it's like impressive until you try to use it and then
424. 1:07:17deliver it in a way that you could give it to your CEO. >> [laughter] >> Right? And then you're like, "Uh, not quite."
425. 1:07:25Still got to get in there. >> [laughter] >> But we're we're really close. We're really close. What's powerful about being in a
426. 1:07:38community like the AI Salon is that you've got people in here that are that geeky, that are willing to experiment, that are willing to just
427. 1:07:47play with stuff. Um people like Jim Ross who's in a in a in a sector self-storage that it's not technical at all, but he he explores
428. 1:07:58this stuff all the time. So you you get to hang out with people in here that are that are willing to do that. And they're also
429. 1:08:04really generous and willing to help. Like everyone here that's who we call the irregulars, the people that show up
430. 1:08:09to these things regularly, which you're like, "That's weird." I'm like, "I know." They're irregular. They're not normal.
431. 1:08:16They're also some of the nicest, most generous people you'll ever meet. So they'll help you, they'll give you a leg up.
432. 1:08:25So, I would encourage you and anyone who's thinking about it right now, start playing. Understand that where it is right now,
433. 1:08:34you're going to have to be a little geeky, but that's where the that's why there's opportunity. Right? You're willing to put in
434. 1:08:42the elbow grease, you know, you get there before other people. And even if you don't succeed right now, you will understand where the
435. 1:08:52limitations are. And when the models show up that are really good, you'll be able to go, "OH, [ __ ] WE'RE HERE. Let's go."
436. 1:08:59I know what this is. Even if you feel like you do not belong in a community about AI, the generosity of the community will keep you coming
437. 1:09:10back. Oh, that's very sweet, Mary Carol. That's true, though. That's true. This is a group that is like weirdly defensive
438. 1:09:21about newcomers that come in and make sure they're taken care of. Like defensive not in a bad way. Like
439. 1:09:27they they will defend newcomers. Like come on in, we got you. We got you. Trolls come in. They're like, "No, no.
440. 1:09:35No, you're not trolling in here." >> [laughter] >> Gareth in the house.
441. 1:09:50What else we got? Mhm. Next week I am without producer Brandon on Monday and Wednesday at least. It may be Friday.
442. 1:10:02So, next week I'll be here, but I'll be on my own. So, bring your patience. Protective, that's the word I was
443. 1:10:09looking for. Thank you, Claire Jacobs. >> [laughter] >> Defensive. It's the most defensive community in all of the land. Uh >> [laughter]
444. 1:10:23>> Daniel H, have you played with the new CreA 2? I have not. So, CreA is is a really good tool. It's a tool that
445. 1:10:30allows you to play with lots of other models. I I used to demo it in here a fair amount, and I just haven't in a
446. 1:10:35while. Um and then CreA 2, that's their new image gen model or is it image gen video? I don't remember what it is, but
447. 1:10:41no, I haven't played with it. >> [clears throat] [snorts] >> Is it good? Have you been playing with it, Daniel?
448. 1:10:52Oh, the other thing that's happening next week that's really cool, um Zendesk, which is where producer
449. 1:10:58Brandon works, is having a big event in Denver, and so producer Brandon and I are going to spend some time together
450. 1:11:04next week. It's really cool. I'm really excited about it. It's like mid-journey. Oh, wow, interesting. Cool.
451. 1:11:12And we get rabbit holes, right? Yeah, this this channel is nothing but rabbit holes. Rabbit holes are important. Um,
452. 1:11:22you know what's funny is is there there will be a time I don't know how long it'll be, 3 years, 5 years, there'll be a time where
453. 1:11:32this AI stuff's good enough that you don't you we won't really need a group like the AI Salon. Um, but what's going to happen is, and this
454. 1:11:43is what we talk about in the Great Repurpose, if you if you're part of the mastermind, the programming that we're putting
455. 1:11:49together for the Great Repurpose is basically like whether you lose your job or keep your job, all jobs are about to transform.
456. 1:11:58And they're going to transform in ways that if you have your identity tightly bound to what you do, I'm an accountant, I'm a lawyer, I'm a
457. 1:12:10Ivy League grad that worked at these fancy companies. AI's about to just decimate the tasks
458. 1:12:24that everyone in every profession does. And not decimate them like destroy the tasks, it's going to decimate them like, oh, it the AI will
459. 1:12:38just handle the tasks now. And if you've spent 30 years of your life learning how to do those tasks, it's going to feel like a loss.
460. 1:12:50And so, that's what we're all about to confront. So, one of the reasons I feel so passionate, one of the reasons we
461. 1:12:55brought the AI Learning Lab back inside the salon is I want to make sure that as a community, we are connecting with one
462. 1:13:04another because we're we're entering what I think is going to be a really intense 3 years. And it's just going to be 3 years where
463. 1:13:12like, literally everyone is going to be like, wait, what? That's my job now? Or like, wait, what? That whole sector doesn't exist anymore?
