6/24/2025 - Exploring AI Agents, Copyright, and the Future of Work

Transcript

1. 0:02I like it. I like it a lot. All right. [Music]
2. 0:18Hey buddy, you ready? Sit. [Music] Woohoo! [Music]
3. 0:46Every time I see you now, get that look in mine. Every time I see your mouth, [Music]
4. 1:01hear that smile. In the early misty morning that I heard the engine turning, the old thought outside. [Music]
5. 1:19Will you believe in me again today? You will convince me again today. You're leaving this hard time looking
6. 1:33for someone else's golden ring. should say so long Suz [Music]
7. 1:49now don't you cry [Music] so love Susanna [Music] don't you cry for Hey,
8. 2:03[Music] sharing jeans and cigarettes and keeping warm out on the road. Chasing down kind of lifestyle out on
9. 2:18Highway 24. New York State was a rolling breeze in the sunshine with a blue sky falling. Jill of old September pre
10. 2:33[Music] you were leaving me again today. You will convince me again today. You're leaving this hotel looking for
11. 2:50someone else's golden ring. Sh say so long Suz now don't you cry
12. 3:09so long Susanna don't you cry for Hey, [Music]
13. 3:28I just realized sitting in here hearing this little like Hawaiian looking short sleeve shirt playing my guitar, playing
14. 3:34like little folky kind of songs like that. I'm like uh Jimmy Buffett. It's like Jimmy Buffett Night without the Jimmy Buffett songs.
15. 3:50Oh yeah. What's happening people? What's happening good people of the world? Um it is Tuesday night.
16. 3:58There's a bunch happening. A bunch bunch happening. I heard a Tik Tok story. I guess there's a New York Times piece out on on uh
17. 4:13it looks like um OpenAI might be coming after Microsoft on antirust grounds um to maybe get out of their deal with Microsoft.
18. 4:28Um, so we didn't we didn't know that that relationship would would last. You know, we assumed it maybe wouldn't, but you
19. 4:38know, I don't know. That relationship's always been weird. It was like OpenAI got their $10 million and they were like, "Who are you?"
20. 4:48Um, here, let me switch cameras. Let me switch yield cameras. Um, welcome everybody. This is the AI learning lab. I am Professor Shannon.
21. 5:01It's what they call me around here. No, it's not. We don't have credentials. I I really could pretend to be smart
22. 5:17good sometimes. Um, so what we do here is we talk about AI like deep machine learning algorithms and and mathematics and science. No, we
23. 5:32don't. We don't. It's that we we really just talk about like chat GPT and like like video tools that make cool [ __ ] Um, but like
24. 5:44I had chat GPT like summarize a paper that those really smart mathy guys did. I like I we summarized it once on this
25. 5:52channel. So, it's kind of like getting a degree according to an MI MIT study that that studies brains when they use AI. And
26. 6:02they're so much better. They're so much better. So that's that's what we focus on is is the the the good part of the brain after AI. Um
27. 6:14yeah, so hopefully hopefully someone told you to be here and that that this was a really impressive channel. [Music]
28. 6:34Um, we have fun here. We like to kid. We like to joke. We like to We like to uh ring around the rosie pocket full of
29. 6:42posies. I I don't know the rest of that poem, but I know it's some horrific 19th century disease or murder. Someone
30. 6:50poison someone. I don't know. All those nursery rhymes are just a disaster if you actually look at what they mean. Um,
31. 6:56we don't we don't talk about that here. We make we make happy little turtles. Happy little turtles. Like there's our little sticker for
32. 7:08for the irregulars for Serena. Am I late? We're not weird. We're irregular. So, I got this from Tobias yesterday. Let me go ahead and
33. 7:19share my screen. If you go to the AI salon to the irregulars channel, let me see. Did I not share my screen? I don't think
34. 7:33I did. And producer Brandon can do nothing about it because that's on me. But okay. So, here I am at the AI salon.
35. 7:48Got to do the black bar. Brandon's going to yell. We don't want Brandon Brandon to yell. It's very uncomfortable for everyone. All right,
36. 7:59so you come to the irregulars channel and um you know there's share cool things and there's the chat but over
37. 8:05here there's a tab says remembering Serena and if you go there it's got this sticker set that has been generously donated by Tobias who's an
38. 8:15irregular here. Um, and this is in remembrance of Serena Pelichi, who was the the original irregular. She's the
39. 8:24one that coined the term, and we lost her, I don't know, a week and a half, two weeks ago. Um, and this sticker set, it includes the
40. 8:34irregular sticker, the salon sticker, and then a memorial sticker. Um, Tobias is covering 100% of the cost.
41. 8:42Um, and uh, so you pay 20 bucks for this thing and you get those three stickers and all of the proceeds of this go to
42. 8:51Serena's family. So if you want cool stickers and you want to support the channel and support a really amazing amazing, you know, person
43. 9:05uh, and and her family, then then go do that. All right, go do that. Sorry I missed the songs. I was having
44. 9:15chat GBT tell me how to add chat GPT to my iPhone lock screen. Goodbye Siri. There you go. Oh, that's a good idea.
45. 9:22That's a really good idea, Sharon. Really good idea. Valerie Cox got hers today. Yeah, the stickers look great. I
46. 9:28gotta say Tobias like the Sometimes with stickers you get them on like either crappy vinyl or they're just printed
47. 9:36bad. Like they're really saturated. They're really good. They're really really good. So, okay. Um, all right. Let me let me go back out.
48. 9:47Let me go back out. We're going to talk about what are we going to talk about tonight. There there's so much going on.
49. 9:54Um, I didn't read this myself, so I'm hearing it secondhand from Brandon who's hearing it secondhand from some guy down
50. 10:03on the street corner. So, th this is not reliable. This is more like chat TMZ, but apparently this is true. Um, so, so
51. 10:14Black Bar. Yes, sir. Sir, yes, sir. Um, so last night someone asked the question about I heard MIT did a thing that said
52. 10:22chat GPT makes you stupid. And I kind of lost my [ __ ] and went off on it. And basically what I said was what the MIT
53. 10:30study proved with a sample size of 55 people or 53 people, whatever. It's not enough to be a science scientifically
54. 10:36relevant, but it's MIT, so whatever. Um, but what the study basically said was if you don't actually use your brain, uh,
55. 10:47you don't learn anything. Like, so if you just have chat GPT output something and you don't look at it and just hand it in,
56. 10:55you don't learn anything. Okay. Wouldn't the same be true if I took someone else's paper and photocopied it with my name on it?
