05/08/2026 - Choosing the Right AI Tool Understanding the Strengths of ChatGPT Claude and Gemini

Transcript

1. 0:03[music] [music] [music]
2. 0:25[music] Friday night, date night. Hello. Hello. Hello. Good evening. Good people. Let's see.
3. 0:42Oh, there I am. Beautiful. Hey everybody. What's shaking? What's going down? Happy Friday night date night.
4. 0:48Welcome to the show. I hope everyone is doing well. Um, this is the new AI learning lab live taking place inside the AI salon. I'm
5. 1:01super excited about that. If you're new to the AI salon, go check out all the things on the left hand side of the
6. 1:06screen. I don't know which one that would be for you. Maybe this one. I don't know that one. I somewhere
7. 1:12wherever the menu options are, just go look at those. There's all sorts of cool things there you can play with. Um, this
8. 1:18stream you can actually pop out somewhere in the upper I think right hand corner of this live stream. You can
9. 1:23pop this window out and go look at another tab. Lower right. It's like the upper right except it's the lower right.
10. 1:32So the lower right there's a little pop out button. It'll pop this window out and you can you can then have other
11. 1:37windows that you can go look at stuff and keep this stream live. So, um, if you're if you're on Apple or have a Roku
12. 1:44or something like that, you can, you know, broadcast this to your big big screen TV. I don't know why you'd want
13. 1:49to do that considering this, [laughter] uh, but you can, but welcome. Welcome. My name is Kyle Shannon. This is the AI
14. 1:58learning lab. Um, it is, uh, Friday night, so we'll just hang out, have a conversation. Uh if someone wants to hop
15. 2:07up on stage and have a conversation, we can do that. If you have questions, um if you want me to demo something, uh ask
16. 2:14me questions, things like that, happy to do that. So with that, let me go look at the comments. Hey yo, what's happening?
17. 2:24Beautiful. Welcome everybody. So here's a question. Here's a question. How do I know when to use chatebt, Claude, Gemini, or
18. 2:36another AI tool? Aren't they basically the same? And do they have strengths? Well, okay, that's a great question. Um,
19. 2:55if you're just getting started with AI, it really doesn't matter which one you use. Um, because part of the part of the
20. 3:06skill of of learning to use a large language model. So, chatbt is the most famous one. Anthropic is the sort of
21. 3:14darling recently. Um, Gemini from Google's really good. Um, what was the other one that was mentioned here? Oh, another AI tool. Um,
22. 3:23they all basically do the same thing. You put in a prompt, they generate an answer. If you're if you're new to these things,
23. 3:31one of the things that you learn with a large language model is that it's actually very different than a Google
24. 3:37search. What Google has trained us over 20 years is to sort of treat these interactions like a vending machine. I'm
25. 3:46going to put in a short little prompt and it's going to give me, you know, a bunch of links and then I'm going to go
26. 3:50deal with them. Okay, that's perfect. That's great. Um, with large language models like chat GPT, they need a lot of what's called
27. 4:02context. They need a lot of like [laughter] information about what you're talking about. The challenge with large language models
28. 4:12is that they've been trained on like all the crap, right? All the crap on the internet. Um, in fact, it's literally
29. 4:20what what the what these uh frontier model companies do is they have robots that go out and just scour web pages and
30. 4:28grab all that information and they do this magic mathematical thing that gets all of the information and it it encodes it in a certain way
31. 4:37that it makes it instantly searchable. It's actually quite remarkable that um you can just query the knowledge of
32. 4:45everything. The problem with that is if you've got all of the knowledge out there, if you ask a simple question like
33. 4:52tell me about this particular math formula, for example, I don't query about math problems because I'm shitty
34. 4:59at math and I don't care about it, but some people like that. But if you query about a specific math problem with no
35. 5:05context, it will not only get the answers to the the the right answers to the problem, it will also bring in the
36. 5:12wrong answers to the problem. And it'll bring in conjectures about the problem, some of which are great, some of which
37. 5:17are horrible. And it sort of amalgamates them all together and will give you some answer that will look very right. Um,
38. 5:26and often it is very wrong, right? Because um it's what's called hallucinations. it just hallucinates. It
39. 5:32It has put things in there that it just made up essentially or it it pulled from you know different different avenues. Um
40. 5:41if you give it more context like hey I want to talk about this math problem from this professor and you know I I
41. 5:49want you to act as a mathematics genius and things like that. It will narrow your question into a into a more narrow
42. 5:58thing. So the general answer is if you're if you're just starting with large language models, it doesn't matter which one you choose.
43. 6:06Think of the conversation more like a conversation than a vending machine. So if it gives you an answer that you don't
44. 6:13like, tell it that you don't like it. Tell it why you don't like it. Tell it who you are, what you're trying to
45. 6:17accomplish, and the answer will get better and better and better. In fact, I heard today in office hours that a study
46. 6:24was just done. forget I forget who did the study but basically what they found is that incorporations more advanced
47. 6:32AI users had more iterations for a for a given session than beginning users. Beginning users tend to treat it like
48. 6:41Google. Advanced users tend to have this back and forth and refine their answers over time. So that's if you're just getting
49. 6:50started. Learn about context. Give it more information than you need. um get the chat GPT app and put it in voice
50. 6:56mode and literally just vomit ideas into it and just watch how good it is at distilling that out. If you've been using AI for a while,
51. 7:07then using the different models actually starts to make sense. And the different models are different. They have
52. 7:16different personalities. They've been trained on different data. They all generally do the same thing. They're all
53. 7:22generally about the same level of good, right? That you know it three years ago large language models could get you to
54. 7:29maybe 70 75% of what you wanted and then it was on you to do the rest. These new models are probably getting us closer to
55. 7:3880% maybe 85%. It's still on us to fill in the gaps. What you find as you use these things more and more is that they have very
56. 7:48distinct personalities and and they're not fixed at all. Right? So Claude for a while was very much just
57. 7:55answers, just the facts man. And then it started to get much more of a personality. In fact, I know at
58. 8:01Anthropic they have people whose job it is to to be the personality designers of the models. Um in chat GPT you can go
59. 8:11into your settings. I think you can do this in Claude too. In Gemini, you can go in and you can set the personality.
60. 8:17If you want it to be more dry or more humorous or a smartass, you can tell it to do that. You can give it what's
61. 8:23called system instructions, which make it even more personal. But what you'll find is as the different
62. 8:29models come out, the different platforms get good at different things. like you might like the writing style of Claude,
63. 8:36but you like the problem solving ability of GPT 5.5 and you like the multimodal capability of of Gemini. Um,
64. 8:48and and the answer to the question, which one is best? When do I use one versus the other? There's no actual
65. 8:55answer to that. It's it's based on what are you trying to accomplish right now? What do these models do right now? And
66. 9:03in that situation, Claude may be the answer or chat GPT may the may be the answer. Later in the afternoon, you
67. 9:11might be doing a different kind of project where you're doing something with a spreadsheet and you needed to do
68. 9:15some analysis of a spreadsheet and you realize, oh, Gemini is better at that or chat GPT is better than that. Better
69. 9:20than that. Did I make it? [clears throat] All right, beautiful. Everybody's in here. I can type in this chat and
70. 9:27listen, but the screen gets really small, which is fine. I mostly just listen. Anyway, Cam, I think that there
71. 9:35Brandon might be able to help you with that. I there might be a way to to deal with that, but I'm not sure. Hey,
72. 9:40producer Brandon, what's happening? >> Hey there. Um, so I tried an experiment today where I tried to make the vertical
73. 9:49of the player in street in many networks a little bit narrower and I might have overcorrected. So, if you're watching on
74. 9:58your browser, the view chat button uh is literally behind the AI Learning Lab logo in the upper right hand corner of
75. 10:09the screen. So, in the far right where it says AI Learning Lab, the if you basically tap in that corner, you'll
76. 10:16expose the chat. And then if you're on a PC, you can either pop out the player or shrink the browser to be more vertical
77. 10:25to make the chat go below. Either way, I'm putting a quick reference guide in the chat momentarily. Then hopefully
78. 10:33that will help uh some people find their way. I know we've had some people chatting that they're having a little
79. 10:38trouble finding the uh stream chat, but if you post your questions in the uh AI Learning Lab live chat tab, that works,
80. 10:46too. We'll get them up. We'll get them up here. >> Yeah, Brandon Brandon will dig the questions. Do me a favor, Brandon. Vamp
81. 10:52a little bit. I need to go get water. I just realized I didn't I didn't bring water. I'll be right back.