464. 1:13:21What are you talking about? I went to school for that. What am I supposed to do? Oh, you're an entrepreneur now. Here's some AI [ __ ] to
465. 1:13:30help you do it. I I don't want to be an entrepreneur. Well, your industry is gone. Right? Like, that's what we're entering.
466. 1:13:39We will always need the AI Salon. That's very sweet. Oh, there's a subreddit for Windows XP that gets 39,000 visitors. We're going
467. 1:13:49to be fine. That's hilarious. Gareth Hood, um I've had a job I've had a job, lost a job, got a new job all
468. 1:13:57while being part of the AI Salon. Yeah, exactly. Well, that's you you know, it's funny. I used to think that being early
469. 1:14:04in AI, the advantage was going to be about understanding the technology. What I now think the advantage in being early
470. 1:14:12in AI is is that we go through the existential angst first. That [laughter] That we're going to get it out of the way early,
471. 1:14:21right? Like Like the amount of people that are either laid off and looking or, you know, coming out of retirement or
472. 1:14:29trying to figure out what their next thing is or decided they want to completely change their career. Like,
473. 1:14:34the AI Salon is full of people like that. And so, when new people come in and they're like, uh I just got laid off. Uh not a problem,
474. 1:14:42come on in. >> [laughter] >> Right? AI Salon is a living and breathing support and networking platform. It
475. 1:14:48won't go away, it will grow. Oh, that's very sweet, Valerie. I agree. I think it's it's a really amazing community.
476. 1:14:53Todd Waller, I used to sort packages. Yeah. I and you know, there's just this weird hubris right now. There's there's all
477. 1:15:04this anti-AI sentiment, which is growing, right? It's you know, AOC and Bernie are like, let's block all of the data center projects. Um
478. 1:15:15Commencement speech givers that mention AI are being soundly booed off the stage by the students. HEY, BOO AI. AH.
479. 1:15:26SO, THE RESISTANCE IS INCREASING. The willful ignorance of AI is increasing. I'm not going to do it. You tech bros with your stupid
480. 1:15:42>> [laughter] >> That's increasing. And the tools are getting so good that we're just on the cusp of them being able to
481. 1:15:53just do work for you 24/7. I have an idea for a business. Sure thing, boss. Three days later, here's your functional business.
482. 1:16:07A month later, you're starting to turn a profit. Like, we're within a year of that. And I don't want to.
483. 1:16:1895% of the people, I don't want to. Nah. Really? Staggering.
484. 1:16:34So, what we're doing here is just talking about this [ __ ] Do 40,000 acre data centers suck? Yeah.
485. 1:16:49Are they going to use up too much energy and too much water? Yeah. Do we not have enough electricity to make them all function? We don't.
486. 1:17:00Is any of it going to stop? No. Now, [snorts] people in communities like this might get together and say, "Hey, we
487. 1:17:11understand we need data centers, but let's roll them out smartly. Let's find ways that those data centers
488. 1:17:17can actually benefit a community rather than just exploiting it or exploiting the land around it. Let's find creative ways to
489. 1:17:27put these data centers in the ocean and in ways that are environmentally neutral. Like, yes, do all that. Let's have AI help us do that.
490. 1:17:45But we're at we're we're we're approaching max Q. We're approaching that point where either the rocket's going to disintegrate,
491. 1:17:54it's going to bounce off the atmosphere and crash back to the earth, or we're going to break through the atmosphere and start floating.
492. 1:18:04We're we're right at that very stressful, maddening, dangerous moment.
493. 1:18:21It's exhausting. Valerie Cox, everyone is so exhausted. I've been asking people, "If you don't need to work to support
494. 1:18:31your basic needs, what would you want to do with your life?" That question, like That question Brandon just said what's what's the
495. 1:18:43weekend homework? I like to give homework over the weekend. That's a really good question for weekend homework. If you could do anything
496. 1:18:59what would you choose to do? Think about who you are, think about what you value, think about who you care about.
497. 1:19:09Look at those three things. Well, this is what excites me. Here's the things that are really important to me.
498. 1:19:20And here are the people that I care about. Given all that and if I could do anything like if my limitations weren't limitations I would X.
499. 1:19:33If you know what X is if you're clear about who you are, what you value, who you care about and have some idea for how you might
500. 1:19:44want to change the world for that combination of goodies you will think about AI in a completely different way.
501. 1:19:55You won't think about AI like how can I use AI? What am I supposed to use midjourney for? I'm not an art person.
502. 1:20:03I don't like spreadsheets. Why would I use a spreadsheet? Well, if you understand here's who I am, the things I I get
503. 1:20:11excited about here are my values. This is This is what I want to make sure if I'm going to do something, it's going to live up to that.
504. 1:20:19Here's these people that I care about. It's my family, it's a community center, it's my city, it's the world. I want to
505. 1:20:24change the world, baby. Doesn't matter. Who do you care about? And then have an idea and say, "Yeah, I want to do that.
506. 1:20:39And I'm going to learn AI in such a way that it's going to help me do that. Because I'm not good at finance and I'm
507. 1:20:48not good at marketing and I'm not good at this and I'm not good at that, but I sure am passionate about this.