57. 11:08same argument, right? But because it's AI, we can go, see, I told you AI was evil. Okay? And then it said it said
58. 11:14some other things. If you know, basically, if you hand wrote it, it was if you like wrote it and did your own
59. 11:19research, your brain was better. And if you sort of did it a little bit and halfass did it, it was not as good. And
60. 11:25then if you were lazy, you got no benefit and your brain got worse is what the the argument was. But here's the
61. 11:33fascinating thing. This is what I got from Brandon. So again, if this is not true, you're going to have to go after him, not me.
62. 11:40See, I take great pride in not having to take responsibility for things. Okay. Apparently MIT, this is so [ __ ] good. MIT buried
63. 11:54in the in the paper that they released an Easter egg that if you use AI to summarize the paper, it actually omits the fourth conclusion.
64. 12:09So there were three conclusions made, right? You're a dumb dumb, you're a mid dum dum, and you're okay. And there was
65. 12:15a fourth conclusion in the paper that if you use AI to summarize the paper, it omits from the output, right? They put a
66. 12:24they put a they prompt injected their paper so that it would do that. The fourth conclusion was if you're actually
67. 12:32engaged and thinking critically and you use AI, [ __ ] gets better. [Laughter] But 100% of the major media sources that
68. 12:45that reported on that thing that said it makes you stupider missed the fourth conclusion courtesy of Joseph
69. 12:58Holston on Tik Tok. Joseph Holston said that. Okay, perfect. But anyway, so I so I haven't seen it myself. I haven't I
70. 13:07haven't looked at it. We could probably go find the the prompt ejected line, but it's [ __ ] That's [ __ ] brilliant.
71. 13:13It's [ __ ] brilliant, right? It it it like that that Easter egg proves my point of the of the the thing I was
72. 13:24yelling about last night, right? Like if you just basically if you're an idiot and lazy, you won't learn.
73. 13:36Okay. I don't think we need to study for that. Oh my god. Isn't that brilliant, Jeff? Brilliant. Okay. All right. Fantastic. Fantastic.
74. 13:51Yeah. Yeah. I don't know what we're going to do then there. Yeah. [Music]
75. 14:08There's been something, baby, I've been trying to say for an age and it seems I don't know how [Music]
76. 14:19with a past and a future now surrounding me. [Music] Surrender to whatever chief can be found. There's been a little trouble
77. 14:34since you came to my rescue. And if you like how of the rest, I would have quit you long ago. But I couldn't do that.
78. 14:51Oh, tell me now. Women and wife never went too well. Make a man crazy, make him cold as hell. I a woman that you wish me well,
79. 15:07but I'm spite of trying. Still going to have to find my own way through. [Music]
80. 15:21A little women and wine from Martin Ston. That was Martin Ston. Martin Sexton singing away, singing away. Um,
81. 15:33a major copyright ruling came down today and we're going to we're going to hopefully they didn't prompt inject the the court ruling
82. 15:42because we're going to take the court ruling. So in the Anthropic court ruling, they basically said that uh
83. 15:51Anthropic cannot be held liable for training on copywritten books, but they can't have stolen them, I think, is kind of what it says. I'm
84. 16:04I'm not clear. I think what that's got to do with is damages. Basically what the what the ruling says is um you
85. 16:14can train large language model. It's fine to use previous work on the input side where they're going to judge you is on the
86. 16:23output side which makes sense. That's the way copyright works, right? They're basically saying the model itself does
87. 16:28not does not constitute an output derived from the inputs or something like that. which makes sense if you
88. 16:34understand how things get embedded and how they get transmogrified I don't know if that's a word I'm pretty
89. 16:40sure it is into tokens um you know in thousanddimensional mathematical space the the original work cease to exist so
90. 16:50if you output something that's identical then you can then you can be held liable but not for the model itself apparently
91. 16:55but if they stole it then the the things are higher so I don't it's law. It's law. So [ __ ] happens, you know,
92. 17:05with law. People people come up with words and then other people go, "See, now you got to pay money." So that happened. Um,
93. 17:19one thing we are going to play with tonight is 11 Labs came out with uh 11.ai. If you go to one.ai, AI that is 11 Labs new voice
94. 17:34MCP tool using talkie thing. What do they call that Brandon? A talkie agent. A voice agent. I think it's a
95. 17:46voice agent. Yeah, voice agent. Right. I got Everybody's got to stop using the word agent. Everyone's talking about everything's
96. 17:55agentic now. It's all agentic AI. We don't know what that means. And now you got voice agents. Didn't they used to
97. 18:03call them chat bots, but now you can't call them chat bots because chat bots have text and this has words. I don't know. Anyway,
98. 18:11it is quite simple to set up a thing you can talk to that uses tools. And if you know the configuration of an
99. 18:21MCP server, which an MCP server is basically a way for large language models to talk to websites and applications and things like that, it's
100. 18:30kind of like an API but a much simpler a version of it that distills it what it can do down to the to the large language
101. 18:39model so it can understand what it can do. Um, so we'll go play with that. We'll go set one of those up and and
102. 18:44look at it. That's pretty cool. [Music] Veil and wash. What's happening? Thumbs up wavy hand. Thumbs up wavy hand. [Music]
103. 18:59Isn't it just a slack kind of? What the the 11 laps thing? Well, slack is one of the things you can connect it to. It's
104. 19:07it's not really that. But as I was setting it up, there is a massive [Music]
105. 19:26there's a massive undefined what I would consider a security risk until it's proven otherwise. Um, I sat down to make this agent and
106. 19:40they're like, "You can connect your Slack to it." I was like, "Oh, cool." And you can connect your Google calendar
107. 19:45to it. I was like, "Oh, cool. And you can connect your Gmail to it." I'm like, "Oh, cool." And then I realized, hey,
108. 19:50wait a minute. In all three of those channels, we talk about our pharma clients, right? And I don't know what the
109. 20:02data agreement is between me and 11 Labs, right? Are they guaranteeing that the data flowing through their chatbot to answer my questions?