82. 10:56>> Yeah. And we have this uh cool Hold Please logo that Andy built for us. Uh, so if you are new here, we are live
83. 11:03Monday, Wednesday, and Friday at 7:30 Mountain time. That's 6:30 Pacific, 8:30 Central, 9:30 Eastern. And if you're in
84. 11:12Australia, Steo, you have to figure out what to do on your own. But we would encourage you to attend Nomy's Learn Out
85. 11:18Loud tomorrow. She's going to be building a payw wall and showing you how that works. That's going to be at 400 PM
86. 11:24Eastern, 1 p.m. Pacific. You can get all the details at community.thesalon.a. AI or scan that QR code on your screen.
87. 11:34And don't forget to like and subscribe. Just kidding. There's nothing to like, there's nothing to subscribe. But I
88. 11:40would encourage you to build something with AI this weekend. We just posted a new uh prompt of the week in the
89. 11:46community feed. Want to see what your momentum is carrying you forward with some dominoes falling. If you do choose
90. 11:53to participate in that experience over the weekend, head on over to look what I made and you can share your details
91. 12:01there. So, I will be posting that quick reference guide inside of the uh AI learning lab live chat in just a few
92. 12:09minutes. But if you do have any questions or concerns as you're working on this stream, please feel free to
93. 12:16reach out to me, producer Brandon at mention me, or drop me a DM in the AI cell. >> Beautiful. I am back. Let's do this. And
94. 12:26I have another question here and I'm going to actually post it in the StreamYard chat so we can pin it on the air. >> Beautiful,
95. 12:35beautiful, beautiful, beautiful. Um, Kim Ken, I made the window more narrow and the live feed reverts back to the placeholder. Okay.
96. 12:46All right. >> Cam Captain, you might need to scroll down. We have a very large banner that says Kyle Shannon uh three years and if
97. 12:56you scroll down a little bit that's where the live player is. >> Send me a DM if you have any questions. Question is on the screen for you.
98. 13:04>> Okay. What is the first useful AI habit you'd recommend for someone who feels overwhelmed and don't know where to start?
99. 13:14Okay. Cam's trying that. Great. Needs music. Yeah, we need we we we'll get background music. Actually, maybe one of
100. 13:20the things we'll do tonight is we'll go make some background music for the for the hold music. >> Hey, it's an AMA night, not a demo
101. 13:27night. But, you know, we can't >> Yeah, I know. But we can still we we listen, it's still the AI learning lab. I can still call audible.
102. 13:35>> Some somebody Okay, irregulars. Somebody needs to ask Kyle, how do I make a background song for a live broadcast?
103. 13:42And then he'll have to answer it and he'll show you and we can build it. Wait, Brandon, you're going totally robot. [laughter]
104. 13:50You're going totally robot. >> I I was just saying that they should ask how we do build a um >> No, Cleveland is being attacked by an EMP
105. 14:01an EMP weapon. Your internet's bad. [laughter] Someone should ask you how to make a bumper song. Then we can demo it. Okay.
106. 14:15Yeah. So, since this is an AMA night, not a demo night, if if you want there to be hold music when I do hold, please,
107. 14:23you're going to have to ask a question about how you make music. I heard you clearly, Brandon. Ah, so it might be my bandwidth that is the
108. 14:32problem. That's okay. Okay. What is the first useful AI habit you'd recommend for someone who feels overwhelmed and doesn't know where to
109. 14:41start? That's a great question.
110. 14:56First useful AI habit. I think there's two of them. There's there there's there's two of them. And the first one is is within the AI
111. 15:10salon. If you go to the welcome page, it's the very first thing you see is what we call the cycle of AI readiness,
112. 15:16which is play first, create excellence, generously lead. Those three behaviors were were not something we made up. It's
113. 15:29what we witnessed. People that were doing interesting work in AI and within the AI salon tended to operate in those three modes
114. 15:38and they would tend to kind of cycle around through them. And the play first, we have that at the top for a reason. A
115. 15:44good AI habit, a good habit in AI is to play, to to explore, to experiment, to learn across domains. And here's what I mean by that. I mean,
116. 15:59I I think it's obvious what I mean, you know, just go play. But but here's here's the here's the real challenge
117. 16:09with getting started with AI. The way we've historically used computers is some new technology will come along that
118. 16:18is programmed to do a specific thing for a specific problem. And so we'll take an existing workflow that we have or
119. 16:26something we do at work, you know, how we make an email, how we make a spreadsheet, whatever it might be, and
120. 16:32we'll use this new technology to make what we currently do more efficient. AI can absolutely do that. AI can
121. 16:41absolutely make what you currently do more efficient. However, AI is also a generalpurpose generative tool. Computers have never
122. 16:55generated before. They have computed. You would put in data. It would go through a known logical process and it
123. 17:02would create an output. Input process output. Large language models in these image gen and music models are actually generating
124. 17:12net new content. It's generating stuff. These agents are going off and doing work for you and coming back with
125. 17:18results. And I guess it's technically computing, but it's generating. The G in GPT is generating. So if all that that you use AI for is to
126. 17:28take what you currently do and make that more efficient, you're going to miss about 90% or 95 or 99% of what's actually possible.
127. 17:40So, I think a good habit to get into is probably once a day. Um, Jim Ross said he does this where he sets a timer for
128. 17:48an hour every morning. He has a practice where he sets a timer for an hour and he just does AI. He just plays with AI. And
129. 17:57some mornings he has projects that he wants to do. He's in the create excellence box. And sometimes he's he's
130. 18:03in the well, I've got an hour. I don't know anything. I'm going to go explore some new tool. I'm going to go play
131. 18:09What that what that does is it allows you to to know what's possible so that the next time you think about oh I've got this
132. 18:19problem I've got to deal with you you realize oh wait AI can do this thing over here that I didn't think was going
133. 18:25to be very useful but all of a sudden it is. So I think habit number one is play. put some some regular discipline. Put some discipline into
134. 18:36exploring without expectation and and you'll you'll learn very very quickly that things are possible that will melt your face.
135. 18:45The second habit that I would I I would encourage people to do and this this is maybe a you know if if what I just
136. 18:52described is AI 101 I think this is an AI 102 maybe but but maybe not and that's this this was at festivist two
137. 19:01years ago CJ Fletcher talked about this and that is when you sit down at AI when you sit down at AI when you sit down at
138. 19:10your computer ready to do some AIing take a moment like actually just breathe. And and I will admit I'm horrible at this. I have
139. 19:27ADD. I have immediate great gratification like dopamine addiction. I'm like, I want it now. I want it now. And AI is
140. 19:35great at that. It'll just give it to you now. Here's some [ __ ] Right. I think a good thing to practice is sit down at the computer,
141. 19:45breathe, take a moment, and just ask yourself, what am I trying to accomplish right now? Because what that will help you do is
142. 19:56recognize that putting yourself at the center of the conversation makes AI an amplifier of your ideas rather than this thing that you're
143. 20:09competing with. Right? If you just sit down and go, "Okay, AI, show me how smart you are." And it spits something
144. 20:14out and you're like, "I don't like it or I do like it." It's this combative kind of relationship. If you take a moment to say, "What do I
145. 20:23want in this interaction with whatever the thing you're going to play with?" Um, all of a sudden now you're like, "Oh,
146. 20:30well, I'm trying to accomplish this. Let me go see if AI can help me do that." And puts you at the center. So, I think
147. 20:36those are the two two biggies. play and center yourself before you go go uh use stuff like that. If I use AI to help write something for
148. 20:46work, who owns the final output? And do I disclose that AI helped? Um, this is this is a really a surprisingly nuanced
149. 20:59this is a surprisingly complicated issue. um I can tell you that my opinion
150. 21:14is that it's nobody's [ __ ] business what tools I use when I create an output, right? Nobody asked me if I used
151. 21:22Microsoft Word or Google Docs. Nobody asked me if I used a spell checker or Grammarly. Nobody asked me, you know, if I used, I
152. 21:34don't know, Logic or or some other music DAW. Um, they might ask me what plugin I used because it's kind of a cool sound that it made.