508. 1:20:53I know what I want it to look like and I know who I want it to help." That's your weekend homework. If you want to spend some time this
509. 1:21:07weekend, think about that. Who are you? Who do you care about? What's the difference you want to make in the world?
510. 1:21:16And write that [ __ ] down. Put it on post-it notes. Stick it around the office. Clare Jacobs, tell me what it is you plan to do
511. 1:21:23with your one wild and precious life. Yes. At a time in history, Mary Oliver. >> [clears throat] >> At a time in history where we have
512. 1:21:38a profoundly intelligent entity that is like a golden retriever ready to do what we ask it to do.
513. 1:21:53Dario Amodei describes it as um a city of geniuses in a data center. Imagine having a hundred PhD level geniuses
514. 1:22:07that work 24 hours a day, 7 days a week to do anything you want them to do. We're 3 years away from that.
515. 1:22:17That that commonplace will be a device where you can deploy a hundred PhD level employees to do anything you want them to do
516. 1:22:31that are probably attached to some physical humanoid robots. They can do physical stuff in the world, too.
517. 1:22:38Let's assume you had that resource. Would you know what to do with it? If you don't have an answer for that in the next year or two
518. 1:22:48>> [clears throat] >> it's going to suck. Cuz cuz your little 14-year-old neighbor Jimmy is going to have that [ __ ] figured out.
519. 1:22:59Jimmy is going to be making $20 million a year cuz he told his phone to go do something. You're like, "Ah, I probably should have
520. 1:23:08learned that AI shit." >> [laughter] >> Longboard surfing looks so zenful. Yeah. >> [laughter] >> Anyway, all right. I would live in
521. 1:23:21economy one, tend to garden, ride horses, live off the grid. I I'll tell you what. So, my the thing that I am most excited about is
522. 1:23:29I am excited about people that go all in on economy one that are AI literate. Right? I'm going to go
523. 1:23:40grow blueberries and I'm going to use AI to do something with them. I I don't know what it is. It'll run the business. It'll
524. 1:23:48do the marketing. It'll I'm just going to play with the blueberries and then it'll go do all the other [ __ ]
525. 1:23:55And it's going to fund me with my organic blueberry farm. I think [clears throat] that's a that's going to be a really exciting sector.
526. 1:24:05Like think about that. You're willing to learn AI, but what you really want to do is go play in the dirt. Cool.
527. 1:24:12How can AI help you go play in the dirt? Who are you? What are your values? What do you value? Who do you care about?
528. 1:24:21What's the change you want to make in the world? For yourself, for the people you care about. We live in a moment in history we can
529. 1:24:33just do that. It's Friday night. We're having a date. We're talking about AI. You're There
530. 1:24:48There are 25 people here on a Friday night at 10:54 p.m. in New York on the East Coast thinking about what they're going to do
531. 1:25:02with AI. This It's It doesn't make sense. But yet we live in this remarkable moment. Solarpunk era. Yeah. Yeah.
532. 1:25:16Do you know what Do you know what the the the the world does not is not going to have near enough of? Electricity.
533. 1:25:25You want You want to go solve solar farms? You're going to have customers for life. >> [laughter] >> You know?
534. 1:25:38Every Everyone's They're planning their communal farm. You know what we're going to have? We're going to have a communal
535. 1:25:43farm where we're going to all have little cottages. We'll have a central little farming area. Some people do the
536. 1:25:47farming and then we're going to have a server building. And the server building's going to have just racks and racks and racks of We're
537. 1:25:54going to be a self-contained AI supercomputer that serves the community and does everything we need it to do.
538. 1:26:00We're going to be sitting in hammocks and slinging back drinks. All right. And we will have an actual AI salon.
539. 1:26:08Yeah, we'll have a whole building. Okay. I need to get out of here. I got to go pick up Gabby who's at a gallery opening. Uh
540. 1:26:18brand new gallery here in Denver, the Rollo Gallery. Really excited about it. She's in it. Um All right.
541. 1:26:28Let's see. I did not know uh Peruvians Okay, you're Those are internal conversations. Cool. Is Champy going to
542. 1:26:35be there? Champy will be on the farm, yes. Champy will be on the farm. Um Listen, thank you all so so much. Um
543. 1:26:43please, if you're new here and you haven't introduced yourself, go to the introduce yourself section in the AI
544. 1:26:48Salon. Introduce yourself. Check out the other areas. Share your work. Ask for help. If someone asks a question, answer it
545. 1:26:58to the best of your ability. Just connect. Just connect. Go to the events. Keep coming to these. All right? Happy
546. 1:27:07Friday night, everybody. Um happy Friday night date night. I love you all. I really appreciate you all. I'm so
547. 1:27:13excited to be back in the in the AI Salon. Um Monday is AI for business. AI for business has a meeting, so go to the
548. 1:27:21events calendar and look for that, and then I will be back here Monday night sans producer Brandon. He will not be
549. 1:27:27here. All right? Peace out, everyone. Have a beautiful weekend. Bye.