110. 20:17Actually, it says Oh, good. Actually, it says no. That no what? No, they're not keeping the data, sharing
111. 20:27the data. No guarantees. Okay, good. Yeah. Okay. So, so thank you, producer Brandon. Producer Brandon confirmed my worst fear that
112. 20:41what MCPs are going to do. If you don't know what an MCP is, it's called uh it's the it's an acronym for model context
113. 20:47protocol. Was created by Anthropic. And it basically allows you to take any website or tool and and give access to a
114. 20:56large language model to it. Makes it really easy to do that. Okay, great. If I just connect my [ __ ] to this talkie
115. 21:05agent and it's passing through 11 Labs or Bob's agent hut or some Chinese super slick agent thing they did who Who's using my data for what?
116. 21:22Like, is it possible my data gets it my proprietary conversations about my pharma clients get swept up in that and that ends up in
117. 21:31some [ __ ] model? So, like, and and here's the thing that's going to be weird is it's going to be so easy to set these things up
118. 21:40that are that are super powerful that I think you're going to have people inadvertently hose their data like we did in the early
119. 21:48days of chat GPT. Uh, but anyway, we'll go play with that. [Music] Let's see. Uh, what is it? Eliza. Eliza
120. 22:04Eliza Eliza sent a gift and became the number 15 fan in the fan club. Thank you. Thank you for being a fan of the channel. Eliza.
121. 22:16Eliza. Eliza. Elisa. Eliza. I should never try to pronounce names. It's just not good for anyone involved.
122. 22:27I try. Um, aren't they training on general consensus versus specific dogs? Um, it's possible, but like the the point is Kelly, if you
123. 22:43can't be sure, if okay, my company has master services agreement in place with Merc, Novartis, Fizer,
124. 22:58BI, BMS. Like we've like we've got significant legal contracts with like top 10 pharma companies that basically say if you breach our data
125. 23:15we're going to sink your ship. Guys don't kill your battleship. So we can't [ __ ] around with it, right?
126. 23:23We can't just go, "Oh, we were just prototyping something and yeah, sorry that thing got out that was, you know, protected information."
127. 23:33That that ain't going to cut it. So, I don't know. All right. It's pronounced ah Al Ali Zade. Oh, cool. Ali Zade. Nice. Nice name. Ali
128. 23:46Zade. Welcome, welcome, welcome. If anybody has any questions about AI, I will do my damnedest to answer them. Um,
129. 23:54I will often have opinions on stuff. Um, I occasionally have answers. Even more rare, those answers are right.
130. 24:05It's It's happened a couple of times. I've I've definitely said something correct occasionally. It's been two and a half years. Just
131. 24:16mathematically, you'd have to Um there and there's so much else going on. What was the other thing that I sent
132. 24:26you, Brandon? There was 11 labs. There was Oh, yes. Okay. So about
133. 24:41for those who don't know, he will answer them but not in any particular order. Welcome to chat add. Yes, thank you.
134. 24:49Thank you, Tom. Thank you, Tom, for for uh for clarifying the kind of uh the the level of quality we've got here at the
135. 24:58AI learning lab. Um I will eventually get to answers. Um OpenAI about two months ago made this little this little uh announcement
136. 25:12and they said we are going to release. So, okay, let's go back six months. Well, no, let's go back two years.
137. 25:21Two years ago, three years ago, whenever Elon um Elon was one of the original co-founders of Open AI, right? It was founded as a nonprofit.
138. 25:35Elon was the one that said they should turn forprofit and then he left. And then he started publicly publicly
139. 25:43bitching about the fact that they're not a they're not acting like a nonprofit and they should be sued. Anyway, that
140. 25:48all that stuff all happened. And about six months ago, eight months ago, something like that. Elon as he was spinning up XAI
141. 25:58said, "Well, if OpenAI basically had their [ __ ] together and if they had any integrity whatsoever, they would release
142. 26:06an open-source model." Well, about two months ago, old Samama Sam Alman said, "Hey, we're gonna we're going to release
143. 26:14an open- source model." And then they kind of went silent. It was, you know, it was like the the submarine came up.
144. 26:20We're going to do an open source model. And back down it went. Not to be heard of again. There were some flying fish on
145. 26:27the surface. I think we saw a porpus, but not to be seen. That submarine was gone. I don't know why my language is so
146. 26:36flowery tonight. I don't know what's going on. Doesn't It's not Who knows? Today there are strong hints. There was a
147. 26:48little uh uh a a poll put out from Sama. We're on Kyle 4.7 experimental. Exactly. This is This is definitely Kyle 4.7
148. 27:01experimental tonight. Um, we went Oh [ __ ] I just lost my train of thought. So, oh, oh, he put out he put
149. 27:14out a um a a poll that basically said, um, how long do you think it will be before we have an 03 level reasoning
150. 27:26engine that's small enough to fit on a phone? And it was like 2024 20 yeah 2025 26 27 28 right and people were voting
151. 27:37and then I voted for like 27 and then when I did my voting like most of them were 25 and then this afternoon
152. 27:47it basically turns out that it looks like what that model's going to be is it's going to be it's going to have the performance
153. 27:55of of GPT 03 mini which is, you know, a really strong model. Like it's it's it's better than 04. It's a reasoning model. It's better
154. 28:07than or 40. It's better than 40. Um it's a reasoning model. Um and it's going to be small enough to put on an iPhone
155. 28:19and run it locally. If you don't know what run it locally means, right now the thing I was just discussing about if I'm
156. 28:26doing something with 11 Labs with all these hooks to these data sources, I have to send my data to a large language model to have it
157. 28:37processed. That happens out on server farm somewhere, you know, in the desert and I don't have control of that data.
158. 28:46This model is apparently going to like run on your iPhone. So you can have something as powerful as 03 mini where
159. 28:53the data never leaves your phone coming from open AI like soon like like it feels like it's eminent. We are
160. 29:01supposedly safe if we opt out of memory. Yeah. But that's so so um source camp that's that's only for open AAI like I
161. 29:12don't know what I think I think 11 Labs uses um OpenAI servers but we're using the API so we didn't opt
162. 29:24out of memory the only thing where you can opt out of memory is in the chat GPT app itself when you're using these
163. 29:30third-party services that put something in front of other things. Oh, yeah. Oh, and and the other thing
164. 29:38source camp uh Brandon just reminded me that there's now an injunction because one of these copyright lawsuits, the New
165. 29:45York Times copyright lawsuit against OpenAI, there's now a court injunction that OpenAI has to save your deleted
166. 29:54chats, your temporary chats, and your private chats. All of the chats, all of the chats that they said in their
167. 30:05Swiss cheeses like privacy agreement that they say they're not going to do anything with, they now have to store
168. 30:13it. Even the API. Yeah, even API calls. So, so we're in data weirdness. We're in data weirdness. All right. Um, I need water.