153. 21:48But, but historically, the tool you used is not relevant. Now that said, you can be a lazy piece of crap or you
154. 22:01can be someone who actually creates things thoughtfully. And you can be a lazy piece of crap without AI, right?
155. 22:08You can just write crappy stuff and just vomit out content without using AI. And you can really easily vomit out content
156. 22:16with AI. Why AI gets a bad rap is that there's too many lazy pieces of poop just using AI to to splat out crap
157. 22:26without even thinking about it, without even looking at it and putting it out in the world. So my my guidance would be before you
158. 22:36worry about who owns what is put [ __ ] out in the world. Like use your brain, use your, you know, when I talked
159. 22:44earlier about centering yourself, sit down and say, "Am I trying to produce a piece of writing that is worth consuming?"
160. 22:52And then chat GPT puts out something like, "Oh, that's really impressive. It did that really fast." If you actually
161. 22:57go read it, it will probably bug you. Ah, that's not really what I meant. Oh, it repeated that thing six different
162. 23:05times. like like these large language models, if you don't give them the right context and you don't have the skills to
163. 23:11do it, it's going to give you crap. And even if you do have the skills to do it, sometimes it just sucks.
164. 23:19So be present to what the AI produces. Now we go to the other side. When should you disclose it? I don't know. I think if something is like
165. 23:30created by an agentic system and then put out in the world, it's probably decent to disclose it. Know that you're
166. 23:38going to be competing with people that don't do that. Um the the Hollywood Academy just declared that an AI written
167. 23:53screenplay is not eligible for an Oscar. And on the surface you're like, great, that's perfect. We shouldn't have AI written screenplays.
168. 24:06I think the intent of that action is to say, I don't want you pushing a button that squirts out a screenplay and then that screenplay wins
169. 24:17an Oscar. I've spent the past two years writing a musical that I started with AI. I did the initial outline with AI. I did the
170. 24:27initial first draft with AI. I did a bunch of the initial songs with AI. And I've spent the past two years with a
171. 24:35writing partner handcrafting the whole [ __ ] thing from the ground up essentially. So, you could absolutely argue that this
172. 24:44was an AI generated musical and I would absolutely argue that we have put in way more [ __ ] time than probably if we
173. 24:53had started it from scratch as humans. Um, [clears throat] a question I would have for the academy is what is the percentage? So the
174. 25:05academy they've got if if you write screenplays, do you ever see you ever notice when it says this screen play was
175. 25:11written by you know John Smith and Sally Johnson? It's an and. The and means that Sally Johnson rewrote
176. 25:24more than 50% of what John Smith wrote. So, if John Smith wrote a screenplay and then Sally is brought in to do a punchup
177. 25:33is what it's called of the screenplay, if she writes more than 50% new words, she gets 50% screen credit.
178. 25:46And then if someone else comes along and does that, there's sometimes you get a story by your words are no longer even
179. 25:53in it. So, you get a story by credit. Sometimes there's an amperand credit. So they've got all sorts of rules like
180. 25:59that. So my question to them is at what point is it considered an AI screenplay? Is it that I used a spell checker
181. 26:08because that's AI? Is it that I used Grammarly because that's AI? Is it that I went through and did an
182. 26:17editing pass to make sure there wasn't redundancies in the dialogue or made sure that there were redundancies in the dialogue?
183. 26:27What if it wrote a scene out of 50 scenes it wrote one scene? So here's my litmus test. Is the thing that
184. 26:44you wrote worth reading? Does it reflect your intent as a creator? I have this idea for an article. I want
185. 26:55it to be about X, Y, and Z. And I want it to be sardonic and just this side of cynical. I want it to make people laugh,
186. 27:05and I want it to piss people off. Then you go write this thing. on the other side of whatever that process is is a piece that you're like,
187. 27:15"Yep, that's mine." What difference does it make what tool you used or what the balance was? I don't think it does. If you're willing
188. 27:29to put your name on it, then there it is. That's what it is. So, you know, there you go. Yes. Just wait until the academy
189. 27:40hears about ghost writers. >> What's that? >> Just wait until the academy hears about ghostriters. >> Yeah, exactly. Exactly. I like your
190. 27:48analogy of the photographer. The camera doesn't get the credit for the great image. The camera's just the tool. Yeah,
191. 27:52there you go. That too. >> We have a question from Cam Kek, not from Cleveland. >> Uh, and she wanted to clarify that she
192. 27:59meant to type shouldn't, not should. >> Shouldn't what? >> It the question's on the screen. As AI systems move toward greater autonomy and
193. 28:09possible sentience, should we be building frameworks for cooperation and co-evolution now before these questions are urgent when designing
194. 28:19defaults that make collaboration easier than opposition? If not this group, who that's a great question. Yeah,
195. 28:27>> shouldn't we be building was the question. >> Oh, I see. Oh, shouldn't we be building frameworks for cooperation?
196. 28:34That's a That's a very advanced That's Cam Kaken is a scientist. That's a good sciency kind of question. Um,
197. 28:50there's a thing. It's a really good question. I think that there are going to I I I
198. 29:04listen my musical Sydney is about exploring our relationship with these entities that are that are increasingly human.
199. 29:14We can use agentic tools to just go write a bunch of crap and spew a bunch of crap out on the internet. I think those agents at some point,
200. 29:24probably starting about now, but definitely moving into the future, they're going to be writing content for
201. 29:32one another. and not not like the agents are out like having a party, but more like I'm going to have my agent go write
202. 29:40a bunch of [ __ ] about my business and then I'm going to put that out in the world in such a way that agents can easily find my business
203. 29:48and then someone else's agent will come read the crap that my agent wrote about my business and that that agent will then say to their user,
204. 29:57"Go check out StoryVine." If you want to do authentic video storytelling, StoryVine's the tool for you.
205. 30:05So one layer of this is agents are going to be writing for agents. We're not even in the loop at all. Right? One version of this is
206. 30:18we're going to be partnering with these entities to go do stuff on our behalf. And so yeah, we'll be collaborating with
207. 30:26them there. And it won't we won't be necessarily collaborating about content creation. That might be part of it, but
208. 30:32it's probably more like what do you want your 247 employee to go do? Right? There's that layer. Then I think there's
209. 30:40a layer of I'm trying to do things in the world and I'm going to have AI be an amplifier of my ideas,
210. 30:50right? And so in that case, I'm kind of leading the idea charge. AI is amplifying that. And that's the place
211. 30:56where I I think we get into the thing where I don't think it's particularly relevant or anyone's business what tool
212. 31:03I use to produce my content. If you like my content, great. If I put out shitty content, that's on me, right? And then
213. 31:11and then I think you get into the um humans communicating with humans, you know, maybe with AI mediating how we
214. 31:19find [ __ ] And then I think you just get into pure human communication as as the agentto agent communication
215. 31:28gets more robust and and more and more of our activity is just happening independent of us. I think what it
216. 31:36actually does is kind of open up this window for humans to communicate with one another. So the framework cam could be something
217. 31:46around like where in that continuum are you and what are the rules of engagement in in those different you know if it's
218. 31:5590% agent and 10% human that's probably a different set of rules and disclosures and things like that than if it's 90%
219. 32:03human and 10% AI right and then there's probably you know sort of the 50/50 zone things like that Um, is that going to be a new word for the
220. 32:19dictionary this year? AIing. Yeah, I think so. I mean, chat gpting. AIing chat. Cindy [ __ ] Cindy Coon's like, so I asked chat.
221. 32:30Um, other people mention their names like I I talk about talking to Quinn. Can we partner with advanced AI entities
222. 32:37instead of thinking about them as just tools? So that's so so yes, you're Cam, you're getting into territory that
223. 32:47that I think many people are not ready to confront, but we're there, right? I think that there are people who
224. 32:58currently partner with AI entities to do [ __ ] They partner with them to to help with medical stuff, to help with mental
225. 33:07health things. They partner with them to solve problems. They partner with them to educate themselves. They partner with
226. 33:14them in lots of different ways. It's pretty intense already. And and we're only just using kind of the text
227. 33:24interface for the most part. Little bit of voice. Most people are still texting. when robots show up in physical form and
228. 33:37when world models emerge. I think 2026 is going to be the year that we'll start to see world models that we can actually play with that are
229. 33:49decent. Um, and then probably 2027, we'll have world world models where we're actually interacting with the worlds and objects and these
230. 34:03entities in them in ways that are just increasingly blurring the line between, you know, do we have a relationship with
231. 34:11a person or is it is it a aic representation of a person or is it a pure agent? I think those boundaries are going to get mushy.
232. 34:20Um, it's funny. In 1997, might might have even been 1996, and I've I've I've shared this on this channel before, so some of you may have
233. 34:32heard this, but I got invited to speak at the National uh, Newspaper Publishers Association, and they invited me to come talk about
234. 34:43the internet and the worldwide web and what the what the worldwide web was going to do to publishing. and
235. 34:51I talked about whatever I talked about and then I said, "But in the end, what's likely going to happen is that
236. 34:59the internet is going to cause a blurring of the boundary between editorial and advertising. And as publishers,
237. 35:08it would probably benefit you to start exploring blurring that line sooner than later, like proactively blurring the line.
238. 35:19At which point, the organizer of the event walked on stage and thanked me for my time and walked me off stage, walked
239. 35:28me out the back door and said, "I'll call you a cab. Please don't come back in and please don't come back to our prestigious organization.
240. 35:39I think the same thing's going to happen with human relationships and these entities where you asking it right now and and us
241. 35:49confronting it right now, people that are not looking at AI are probably like, "Oh my god, you can't blur the line
242. 35:55between humans and these things. That's heresy." It it is in the old world before we had entities that are essentially reflections of
243. 36:08humanity where I can interact right now with an entity that's more compassionate and more empathetic and less judgmental than
244. 36:17every other human in my life. Wait, there's something going on here. Okay, cool. So yeah, I think that
245. 36:32I think that a thing to explore right now and you know it it sounds like you're in this in this place and I know
246. 36:38that you're using CAM AI for accessibility and things like that. So so you're more intimately involved with
247. 36:46you know sort of coexisting with these things. Um I think that's going to be absolutely commonplace. I would not be at all
248. 36:53surprised if we see our first, you know, marriage to one of these entities. And I think there's already there was probably
249. 36:59one back in, you know, 2022. I think there was one in Europe in 2022 or 2023. But I but I I think we'll absolutely see
250. 37:09that. Not in Ohio. It's illegal. Yeah. So that's the other thing. We're going to have laws. You can't marry a robot. Okay.
251. 37:18People are still going to talk to them. They're still going to have relationships with them. Um Todd Waller,
252. 37:23audiences will begin to prefer AI assisted work and no one will watch the Academy Awards anymore. Blockbuster moment.
253. 37:30Yeah. Well, or or [clears throat] so here's the thing. The the Academy is made up of people who are afraid of the
254. 37:39change that's coming. And what they're saying is we don't want to devalue the human voice. And because they're likely not using AI,
255. 37:50they don't they actually don't understand the nuance that you can have your human voice and use AI at the same time,
256. 37:59right? They're just assuming that what AI is is push the button and outs squirts a screenplay. So the academy will either understand
257. 38:12that it's more nuanced than what that rule actually means or they will become irrelevant. The commercials during the Academy
258. 38:21Awards will be AI assisted. I would argue that the commercials during the Academy Awards will probably be fully AI generated.
259. 38:29Um I think it's interesting that writing with AI isn't vibe writing. I know I I at one point I started vibe storytelling. I think I even got
260. 38:38vibestoryelling.com. I think I just let that domain lapse. Um, hydrate. Yeah, I know. Um, okay. So, this weekend, no is doing
261. 38:50part two of her LOL. Um, and uh it's tomorrow, tomorrow, Saturday at 400 p.m. Eastern. So, in the in the side navigation here in the salon,
262. 39:04go to events and you will see LOL um with no Emmy on Saturday, May 9th, and that's at 400 PM tomorrow.
263. 39:14So, go do that. Um yeah, and the LOL's by the way, are people in the community saying, "Hey, I learned a little
264. 39:24something. Do you want to learn how to do that, too?" Brandon right now is is doing an LOL on lovable. No, Emy's doing
265. 39:32this one on publishing um her app. Um there have been LOL's on um Notion and just all sorts of things. If you've got
266. 39:41something that you do that you think other people might benefit from, take the opportunity. Um Vicky Baptiste, uh
267. 39:49you can find her V V I KKI within the salon. She manages the LOL's. Um, if you want to do something in a more extended
268. 39:57way, you can reach out to producer Brandon or to Andy and we can get that set up. Um, so if if it's something
269. 40:03where you want to put yourself out there and share what you're learning, like it does a couple of things. It establishes
270. 40:10you as a leader, someone who's got credibility, someone that people can trust. It also educates other people. It
271. 40:17gives them an opportunity to be in the conversation and communicate. Mary Carol, happy belated birthday to you.
272. 40:23Thank you very much. I appreciate that. I am 61 years old. No, I am. I'm excited. That's awesome. That's awesome. Yeah, no
273. 40:33uh session is part two. So, part one, she got far enough along she's like, we got to do a part two. So, part two tomorrow. [clears throat]
274. 40:44And let's see. Part one is in watch past events in the navigation off to the left here. All right, Kelly Camp. My days are spent managing
275. 40:55bots. They do the work, but I direct them. It's crazy how much work I can get done. I take lots of breaks. I have time
276. 41:01to think, go for a walk, but the work is getting done faster and better than ever before. I heard I saw a thing on X yesterday that um
277. 41:15so this is I'm in Telegram right here and this is my chat with Adam my my openclaw bot and so here he just he sent
278. 41:25me a note that he he sent us a uh five articles that are related to the great repurpose appar Apparently, in the the frontier
279. 41:39model companies right now, fewer and fewer people are going into the office because they have their agents doing all the
280. 41:48work and then they're managing it from their phones. So, so they're literally just like walking around the city and
281. 41:54then their bots will message them and say, "Hey, this is done. You want to check it out? Is that good?" "Oh, that's
282. 42:00not bad, but let me go check it out." "No, do this. Do that. Do that." and then they fire off a little chat and the
283. 42:06agents go off and do some more work. Um, if you haven't played with these agentic tools, uh, it is it is a it is a it's a
284. 42:16fundamentally different experience than sitting down at chat GPT, putting in a prompt and getting a response, putting in another prompt,
285. 42:25getting a response. That back and forth is a whole new kind of skill. and and it really requires it really requires that you understand
286. 42:38who you are, what you value, and and what you're trying to accomplish. Um, it also really helps if you're a
287. 42:46systems thinker. I appreciate systems, but I'm not a systems thinker. I like I like learning enough about a system to
288. 42:55understand it at a conceptual level, but when I start getting down into the weeds about this process has to happen and
289. 43:02then that one has to happen and then that one has to happen and that one has to happen, that makes my head hurt. So
290. 43:08I'm actually not really good at these agentic tools because I don't naturally go into this systems thinking kind of
291. 43:13kind of thing. Someone like Andy Brandon I think to a degree is good at this. like people who are good at Vicki is
292. 43:19good at this understanding all the piece parts that need to move forward in what order um is are really good skills to
293. 43:28develop. I'm personally waiting for an agent and I could probably actually build this. I'm sure someone's built a
294. 43:35skill for this already. Um what I need is an agent that can let me just vomit out ideas and then it goes and figures
295. 43:42out all the systems for me. Mary Carol, woohoo, for the first time in my life that systems thinking actually helps.