169. 30:31So, is that the injunction order on fair use? Yes. Okay. iPhone will have Open AI. Oh, cool. Android will have Gemini.
170. 30:42Yeah, that makes sense. But has has has Google come out with sort of a reasoning model small enough to fit run locally on
171. 30:54a phone? If not, they probably will. I I don't have This is just purely chat TMZ. Okay. Um but they've rumored for a while
172. 31:04that um the Pixel 10 um which is their next flagship that comes out this fall will have some sort of ondevice
173. 31:14um processing. Yeah. Um and that should align with the timing of this announcement of the O3 um coming to
174. 31:22Apple devices. Yeah, it makes sense. Like like you got to figure they've been optimizing like the 03 mini model we've
175. 31:29had for what eight months, six months. Eight months probably probably eight something like that. Um so they they've
176. 31:36been probably working on optimizing these things for you know close to a year if not a year. So so yeah I I would
177. 31:44think that both Apple and and Google are probably going to get there at a similar time on that one. So that makes sense.
178. 31:49Cool. Thank you sir. Um, okay. Let's um let's go let's go to the videotape. Do I have a mouse? Where's my where's my
179. 31:59mouth? Hey, Marge. Marge. Marge. Okay. All right. So, here we have Yan Court order. So this is the anthropic
180. 32:15uh the anthropic uh artificial intelligence firm downloaded for free millions of copyrighted books right
181. 32:25in digital form from pirate sites on the internet. Okay. Can I get an injunction? Oh, for the black bar. That Brandon, he's got the comedy
182. 32:41tonight. Okay. Um, so what I'm going to do is I'm going to do what the MIT folks knew everyone else would do with their paper is we're going
183. 32:57to use chat GBT to read it. Um, so I got to have a chat GBT window. There we go. New chat.
184. 33:11And then in there. And then what did I do? Did I copy all that? Was that That was a dumb dumb move. Why did I download this as a PDF?
185. 33:22We'll throw that right on a desktop. Oh, severe weather in Ohio. So, if Brandon goes bye-bye, hopefully hopefully he's not in Ohio's
186. 33:36tornado valley. Um, the weather's been weird. It was It was more than a hundred for two days in a row here in Denver,
187. 33:43which is pretty rare. Um, okay. So, I'm going to upload I'm going to upload that injunction or that ruling. I don't know legal words.
188. 33:58Failed to upload. Please consider having a computer that works. All right. That good. All right. So, it's in there.
189. 34:11So now what we're going to say is um act as a fine parallegal and I want you to give me a plain English
190. 34:28version of what this ruling means and exactly
191. 34:42the argument for why the rulings were reached. I think that's sort of quasi
192. 35:00competent language. See, you don't need to be smart anymore. You just need to be quasi smart. And if you use you if you use words like quasi
193. 35:14in conjunction with a word like smart, you sound smarter. I'm quasi smart. People like, "Oh, I didn't realize he was that smart."
194. 35:29Okay. Plain English summary of the ruling. The context. The AI company Anthropic maker of Claude used millions
195. 35:36of copyrighted books to train its AI. Some of these were illegally downloaded from pirate sites like books 3, libgen, and pli mi, whatever.
196. 35:47They purchased in print and then scanned and digitized, destroying the originals. The plaintiffs, author of several of the
197. 35:56books included for the data sued for copyright infringement. Enthropic asked the court to throw out the case claiming
198. 36:03fair use, which is effectively what I think they just got. The ruling is the this this ruling is the judge's
199. 36:10response. What the judge allowed as fair use. Training AI training AI on books was fair use if lawfully obtained. Okay.
200. 36:25The judge said, "Training an AI model using books by analyzing their structure, writing style, and content to
201. 36:33help the AI learn how to write like a human is highly transformative." Again, if you understand the way the transformer works,
202. 36:44the original work ceases to exist. So you take a book like The Jungle Book or you take a book like I don't know
203. 36:52something more modern, some Tom Clancy novel. You got the whole novel in there, right? What the transformer does is it shatters
204. 37:03that single document into tens or hundreds of thousands of tokens which are fragments of words, punctuation, spaces, fragments of words
205. 37:15that all get clustered. they they get converted into a a a mathematical location in thousand dimensional mathematical space.
206. 37:27And so you might have the word dog that lives in one cluster, semantic cluster, where a dog is the thing you pet and
207. 37:35sings with you when you play guitar. And then the other part of the latent space is the word dog that's related to he did
208. 37:44his girlfriend bad so he's a dog right two different meanings of dog but the original work ceases to exist.
209. 37:52So it's highly transformative but Claude doesn't just spit back the books but learns from them to generate new
210. 37:59content. This counts as fair use. Thank god somebody took the time to figure out the the anthropic lawyers took the time
211. 38:07to explain to the judge or the lawyers or whoever how this [ __ ] actually works because the the trope that you hear people say it's
212. 38:18the world's greatest plagiarism machine. No, it's not. No, it's not. It's not copying. Okay. The judge compared it to a human
213. 38:29learning from books. They can be sued. They can't be sued for absorbing knowledge. Interesting. Okay. Digitizing
214. 38:38purchased print books was fair use. Anthropic bought print books, scan them into searchable PDFs. This is amazing.
215. 38:46Who knew they did this? Uh, and use the digital versions internally. The judge says this was like converting VHS tapes
216. 38:54to digital for storage and searchability, not making extra copies or selling them. That's also fair use.
217. 39:01Fascinating. Okay. X. What the judge said was not fair use. Keeping pirated books in their library was not fair use. Anthropic downloaded
218. 39:14millions of pirated books even though they could have been purchased. They used some for training but kept all of
219. 39:22them forever, even ones they never used. The judge said, "You don't get to steal books just because you might use them
220. 39:31for fair use later. That's still copyright infringement." The judge emphasized there is no special carveout for AI companies.
221. 39:41So that's fascinating. So, so the fact that they kept the files. So, so if I get if I get this right,
222. 39:57I'm confused. Well, we're reading a we're reading a chat GPT thing. I'm not going to even bring it up. Okay. Legal
223. 40:04reasoning summary. Why the judge ruled this way. Yeah, this feels like a notebook LM use case. It does. Training AI models was
224. 40:16considered tr a transformative use, especially since the outputs didn't reproduce the books. And and that's not
225. 40:23the what the argument that that the AI haters will make is I can get it to spit out an exact copy of a book. You can get
226. 40:31it to spit out something really close, but it's generating something new every time. It just is. Um, but building a permanent pirate book
227. 40:44library was not transformative. It was just theft. That's makes sense. Fair use factor number two, nature of the work.
228. 40:52Most of the books were creative and expressive, which usually tilts against fair use, but factor one outweighed this. Again, factor one,
229. 41:03it was considered transformative. C. Fair use factor three, the amount used. Full books were copied. This usually counts against fair use,
230. 41:14but again for training models, this was seen as necessary and transformative. Fair use factor number four, market
231. 41:22harm. Claude wasn't shown to directly harm sales of author's books. However, pirated copies undermined the legitimate
232. 41:31market for licensing books. All right, bottom line. If Anthropic bought the books and used it for AI training,
233. 41:37that's okay. If they stole it and hoarded it just in case they might use it, that's not okay. The court is
234. 41:45drawing a sharp line. AI companies can't use piracy by claiming innovation. All right? So, pay for your [ __ ]
235. 41:53All right? Makes sense to me. You go out and you raise 10 billion dollars, you can pay some some book fees.
236. 42:00All right. There we have it. There we have it. Should we go do something? Should we go make pretty pictures or
237. 42:05something? We go make pretty pictures. We can go make videos. I We go make an agent. We go make a talkie agent. 11 Labs 11.ai.
238. 42:15I think we'll do that. Mind your tabs. Oh yeah, we're going into uh baby videos. We make baby videos.
239. 42:32Okay, I am gonna go [Music] one.ai and I'm going to switch tabs so you're going to be able to hear this.
240. 43:04All right, we got it. There we have it, people. What's he doing? Why is there a Why is there a CD on screen? I thought we were doing AI.
241. 43:15Isn't a CD like really 1990s technology? Oh, right. He's old. He's one of those olden timey people. He probably had like
242. 43:23one of those Walkmans that was like the size of a of a pancake. Carried it on his hip and a little holster. Played CDs
243. 43:30on it. Had the anti-skip technology. He's probably one of those. Okay. So, we'll start one from scratch. So, the way this works,
244. 43:45when you get in here, we're going to say create new agent. First thing it has you do is pick a voice. All right. This is how we'll refer to
245. 44:00your assistant 11. Okay. So, we're going to call this um the irregulars. Irregulars. [Music]
246. 44:15And then here's all the voices we can play with. So these ones it says are perfect for 11 AI. This voice covers a wide emotional
247. 44:25spectrum from the quick brown fox jumps over the lazy dog is an it's not about what we have or what we achieve but
248. 44:31about how by repeating what's I'm a sucker for British accents by repeating what students say. Teachers can
249. 44:37demonstrate that they are listening by. So essentially well what I'm trying to say is that you deserve more because it's not your fault
250. 44:47that All right. Are we cracking codes, solving mysteries, or casually taking over the world? For the good old
251. 44:52American life. For the money, for the glory, and for the fun. Mostly for the money. I like Bert Reynolds. That was
252. 45:01kind of nice. So, essentially, well, what I'm trying to say is that you just by repeating what students say, it's not
253. 45:09about what we have or what we I go with Nova Westbrook. Oh, the other thing you can do here. So, they give you the ones
254. 45:15that are perfect for 11 labs. You can also uh look at any voice clones that you've made. So like I did one of me. We
255. 45:23make our own fortunes and we call them fate, right? We make our own. It sounds thin. Like I I just did it on a shitty
256. 45:30microphone and I didn't really think about it. So oops. What did I do? Oh no. Did I screw up? Yes. Damn it.
257. 45:39Create a new agent. Okay. So we're going to go with uh perfect for 11 Labs. We're going to go with Nova. So essentially, yeah, I
258. 45:47like that. All right, so Nova Westbrook is our is our lady added. Nova Westbrook chosen create 11 Labs thing. Okay, so we
259. 46:03can now talk to her. Hi there, 11 here. Hey 11. I'm ready to jump in whenever you are. What would you
260. 46:14like to explore? I I want to change the name of this thing. Hey there. I appreciate you wanting to personalize things, but my name is 11.
261. 46:25That's just who I am. Is there anything else I can help you with today? Delete agent. Delete agent. We're going to go back. We're going to
262. 46:32start a new one. We're going to go grab perfect for 11 Labs. Wait, why' we lose her? Oh, there we go. There's Nova. she's selected. We're
263. 46:43going to call it um irregular because it sounds like she uses this to just to introduce herself. I just want
264. 46:53to call out that when the robots completely take over, this live broadcast playback is going to be exhibit A of Why? [Laughter]
265. 47:09How can I help? Delete. Delete agent. Okay. So, let's talk to her now. And now her name should be irregular.
266. 47:25Hey there, there. Got something I can handle for you? Yeah. What's your name? I am irregular, your personal assistant.
267. 47:32It's a pleasure to connect with you. You know what they say, don't you? We're not weird. We're irregular.
268. 47:40That's a fantastic way to put it. I completely agree. Irregular sounds a lot more interesting than weird. Don't you
269. 47:46think? It's all about embracing. Okay. Exactly. So, let's hang up on her. Come on. Hang up. Irregular. Why can Understood. Why can I
270. 47:59not hang up on you? Uh, that's a good question. I don't actually have the ability. I can reload you. Ha. Okay. So, 11. So,
271. 48:11we're going to go to irregular. Okay. So, so we now have uh a thing that you can send people to. And I think you can publish these. I
272. 48:22forget how how it works. Custom behavior. You can you can put in uh a system prompt, right? So, you can
273. 48:31put in a custom thing. And then there's custom uh MCP servers. You can customize the color of your um the circle
274. 48:43and then I think you can publish it somewhere but whatever we'll we'll figure that out. But add your first integration.