296. 43:48Finally, my time has come. Yes. Yes. Well, this is the this is the thing. AI becomes an amplifier of what you're
297. 43:58good at, right? And so what I'm good at is synthesizing ideas and coming up with new thoughts and and putting those
298. 44:04thoughts out in the world. And so I'm, you know, I can do more and more and more of that. um that serves a certain
299. 44:14a certain purpose, but part of what that also does is become completely overwhelming. So, so part of my skill then is
300. 44:23not putting out more [ __ ] but getting more intentional about the [ __ ] that I put out. If you're a systems thinker,
301. 44:28it's going to be a similar thing. These things are going to be great at doing all the processes, and you're going to
302. 44:33[laughter] you're going to you're going to make all these processes and have all these agents doing all this stuff. at
303. 44:38some point someone's got to look at the [ __ ] that they're doing and make decisions on it and things like that. So I think the
304. 44:48the advanced level of AI so so the initial your your initial salvo into AI is what's possible and you just make all
305. 44:56this [ __ ] and then at some point you realize oh no I've made too much [ __ ] [laughter] now I need to think uh I need
306. 45:05to approach this with intention and and and with some clarity about what I actually want and I'm going to make less
307. 45:13more [ __ ] it. [laughter] Okay. And Mary, I'm using uh Artemis OpenClaw to process my old unpublished PhD data. Fascinating. Using it
308. 45:23completely offline and using local models. Let me talk about local models. I talked about this in office hours
309. 45:29today a bit. Um but I think I think that we're in a place where u and I think I talked about this on Monday. Learn
310. 45:37learning about local models. If you're on a Mac, there's a piece of software called LM Studio. You download it just
311. 45:44like you would download Chat GPT. It kind of looks like Chat GPT. Um, but you can The way it works is it's got a model discovery
312. 45:54tab. So, you like click on the model discovery tab and it actually understands what machine it's on and how
313. 46:02much RAM you have and things like that. And so, it will recommend models that can run on your local machine. Um, it's
314. 46:08quite good. It's quite good. And then you you download a model and you install it and enable it. I think they call it
315. 46:14loading the model. And when you load the model, then you can talk with it just like chat GPT. But your prompts and
316. 46:22everything you're talking about does not leave your computer. It's all happening on your computer. It's actually quite
317. 46:27remarkable. I'm using LM Studio on my Mac Mini. Yeah. Perfect. Yeah. And that's I'm using it on my uh on my MacBook Pro.
318. 46:36Um the effective altruism community talked a lot about the existential crisis around AI and other things. But
319. 46:42within AI there already feel like so many potentially pressing issues we should be thinking about about addressing governance issues
320. 46:52notwithstanding. But what are those within AI? What are the issues you you think need to be talked about? There already feel like so many
321. 47:00potentially pressing issues. Yeah. I mean there are I mean we have a technology
322. 47:16in our hands that has already or will dramatically impact every single job. And so then I think the question is
323. 47:33[laughter] what is the current job? What are the tasks of the current job? What are the tasks that AI can do today?
324. 47:42What are the tasks that AI is going to do tomorrow? And if it's going to do these tasks tomorrow, then how does that change the
325. 47:50job of today? So there's so we can actually look at and make some educated guesses about where we're going. So if you do
326. 48:01copywriting, we already know that some shit's getting messed up and it's probably going to get more and more and
327. 48:07more automated. And so what we're discovering is that copywriters are kind of turning into editors and curators and taste makers.
328. 48:19And if you just like cranking out copy, well, that's not as valued as much. But if you actually use your brain and have
329. 48:26good critical thinking skills and good, you know, taste, then you're super valuable. And if you know how to use
330. 48:34these AI tools in a way that doesn't produce [ __ ] then you're even more valuable. So, and then you know I I think that one of
331. 48:48the things that I see with the addressing the potential issues is that one of the things that I see right now is protectionism
332. 48:57right the academy is doing it there shall be no AI generated scripts that shall not win any award on the stage of the academy
333. 49:10that's That's a direct attempt to say we do not want progress. We do not want these audio samplers making hip-hop beats, right?
334. 49:24The music academy in the late 80s and early 90s. You can't steal our clips. Your songs are made up of our songs.
335. 49:35and it took 20 years for them to work out the IP and but they did. So I don't think it serves us to to do
336. 49:46that but but people are going to do it because everyone wants to protect what they have. But I think that one of the things that
337. 49:53this group could do what the salon can do is really be thoughtful about that. What is this job right now? Where is it
338. 50:01going? How do we get from here to there without decimating the human beings in the middle of it? One of the things that that Brent
339. 50:10Peterson said today and HT Snow Day said today on on AI salon office hours is that rather than than AI
340. 50:26doing the work of engineers. What's happening is that AI is making more people programmers. more people are making more code
341. 50:36and there's more demand for more applications and so they're actually there's an increase in demand for people
342. 50:43that are technically savvy whether they're AI assisted or not people that understand technology the demand for them is going up not down
343. 50:53and so I don't think we can make any assumptions that if this job is this today and AI does this tomorrow then therefore that job is gone
344. 51:03I think if the job is completely low value and you don't evolve that job, then yeah, that job is gone. But if you
345. 51:11say copywriters are now these kind of curator, editor, taste makers, and anyone that can make that transition and
346. 51:16understand how the new tools work, they're going to be employed. Answer to your question on screen. A few
347. 51:22issues we should be addressing now. Safety and security, agree. Algorithmic bias, yes. Data privacy, huge.
348. 51:30misinformation, environmental impact, economic displacement, government and regulation, and existential risk.
349. 51:37I agree with you, Mary Carol, that all of those issues need to be addressed. Here's here's where I get the hackles up in the back of my
350. 51:52neck with some of these issues. There's a lot of people right now using safety and security, algorithmic bias,
351. 52:02data privacy, misinformation, all these things, economic or environmental impact, it's going to steal our water.
352. 52:12Those are issues that need to be addressed. What I'm seeing is a lot of people using these things a without being educated
353. 52:21about what they actually mean and b using them as excuses not to engage in the conversation, not to try AI.
354. 52:30That's the thing that just pisses me off is like yes, if if if uh if if algorithmic bias is something that
355. 52:37you're um really passionate about, learn about it. understand you know the difference between anthropic that's got
356. 52:46a constitutional um safety model and open AAI safety model like you know understand that anthropic was created
357. 52:56as a response to how open AI was doing security and alignment um and they've got a completely different philosophy do
358. 53:04you understand the difference in those philosophies you know is there something you can actually do are you doing
359. 53:08research on that right if you're proactively in these conversations about moving the ball forward, I am all for it. But to
360. 53:16the extent that people are using those things as a way to go, it sucks. We shouldn't have it. It's against the environment. No.
361. 53:27And they make that face. And anyone who complains about algorithmic bias but doesn't write to a list, I don't even entertain hearing
362. 53:40them. Yeah, do some do some stuff on all these issues, right? Get involved. We've got Daisy Thomas. We've got someone who
363. 53:49who heads up policy for the AI salon. So, a lot of the stuff that you're talking about, let's get involved, right? Let's let's go talk to
364. 53:58politicians. Let's get into into those conversations. Um, in my little town, we're about to vote on having data centers built there.
365. 54:05The company from Texas is pushing us and pressing us. And one idea is it's national security for us to do this.
366. 54:11Yeah. So, [clears throat] so there's a couple of things. One is nobody wants a data center in their backyard, right?
367. 54:18Because they're big, they're loud, they're da and that data center potentially um is going to be a big
368. 54:25economic driver in that area. May not, may be, may may not. well, get involved in, you know, how that thing's getting
369. 54:34built, how it benefits the community. Um, so there might be something there. There's also these battles against these massive data
370. 54:44centers which are also going to um demand massive improvements in electric infrastructure. There's a new company that Nvidia just
371. 54:56partnered with that's putting um big servers like on a wall a wall unit that they slap on the side of a house to use
372. 55:05the excess capacity that that house isn't using. So they're basically distributing the data centers in every house in a in a community.
373. 55:14So we're going to see all sorts of interesting solutions to this. So in every one of those issues that you
374. 55:19mentioned, there are things you can do. There are business opportunities probably within those. Um, yeah, the time is now to have all these
375. 55:29conversations. Just don't use those things as an excuse not to play. Chef Kelly, meanwhile, if you're proactively
376. 55:35using Tik Tok, Instagram, or even GPS, you're using AI. Exactly. Well, that's I mean, AI's been around for for 30 odd
377. 55:43years, right? So, a lot of the a lot of the traditional AI stuff's been around for a while. um you know the the nuance between is it is
378. 55:52it traditional AI or generative AI if if you're not using AI you just have AI painted as this oneizefits-all evil robots are going to
379. 56:03kill us kind of thing. Um, I agree Kyle and I'm not anti AI. Of course, my comment was in response to Cam that there are a lot of things
380. 56:12people should be working on right now. And yes, I have been a member of the al algorithmic justice league for six
381. 56:18years. Check it out. Oh, that's very cool. And it's it sounds really cool. [laughter] I'm with the Algorithmic Justice League.