275. 48:49So if you click on this it gives you these five Google calendar hacker news linear manage your issues in linear
276. 48:59perplexity search the web with perplexity and slack. So, what I'm going to add is I'm going to add perplexity
277. 49:08and I'm going to add hacker news because the other ones I don't trust the data piece. But those are and and then the other
278. 49:22thing you can do is you can add custom MCP servers. So, if you've got MCP servers, I think hey Brandon, do me a
279. 49:29favor. See if you can find I think someone was doing like a Yahoo like server for for all the MCP servers that
280. 49:37were out there. It was like a was like a single interface where you could get all the whatever the information for all the
281. 49:44different I don't know whatever. If you can find that find it. If not that's okay. Um why I guess it's saved. Changes
282. 49:51are saved. Okay. So now Oh mcpservers.org Let's try to Let's try to go get something. mcpservers.org.
283. 50:06This could be good actually. Okay. Only on cling. Awesome. MPC servers web scraping. Oh, no one can see what I'm doing
284. 50:20because it made a new tab. Of course it did. Oh, no. I made a new tab. Yeah. No, we should show people. We
285. 50:29should show the good people what I'm doing. Okay. Okay. So, here I am at this site.
286. 50:47Awesome. MC MCP servers. Bright data everything. Fetch. Get memory. Sequential thinking, time zone, and time conversion. Yeah,
287. 50:59but it can just do that anyway. Um, let's see. File system, cloud storage, productivity, web scraping, official.
288. 51:23Oh, this is [ __ ] exhausting. Let's go web scraping. Fetch appi browserbase. But all these we can get from playright
289. 51:38MCP server. All these we can get from perplexity. All right. This is why we don't do MCPs yet because no one's figured out.
290. 51:46They're all just if you're a developer, fine. You're going to hook into some [ __ ] Whatever. Just
291. 51:53Just ignore that [ __ ] for now. If you're just trying to do some [ __ ] Okay, let's change our sharing.
292. 52:08Okay, so we now have a an agent. Wait, what did I do here? Did I do
293. 52:24something dumb? No. Okay. What's going on? Okay. Okay. So, we've now got this agent
294. 52:42that is connected to Perplexity and whatever that news source was. Do you remember what it was? Doesn't matter.
295. 52:50Hey there. Need a hand with anything right now? Yeah. Um, can you tell me what the latest um stories are for hacker news?
296. 53:03Sure thing. Just a moment while I fetch the latest stories from hacker news for you. So notice it's it's doing a call here.
297. 53:10Yes, that's right. I just ran a quick check to get the latest posts for you. Is there anything specific you were
298. 53:16looking for or did you want me to delve into any of those stories? I'd like you not to use the word delve ever again,
299. 53:22but if you could just tell me the top three stories. All right, understood. No more delve. My apologies for that. That's right. The
300. 53:32top three latest stories from hacker. Okay, got it. The top three latest stories from hacker news are I was able
301. 53:39to recover my money from crypto scammers, a simple way to view office files online, and few Americans pay for
302. 53:45news when they encounter payw walls. I agree with that. Pay walls make me crazy. Um, okay. Um, let's see. I want
303. 53:55you to go find Can you can do this in perplexity? Um, maybe you can help me figure out something that might be interesting to
304. 54:07search for using perplexity about um, marzipan cooking.
305. 54:22Uh, pay walls are definitely a pain. I hear you on that one. So, you're looking for something interesting to search for in perplexity
306. 54:30about marzipan cooking. How about we try unusual marzipan recipes? That might uncover some fun and unexpected ideas.
307. 54:37Just a heads up, this can we do it? Maybe break it down by what are the top three countries that that serve marzipan
308. 54:46and then let's get one recipe from each country. That's a fantastic idea. I can definitely help with that. It'll take a
309. 54:56moment for me to do a search for the top three countries that serve marzipan and then I'll look for a recipe from each.
310. 55:02Just a heads up, this might take a few seconds as I'm searching the web for you. Sea slug of doom. Nova irregular
311. 55:09sounds like she can barely stand Kyle. She might use bad words. She does. She sounds like she's got a lot of attitude.
312. 55:16She's like, "Do I have to do I really have to search for something for you now? This is exhausting."
313. 55:23Oh. Oh, and by the way, so notice notice what it popped up here. It popped up a a permission thing. Do I have your
314. 55:29permission to search the web using Perplexity? You can when you set when you set up the uh the services, there's
315. 55:37all sorts of permissions things in there. No, I've got her muted. Okay, I've got the top three countries
316. 55:43for Marzipan. They are Germany, Spain, and Italy. Now, let's find a recipe from each. This will take a few more seconds
317. 55:50as I search for those recipes for you.
318. 56:12Oh, wait. I got to approve it. Hang on. Approved. That's really annoying. Thanks for the go-ahad. It's understandable that the approval step
319. 56:23can be a bit of a pause. Okay, I'm now searching for a marzipan recipe from Germany, one from Spain, and one from
320. 56:30Italy. This will take a few seconds for each search. I'll let you know what I find. All right, I've got a traditional
321. 56:36German marzipan recipe for you. It typically uses finely ground blanched almonds, powdered sugar, almond extract,
322. 56:42and a touch of water or rose water. Next up, I'll find a traditional Spanish marzipan recipe, and then an Italian
323. 56:48one. This will take a few more seconds. So, what I don't know
324. 57:03what I Oh, sorry to interrupt. What were you going to say? Um, actually, I'll ask it. So, what I don't know is where is
325. 57:12there is there anything that you're finding? Is it stored in my account anywhere? The the transcript of our chat here?
326. 57:23That's a great question. As your personal assistant, I don't actually store the transcript of our chat in your
327. 57:28account. Your interaction with me is happening through the 11 Labs platform and any saving of the conversation
328. 57:34history would be handled on their side. I'm just here for the current conversation. But there's no way I can go look
329. 57:45from my end. No, there isn't a way for you to go back and look at our chat history. I don't have access to the 11
330. 57:50Labs platform storage, so I can't retrieve past conversations for you. Right. Marzipan recipes. I've already
331. 58:02got the traditional German one for you. Now, let's find that Spanish marzipan recipe. Just a moment while I search for it. All right. Um,
332. 58:12okay. I found a traditional Spanish marzipan recipe for you. It typically uses blanched almonds and sugar, often
333. 58:18with egg white to bind it and sometimes lemon zest for flavor. They often calm down, calm down, calm, calm, calm down.
334. 58:25All right, we're going to hang up. Okay. Um, 11 Labs IO app conversation AI history. 11 Labs app. That's an annoying URL.
335. 58:42If you just go over to your 11 labs dashboard and conversational AI, it's it's all there, but it's redacted. No, really
336. 58:55crazy. All right. Well, so okay, so this thing works. Um, I guess here, let me give you my thoughts on this thing. Um,
337. 59:18if you had said, if you had said a year ago that we would we would be able to easily without programming build a
338. 59:32conversational voice app that can use tools in conjunction with the large language model, we'd have probably been like, you going
339. 59:42to be able to do that. It's now trivial to do that. Um, wait, go here. Bury delete. Hang on.
340. 59:57You'll need to change your tabs, but this is way more powerful than they let off on today's announcement. Um, I just
341. 1:00:04uncovered the dashboard for conversational AI. Okay. And, uh, it's really interesting what I'm seeing.
342. 1:00:28It cost us a penny. So far, we've had five calls of a minute and 42 average duration. Oh, this is cool.
343. 1:00:43Nice. So, here's our agents. Oh, look. Irregular. Okay, there's the irregular one we just built. So, okay. So, so I
344. 1:00:52get I get what they did there. Okay, this is starting to make more sense for me now. So, 11 Labs has had agents for a while. it one.ai
345. 1:01:0211.ai. Think of that as the chat GPT for creating voice agents. That's what that's what that is. It's actually writing stuff into this
346. 1:01:16big complicated dashboard. But that's the simple interface. Okay, this is starting to make more sense now. Okay,
347. 1:01:24so here's my thought on all this stuff. That's really good. Thank you for that, Brandon. This makes good sense now.
348. 1:01:31So, there's all sorts of controls you can put in these things. You can Yeah, Brandon. All right. So, so just uh if
349. 1:01:37you could pull that pop that back up on screen. Actually, I can pop that back up on the screen. Yeah. Um the thing that
350. 1:01:41stood out to me aside from the phone numbers, so you could d a phone number and call your agent. Yeah. But also the
351. 1:01:48knowledge base. You could actually rag these things and give them data. Yep. And I don't know that that was called
352. 1:01:56out in the it's not called out in the 11.AI UI and you wouldn't know about it. It's not Yeah, it's not it's not in the 11.AI interface,
353. 1:02:07but but you could come back here and modify like we can I assume I can come back in here. I can go in and edit.
354. 1:02:14Yeah, I can edit this. So there's voice. I can change settings for voice. There's evaluation criteria. Which data
355. 1:02:24collection are we going to use? There's security settings. So, you can do full-on stuff here, right? And then
356. 1:02:33widget. Let's see. Authentication notice. And then, yeah, this is where I think you can do a Twilio phone number
357. 1:02:40or something like that. Yeah. So, these are really powerful. I think these are going to be especially in the customer
358. 1:02:47service space that I play around in. Yeah. this is uh you know Yeah, this is my jam. Yeah. Yeah, exactly. So, so
359. 1:02:55that's that's a really good point and that's kind of where I was headed with this. Um this makes it super easy to make these
360. 1:03:06kind of voice agents. It's not going to be immediately apparent to all of us why we would need a voice agent,
361. 1:03:15but over time we're going to start to figure it out. So Brandon in the in the space he works in. He understands the
362. 1:03:22value of this immediately, right? Um, imagine Kelly Camp, I just saw Kelly Camp in here. A lot has happened in two
363. 1:03:31years. Wait, hang on. I got to do physical dexterity on TikTok. Okay, I did it. I'm getting better at that.
364. 1:03:38A lot has happened in two years. We should have written down predictions. You know what's funny, Kelly? We have written down predictions.
365. 1:03:45I've got every recording since April of 23 recorded and transcribed. So, we've actually written down everything we've ever said in here. Um,
366. 1:03:58plus the AI salon meetings. Um, we've got that all that history as well, which is super cool. There's probably about 10
367. 1:04:04books we could extract out of all that data. Um, but imagine someone like Kelly Camp where she she meets with these different clients
368. 1:04:14and one of the things she does right now is she'll build them custom GPTs. Well, maybe she builds them a custom agent
369. 1:04:23like this where uh, you know, someone can call the bar and it gives the hours and and the drink specials of the night
370. 1:04:29or something like that. Or maybe she does a digital twin of herself and they can call her agent any time. And then I
371. 1:04:38don't I don't know if the if the agents have call forwarding, but I assume at some point they will where if the agent
372. 1:04:44can't answer it, it could actually forward, you know, a text over to uh to Kelly. I don't know if it can do that
373. 1:04:49yet, but that's where we're headed. So for me, this starts to represent kind of what VO represents for video VO3
374. 1:05:03where you can now just make a video with actors talking or animated animals talking or explosions happening. Um or
375. 1:05:11midjourney um animation, midjourney video where you're just people know how to make images in midjourney. They can
376. 1:05:17now just hit the animate button. they don't need to actually do a prompt and it gives them really good results.
377. 1:05:25All the different disciplines are going to start to have these kinds of easy interfaces where literally anyone
378. 1:05:33can do something that a year ago would have seemed impossible to even exist. It is now trivially easy to make,
379. 1:05:44but it's not obvious what to do with it. So again, that goes back to if you want to have value in this world. Tik Tok pin
380. 1:05:53if uh let's see what's it say. We should have our own AI learning lab LLM. Hey Pavin. Um did I create one of those? No, I created
381. 1:06:05a I've got a um I've got a notebook LM notebook for my content evolution collab calls and that's it's really powerful. Um,
382. 1:06:20we could probably do that, but it's a lot of data. So, I think it probably would fall into um I think it would probably fall into
383. 1:06:31custom software development. We could probably do something with digital gods. Here's a thing that I cannot figure out
384. 1:06:37why they haven't done better. You've got Gemini has a multimodal model that can read video. So it can read like an
385. 1:06:47hourong video with a million tokens. You can do an hourlong video with two million tokens. You can do twohour videos.
386. 1:06:55And they've got all the transcripts of all the videos on YouTube. Why we cannot right now in Gemini just
387. 1:07:04say, "Hey, where were all the times in the AI learning lab they mentioned irregular? Where was the first time the
388. 1:07:10word irregular showed up in this channel? Like, why can we not do that? They're [ __ ] Google. They invented the transformer. They own
389. 1:07:21all the videos. So, that'd be cool. Complete with bloopers and outtakes. Exactly. And because it's multimodal,
390. 1:07:33you could go find the clips in the transcripts and then say find, you know, find any clips where the guy on screen
391. 1:07:39is wearing a pink bow and it would go find that. Like you can do that right now in a little bit of an ad hoc way.