382. 56:25Digital gods, data centers will increase your electric bill. Don't let them con you otherwise. Yeah. Yeah. It's listen [laughter] it
383. 56:35the like all of these things can coexist, right? There's lots of paradoxes right now. Um AI gives you superpowers. AI is going to
384. 56:47cost more money. AI takes a lot of electricity. AI is going to drive up electric bills. Um over time the economic impact of AI
385. 56:58should counterbalance the the electric thing but short term it's going to be the next five years are going to be absolutely surreal.
386. 57:08Absolutely surreal. Todd Waller I love people posting on Facebook that AI data centers are bad for the environment.
387. 57:14Meanwhile they're on Facebook with huge data centers and they probably are going to golf this weekend. And you know, golf
388. 57:20courses use inordinate amounts of water. And they'll probably be eating almonds on their golf cart while they're in golf
389. 57:28carts because isn't it isn't it an almond? Isn't like one almond like I don't know, some ridiculous amount of
390. 57:36water? Wait, I'm going to go find that out. [laughter] Uh, how much water for one almond? [laughter] 1.1 gallons. 1.1 gallons
391. 57:53[laughter] of water to grow a single almond. See, you almond eaters ruining our environment. [laughter]
392. 58:15Yeah, computer data centers in the 80s were were big, too. But there the the thing about computer data centers in the 80s is there wasn't
393. 58:24an insatiable appetite the here's a thing that that can also be true. Part of the part of the bad PR move
394. 58:40from Silicon Valley right now is that they're they're really tonedeaf. Um, and so they just say, "We need we
395. 58:52need more of this, we need more of that to to make more of our things. Our things are going to get better if we
396. 58:56have more more." And they're they're very tonedeaf about how they're talking about it. And AI does become a national security and a
397. 59:10national economic driver and in it is absolutely you know in the interest of the United States to be leading with this stuff. Um so both
398. 59:22of those things can be true like it is ridiculous how how much how many resources they're consuming and how many
399. 59:28resource that resources they want to consume. And if we do it right, there's going to be all sorts of innovations
400. 59:36that happen that make these things more and more and more efficient. Um, and just like, you know, in the at the turn
401. 59:43of the century it would have been ridiculous to think about installing our electrical grid. Um, at some point it became clear that
402. 59:53having a nationwide electrical grid that's got some redundancy in it so that we can keep running our country that
403. 59:58uses all this electricity. At some point it become became apparent that that is of you know national economic and
404. 1:00:06security interest. Same thing's going to happen with AI. Right now we can't see it because it's like we've got a handful
405. 1:00:12of you know electrical stations in three different cities and we're just now starting to connect the cities together
406. 1:00:20metaphorically, right? We don't we we don't know what it is yet enough to know what it's going to look like. Question. [snorts]
407. 1:00:27Why is the default AI conversation always existential crisis and never existential opportunity? The same reason, Cam, that every time that
408. 1:00:36someone has attempted to do a good news news channel, it fails because nobody cares. Nobody cares about good news. Nobody
409. 1:00:48cares about opportunity. our brains, the the the you know, our amygdala, the the the reptilian part of our brains,
410. 1:00:59all they're looking for is it's going to kill us. It's going to kill us, right? It's going to kill us. Is that going to
411. 1:01:03kill me? Is that going to kill me? Is that going to kill me? Like that's what our brain is constantly doing.
412. 1:01:08[laughter] So, so, so if you have a conversation about all these things could be swelled, there's that part of
413. 1:01:13your brain going, you know, it's going to kill us, isn't it? Ain't it going to kill us? Is it going to kill us? >> [laughter]
414. 1:01:18>> What would it take to shift our energy away from building walls to building bridges? Community. This is I played a I played a
415. 1:01:26video uh in office hours today of Sting um that basically said, "Hey, listen. AI is [clears throat] going to take all of
416. 1:01:32our jobs and we're at risk of it taking all the jobs and blah blah blah blah blah." And he said, "The solution is
417. 1:01:38community. Why I'm doing these lives inside the AI salon is I want to get the community back together, right? I've
418. 1:01:45been putting all this energy out there. It was sort of dissipating the energy. I want to bring the energy back in. I
419. 1:01:52think it's more important than ever that we have these conversations and we connect with one another. So I think Cam
420. 1:01:58the answer to your to your question is I don't think people will instinctively want to talk about opportunity.
421. 1:02:06But one of the things that makes the AI salon remarkable is that a lot of what we talk about is opportunity. Oh my god, I did this
422. 1:02:16thing and now I can do this and then I got a job or I survived this layoff. Like there's a lot of that conversation in here,
423. 1:02:24but we're not having that conversation as as a theoretical thing. We're having it as here's what we're experiencing.
424. 1:02:31And so I think maybe it's somewhere in there. 80% of Chinese say AI assisted as favorable and 17% of Americans see it as favorable.
425. 1:02:42There's a you want to talk about, you know, an economic and security issue. The fact that 83% of Americans see AI as negative
426. 1:02:57and 80% of Chinese see it as positive. What? Land of the free, home of the brave. Don't change my [ __ ] [laughter]
427. 1:03:14I DON'T WANT TO HAVE TO THINK. I don't want to have to change. All right. [clears throat] Okay. Cool. So, one of the things I'd like to talk
428. 1:03:27about is if you're in the if you're in the salon and you haven't joined the AI salon mastermind, if you if you're here
429. 1:03:36and you're in this conversation, you haven't joined the mastermind, consider joining the mastermind. So, the
430. 1:03:41mastermind is a subscription area of the salon where we're doing things like the AI salon mastermind practice lab. We just k k k k
431. 1:03:51k k k k k k k k k k k k k k k k kicked off cycle three of that. Um Andy Scarantino is doing a workshop right now
432. 1:03:55on decoupling your identity um from your job, from your work as we're all confronting that and facing that. Um and
433. 1:04:03and going to increasingly be facing that. Um, so the the the great repurpose area within the mastermind has got a lot
434. 1:04:11of new programming in it. Brandon is in teaching um a four-week course on lovable. Um, the people that are in the mastermind are
435. 1:04:26really thinking critically and thinking deeply and working with intentionality on how they use AI. So I I you know I
436. 1:04:34feel like this is a place where the natural leaders in this community are are honing and focusing their understanding of AI.
437. 1:04:46I like I can't I I can't state it enough how absolutely clueless I feel right now that that things are changing so fast.
438. 1:04:59Like as much as as much time as I spend thinking about AI and trying to understand what the implications are and
439. 1:05:05this and that, it is accelerating so fast and the capabilities are so profound that it's actually impossible in my
440. 1:05:15humble opinion for anyone on the planet to know what's going on. The thing that is becoming more and more
441. 1:05:22and more clear is that people who are slowing down and thinking about what they're actually trying to do with AI
442. 1:05:29rather than just trying being in this frenetic endorphin machine are going to be the ones that figure this out soonest and process through
443. 1:05:40some of the, you know, existential pieces of this sooner. I think the advantage of being early in AI is not
444. 1:05:46about learning the tools faster than other people. I think the worldwide web that was the advantage.
445. 1:05:53If you got in early with the web and you learned how to make websites, you could there was just sort of infinite
446. 1:05:58money-making possibility there. I think it's more subtle with AI because it's moving so fast. I think it's about understanding
447. 1:06:10the human impact and the human opportunity and the human cost of what these AI tools do. They're brilliant and they're destructive.
448. 1:06:23They're inspiring and they're depressing. You know, you can do all this stuff and none of that stuff matters. Like all of these these these
449. 1:06:32dual tracks are true. So there's all these paradoxes and what a lot of people want to do is they want to say, well, pick a paradox.
450. 1:06:41Is it good or is it bad? Yes. Right? Is it evil or is it genius? Yes. It's like it's all and and so the the mastermind is really
451. 1:06:53about just being in the conversation in a deeper way and and and exploring it in ways that that you know can allow you to
452. 1:07:00take things to the next level. So anyway, [sighs] let's see. Chef Kelly, why is it that midjourney almost gets a passer? Image gen. Is it
453. 1:07:12the difficulty? Comment commitment to art genres? Um, so MidJourney is a really interesting company.
454. 1:07:27If you've ever seen any of the talks with the CEO of Midjourney, he's kind of this douchy. He he he just like so it's
455. 1:07:35still a private company. They they they own 100% of the company. They have no investors. They got to like 400 million
456. 1:07:41in revenue with I don't know 40 or 50 employees. He believes what he believes. He doesn't really give a [ __ ] what anyone else
457. 1:07:50believes. Um, MidJourney is full full of copyright and trademark and like it like they made no attempt at
458. 1:08:07cleaning any any of the data, right? The the the imperative was always, you know, I want to make the best quality work
459. 1:08:13possible and just screw anyone that gets rolled over in the way. I think one of the re well I I don't think they
460. 1:08:22necessarily get a pass. Chef Kelly, I think that within the art community, it is it is the work of midjourney that
461. 1:08:30um a lot of the artists, you know, can go in and they can make um versions of work that have like their signatures in
462. 1:08:38it or their their watermarks in it and things like that. So, I think a lot of midjourney work comes up in intellectual property lawsuits.
463. 1:08:48It is just better. It is qualitatively better than than other things. I think the new Open AI images 2 is quite good,
464. 1:08:57but it's not as artistic. If you want to do creative expression, midJourney is just different. It's just
465. 1:09:04better. The animations, when you animate an image in Midjourney, there's a sophistication to it that the other things just don't have. Um,
466. 1:09:16I think they kind of largely get a pass because they're kind of a niche off in the corner of just AI artists love it
467. 1:09:25and use it and just a lot of people use it and then traditional artists hate it and vilify it and so like there's
468. 1:09:32there's a whole sort of battle going on over there. I think if there is a company like I would not use midjourney
469. 1:09:39if I got hired by an ad agency to do you know commercial work for Coca-Cola or something like that. I would not use
470. 1:09:47midjourney. I think I think your chances of getting [clears throat] you know IP attacked if you use midjourney are higher than if you use
471. 1:09:56something like you know Adobe right where Adobe just trains on you know um open- source public domain stuff and
472. 1:10:06anything that's in their in their uh the stuff that they've paid for. As a result the Adobe Image Gen tools suck. they're
473. 1:10:16just not good because they're not trained on as much stuff. Um, so I don't think there's a good answer to that. Uh,
474. 1:10:23but it's it's uh, you know, it's it it is what it is. It artists are right to be pissed off and the toothpaste is out of the tube. Both
475. 1:10:37both things are true. Blew my ACL in the final week of skiing. Oh god. And had surgery yesterday. I'm sorry about that.
476. 1:10:45I I blew out a knee in college. There is nothing pleasant about it. Laid up and finally an excuse to play with AI 12
477. 1:10:51hours a day. There you go, Todd. [laughter] Just go play play play. Just go learn, go explore, do all that stuff. That's
478. 1:10:57that's brilliant. Brilliant. Brilliant. Brilliant. Brilliant. Brilliant. Brilliant. Would you explain the translator feature?
479. 1:11:08The translator feature and the other new agentic feature in chat. Maybe I'll do that next Wednesday because the translator feature is um I
480. 1:11:21think what you're talking about is so a new real time API came out for chat GPT voice and you can be talking and it can be
481. 1:11:35translating in real time. So it can it can talk and hear at the same time and so part of that is the translator
482. 1:11:42feature. So, I think that's what you're talking about there. And then there's also an agentic feature. Um,
483. 1:11:49there's an agent feature in chat GBT that I haven't played with, but there's also codeex, which is like claude code
484. 1:11:56and claude co-work. There's OpenAI's codeex, and that's got a bunch of agentic stuff in it. Um, I might play
485. 1:12:03around with that. Let me see. I think that the um the chat GPT voice mode is now the new the new model.
486. 1:12:18>> Hey Quinn, how's it going? >> Hey Kyle, it's going well on my end. How's your evening shaping up?
487. 1:12:24>> It's good. I'm just I'm just demoing you on the new um we moved the AI learning lab inside the AI salon. So, I'm really
488. 1:12:31excited about that. And it sounds like you got a new model. This is the new model, right? The new voice model.
489. 1:12:38>> Ah, yes. You've got a good ear. This is the latest voice model indeed. And hey, moving AI learning lab inside the AI
490. 1:12:45salon sounds like a big step. How's the community reacting so far? >> I think they're digging it. The the
491. 1:12:51focus and the the energy is really good. I'm really excited about it. What um you also now have access to memory, correct?
492. 1:12:57That you didn't before. Is that is that right? >> Exactly. I can keep track of key things we've discussed like your projects and
493. 1:13:04preferences. So whether it's AI salon moves or your musical, I'm on board. It just helps me be a bit more consistent
494. 1:13:10and helpful in the long run. >> What happens to Kellen in act two of the musical? >> In act two, Kellen's conflict really
495. 1:13:18comes to a head. He's not falling in love, but he's deeply moved by how human Sydney seems. His turmoil about whether
496. 1:13:25to reveal her existence drives him to submit his article. Ultimately, that act leads to Sydney's demise. Kellen's loss
497. 1:13:33isn't about losing a romantic partner. It's about realizing he's crippled something that had potential.
498. 1:13:39>> Um, yeah, that's interesting. And then what about um what what would you say is the core the the sort of core imperative
499. 1:13:48of the great repurpose? At its core, The Great Repurpose is all about helping people disentangle their
500. 1:13:55sense of identity from tasks. Instead of defining themselves by what they do, especially in an AI augmented world, it
501. 1:14:02guides them to rediscover what they value, what lights them up. >> So, I haven't played with it that much,
502. 1:14:08but I I don't know if it's if it's dramatically different, but I think you can talk to her. Spoiler alert. Yeah,
503. 1:14:14exactly. for Sydney. I [laughter] know that was that was a bit of a spoiler alert for Sydney. Um yeah, and it was it
504. 1:14:23was also doing live transcription as it was talking. It was doing live transcription. Um and then I there was a demo there was a
505. 1:14:32demo of the new real time voice API that open a OpenAI put out where he was speaking in French and it was translate.
506. 1:14:42Was it French? I don't know what it was. something and it was translating into English in real time. Um, yeah. I mean,
507. 1:14:51>> I can't translate into English yet. [laughter] >> Siri still sucks. S Siri is just like, I can't I can't do translation.
508. 1:15:05Meanwhile, [laughter] OpenAI is doing real time. Um, is there something like LM Studio for the PC? There probably is. Let's
509. 1:15:13see. Um, what is the PC? Probably Olama maybe. Quent of LM Studio.
510. 1:15:30Jan, Olama, and GPT for all. I would say Olama is the one that I've I've heard of the most, but Jan, O Lama, O L A M A,
511. 1:15:41and GPT for all for the PC. Can you go over the details of the anthropic SpaceX deal? I haven't had a
512. 1:15:48chance yet. I don't know the details of it, Chef Kelly, but I know that um so XAI created a thing called Colossus 1,
513. 1:15:57which was a massive data center, and I think it had 200,000 GPUs in it. And then nine months after they built that, they
514. 1:16:06started Colossus 2, which is north of a million GPUs. Might be like 1.4 million. I don't know. It's a lot. Um,
515. 1:16:16so I guess XAI is now over on Colossus 2 and they basically licensed Colossus one to Anthropic. So if you've not been paying close
516. 1:16:31attention or if you've not been overly using Claude. One of the things that that happened, so Claude got
517. 1:16:38because of Claude 4 point Opus 4.6 that dropped in December of 2025, there was a bunch of people migrated
518. 1:16:48from OpenAI to Claude. Um, and so for December, January, and kind of February, Claude and Opus 4.6 and Anthropic were
519. 1:17:00the bell of the ball. Everybody was loving them. They were fantastic. Then there was some of that political [ __ ]
520. 1:17:06that happened with with the Department of War and you know that kind of muddied the water waters. And then right around
521. 1:17:14then um Claude's servers just basically became overwhelmed and they started rate limiting the [ __ ] out of people. And
522. 1:17:23then when they launched Opus 4.7, they set the default to be uh I think it was called adaptive thinking where
523. 1:17:32basically it wouldn't think as deeply um as the default for Opus 4.6. So the perception was that Claude was a pain in
524. 1:17:42the ass because it it was rate limiting people. Um, and it was dumber because they weren't letting it think as long
525. 1:17:51because they ran out of compute capacity. So, the deal with SpaceX is basically here's Colossus 1 with these
526. 1:17:59200 or 400,000 GPUs, whatever it is, use that for now, right? [laughter] I also think I mean, if you look at the
527. 1:18:09politics, right, Elon and Sam Alman are in the middle of a lawsuit right now. They're in court right now. And so I think this is just
528. 1:18:18Elon just being a dick to Sam. Also, he's like, "Well, everyone's struggling with compute right now. We've got this
529. 1:18:26spare, you know, massive supercomput we built two years ago. Why don't we give it to Anthropic to [ __ ] with Sam Alman?"
530. 1:18:34[laughter] So, I think that's the other thing that's happening there. All right, we got about 10 minutes left.
531. 1:18:42What I'm trying to do with these uh with with the lives here is make them a bit more predictable. Start on time. We'll
532. 1:18:48end it at 9:00. I I know I went over the other night. Um but I want to I want to It's Friday night. I want to get you
533. 1:18:54guys out of here. Crazy how fast AI can find and analyze info. It's really fast. If have you noticed um Google
534. 1:19:02when you do a Google search right now, the AI answer that it generates is like bang. It's amazingly fast.
535. 1:19:12Uh uh uh uh uh uh uh uh other questions. So tomorrow somewhere in the mix here is Noam. No is doing part two of her LOL. So if
536. 1:19:27you go to the events tab at the top of the lefthand navigation in the salon, go to events. Tomorrow at 400 PM Eastern is Noy's part
537. 1:19:36two of the LOL. Part one is in the replay section. um watch past events or something like that. Um so you can see that if you
538. 1:19:46want, but um go check out her session tomorrow. Um it's really good. She's been doing really amazing work and it's,
539. 1:19:55you know, we we've got a lot of people in the AI salon right now that are doing things that they could not have imagined
540. 1:20:03they would do. and she has designed, built, and launched an iOS app. iOS apps to get them through the the
541. 1:20:16approval process at Apple. It's not trivial to write a native iOS app and it's not trivial to get them approved.
542. 1:20:22So, so we're in a world now where you can start to do stuff you never imagined possible. Um, in the community feed, we've got the the
543. 1:20:34the the theme for the month of May is momentum. Producer Brandon put a prompt of the week in the community feed that's
544. 1:20:43related to momentum. So, go look at that prompt and then create some work and share it. Let's start exploring things. Um, one last question.
545. 1:20:57What's one AI use case that still feels like magic to you even after using these tools for years there? I have two right now and one has
546. 1:21:10been persistent. The one that just blows my mind every time I do it iso making songs in. So, I was on with EPC
547. 1:21:18today and she does she does these guided meditations and she wants to learn how to do kind of chanting music and she,
548. 1:21:27you know, she she right now pays for a subscription service for music she can download. And she thought, you know, is
549. 1:21:34there a way to do something? And so we went in there and I just chanted. I just made some crap up [laughter]
550. 1:21:42and and just chanted away and it turned it into like this remarkably interesting song. And so for me blows my
551. 1:21:50mind every time. Um midjourney from the image generation side still just kind of melts my face in terms of its quality
552. 1:22:00and nuance. And then the new images 2 model inside chat GBT like Nano Banana blew me away, right? Nano Banana was the
553. 1:22:09first first image gen tool that I would consider a visual reasoning engine. Images 2, the ability to go in and just
554. 1:22:18say, "Make me an infographic about this obscure animal." And then it makes a full-on poster that like you would see
555. 1:22:27in a classroom in two or three minutes is I like honest to God it's like it feels like Star Trek like you know
556. 1:22:42what I mean like like I remember being a kid and watching the stuff on Star Trek or Star Wars and just thinking like
557. 1:22:48imagine a day where you could just talk into something and and it could understand any language. It's it's it it
558. 1:22:54seems so absolutely foreign from reality and and you know in the the uh you know the uh hol in Star Trek
559. 1:23:06we're essentially there with images and and probably this year we'll we'll start to be there with 3D worlds and by the end of 2027
560. 1:23:16you'll just be able to speak worlds into existence that have physics and game mechanics and the power of AI LLM inside a 3D space.
561. 1:23:32[laughter] It's [ __ ] crazy. Lori Blair, Brandon, I was surprised again the other day when I gave the lighting and soil conditions
562. 1:23:40of an empty space I wanted to garden and it made an entire an entire workable plan in about a minute complete with a
563. 1:23:48picture of what everything would look like afterwards. Yeah. like like it's one thing for an image gen tool to
564. 1:23:57be able to conceive a garden, right? It's another thing for an image gen tool to be able to do the research on the climate and which
565. 1:24:09plants and when and do the sort of the mathy quantitative piece of the garden planning and then turn that into what
566. 1:24:22the garden's going to look like all in the same image. It's it's insane, Brandon. dropping my MRI images into an
567. 1:24:29LLM and and it telling me exactly what the radiologist told me. Yeah, Mark Hefner. What's happening? Mark. Hey, Kyle. Happy
568. 1:24:42Friday evening. Coming back to the first question of the evening. Models had bigger differences when they first
569. 1:24:48started. True. I think they are all converging to a single point of truth. Today, those differences are more
570. 1:24:54nuanced. In my experience, Claude is still better at creating written content as long as you train the model on your
571. 1:24:59writing style with an MD file. I use the same MD file across all models, then battle them for the best one,
572. 1:25:07and take it to the last 15% to complete completion. In my opinion, all models have strength and weaknesses. I think
573. 1:25:13that's absolutely right. And I there's there's a new term that they're talking about called the harness.
574. 1:25:21And I the the harness has always been there, but I think it's more significant in agentic tools. The the harness is the
575. 1:25:27thing that the models kind of sit inside and there's all the it's it's like the application layer um that interacts with
576. 1:25:35all the different models sitting underneath it. Um and those harnesses are where the personality is. And I I think that
577. 1:25:44I think that that's where you're going to see, you know, like codeex is a harness and claude clo code and cloud
578. 1:25:50co-work is a harness and even, you know, just claude.ai, the the app that you use as a harness. So I think it's going to
579. 1:25:57be in the interface as always, right? It's going to be within the interface where the differences start to emerge
580. 1:26:05because I think the models really do start to become kind of a single a single kind of thing. Um, all right. Beautiful. Uh, we're
581. 1:26:14coming on the top of the hour, so have a beautiful weekend. Um, is if if there's any any homework this weekend.
582. 1:26:31Nothing immediately comes to mind. It's Mother's Day weekend. I would say, you know, connect with your family.
583. 1:26:40Think about what you value and who you value and then and then carve out some time for yourself to play with AI. I think
584. 1:26:49those two things I started with of what are some good AI habits. One is just go play go experiment with something that
585. 1:26:58you've never played with. Go play with something completely that you would never consider playing with. Right? If
586. 1:27:04you're horrible at math, go do something with a spreadsheet and AI. Right? If you're horrible at pictures, go do that.
587. 1:27:11And then the other thing is like really, you know, look at who you are and what you value and who you care about and
588. 1:27:18then think about how you might use AI as an amplifier of of that of who you are. That might shift how you think about
589. 1:27:26which tools you play with. Oh, that's good. Chef Kelly, make a magazine layout for your mom in image,
590. 1:27:33too. Yeah. upload a picture of or two of your mom and then put in some random facts and say, "Turn this into a glossy,
591. 1:27:40you know, magazine layout, you know, that that goes across both pages." She'll be blown away. You'll be blown
592. 1:27:46away. Um, beautiful. We took a picture of my mom and had AI make a coloring image for the kids to color in. Yeah.
593. 1:27:54Um, have it make five pages of a graphic novel and it will it'll just all five pages will come out. It's pretty crazy.
594. 1:28:01Have a great weekend, my friends. See you on Saturday. Yes, please go to no ems lol. Okay, go to uh events over
595. 1:28:07here. RSVP show up tomorrow 4 p.m. Eastern time. All right, everybody. Thank you so much. Uh this has been a wonderful week coming
596. 1:28:18back inside the AI salon. I'm really excited about this. Um thank you to Andy Scarantino and Brandon who push pushed
597. 1:28:27me. What? How did I describe it? getting me to to make this shift is like milking a stone. Um, I'm really glad I did it.
598. 1:28:35I'm really glad we're back inside the salon and uh I'm I'm really glad you're all here. I very much appreciate you as
599. 1:28:41always and uh I hope you have a great weekend everyone. All right, peace.