392. 1:07:46But why they haven't built that yet? I don't know. Oh, what did I do? Did I screw up your ability to do a post-it
393. 1:07:51note? Yeah, sorry about that. Okay, wait. Go here. Okay, there you go. Okay, back to the yellow. Sorry about that.
394. 1:08:00Your previous thought, if you're someone who knows how to welcome to chat. Oh yeah. So, we're going to have all these all
395. 1:08:14these tools, all these building capabilities. We're gonna we're gonna be able to Ann Murphy can just spin up a video game and
396. 1:08:24Brandon's four-year-old can make a song and I can Kelly Camp can just go spin up a voice agent and have a custom phone
397. 1:08:31number for it and it just [ __ ] works. It's just good and it works. The people that are going to be most valued in the workplace,
398. 1:08:46future of work are those people that can think critically about the use cases. What are we trying to accomplish? Who
399. 1:08:57are we doing this for? How do we want to affect them? Right? People that are thinking strategically and critically
400. 1:09:03about what we're trying to put into the world. Not necessarily people that know how to code the [ __ ] tools or build
401. 1:09:10the thing or go into the agent, the conversational AI agent dashboard and do all the configuration. All this configuration stuff is going to
402. 1:09:22drift down to the bottom of the sea and floating up on the surface of the ocean are going to be these easy to use little tools.
403. 1:09:31And because everyone has access to them, 99.99% of people are going to do the same stupid seven things with every tool.
404. 1:09:43Here's the same stupid videos that everyone put out. Here's the same stupid AI agents that everyone put out. Here's the same stupid
405. 1:09:53uh casual games that everyone vibe coded. Who's going to rise above the noise? Just like the MIT study, the lazy people
406. 1:10:06that just push the button, those that just push the button, their shit's just going to be like noise.
407. 1:10:20The people that can think critically and creatively and have their own personal point of view. Be like Rick Rubin. Be
408. 1:10:30like Gordon Ramsay. No, it's not good enough. No, it's not good enough. Do it again. Right? Those are the people using AI
409. 1:10:42that are going to generate the stuff that's really valuable and they are going to be incredibly valuable people
410. 1:10:47because most people will not think like that. Most people are just going to hit the easy button. Oh, look what I made. I
411. 1:10:54made a conversational agent. I think I'm going to start a business. It's going to be called conversational. It's going to be called the
412. 1:11:02conversationalizer. It's Hi. So, I am a I'm a professional AI agent a agent author and I I created the conversationalizer.
413. 1:11:16And what it is, it's a collection of five conversational agents. What's unique about my service is that um
414. 1:11:25each of the five agents kind of has its own personality and they have different voices. You know how much of that [ __ ] there's
415. 1:11:37going to be out there? It's gonna be exhausting. Pink bow the way of code. Oh, what is the pet rock? Kyle, I'm using the agent where I
416. 1:11:53can find the conversation chat box and copy and paste the conversation. Very cool. Oh, okay. That's good. Yeah,
417. 1:12:00there's probably an interface for it. I just it, you know, it's just not at 11. Yep. Beautiful. All right. Um, I have a
418. 1:12:11bit of a headache. The Kyle 4.7 gave me a bit of a headache. So, I think I'm going to wind I think I'm gonna wind her
419. 1:12:18on down. Anybody have any questions, thoughts on what we've been talking about here. Agents of the round table. I like it.
420. 1:12:27Very nice. Very nice. With pink. Yeah. Now with pink bow technology. Hi Hi. Hi. My name is Jenny Sullivan and
421. 1:12:37I'm I'm the founder I'm the founder of the conversationalizer. It's it's a it's a family of of AI agents that I programmed with my own
422. 1:12:51self with my hand hands. Now with pink bow technology, pink pro users get fuchsia.
423. 1:13:08Pretty good. Pretty good. Did I miss the bow? Joy, you missed the bow. His hair looks great. Well, I Yeah, I don't know
424. 1:13:14what's going on. It's It's It's a little moist, I think. I think is what's happening. Oh my god. I'm not I'm glad I'm glad I'm
425. 1:13:24not the only one who doesn't quite grasp MCPs. I grasp them conceptually, but we're we're at this stage. Okay, here's where
426. 1:13:34we are we are with MCPs. Anyone who's technical has made an MCP server. So, there are literally, you know,
427. 1:13:49tens, hundreds of thousands of MCP servers out there, and you go to a site like mcpserver.org or whatever the hell
428. 1:13:57that was, and it just lists them all. They're not ranked. There's no quality control. They're all done by engineers. So, they've got
429. 1:14:08shitty titles, shitty descriptions. You don't know what they do. You could probably read the manual and figure out
430. 1:14:14what they did, but there's so [ __ ] many of them that you just there's no way to discover them, right?
431. 1:14:20And so we need another year or maybe two for the MCP geekery to to stabilize the technology so that another round of developers will
432. 1:14:33develop the better interface for them. And then it'll probably be the next round of technology after that where you
433. 1:14:41have really easy interfaces where you'll be able to pull in the top 20 MCP servers that do everything you need them
434. 1:14:47to do and just easily connect them and you're off to the races. So the the reason they seem useless right now is
435. 1:14:54that we're just we're we're two year year and a half two years early. So they they will make more sense in the future.
436. 1:15:02Tik Tok something MCP server. Oh, you want to know what an MC MCP server? Okay. So, about four months ago,
437. 1:15:16somewhere in that neighborhood, Anthropic launched this thing called model context protoc protocol MCP. And if you know what an API is, it's
438. 1:15:32it's in the neighborhood of an API. So an API is stands for application programming interface. And if I' if I'm creating an iPhone app
439. 1:15:42and I want to put a map in it, rather than me having to launch satellites into space and write my own map application
440. 1:15:49just so I can have a map in my restaurant finder app, I can just use the Google Maps API and they give me a special key and I
441. 1:16:01basically just tell my app, hey, when you when when you go to display the map, instead of, you know, us writing the
442. 1:16:07code for that, I'm just going to have this AP API that that will talk to Google's servers that has all the maps
443. 1:16:15and it it will put a request, you know, I'm in Denver, Colorado at this coordinates. Go get me that map and
444. 1:16:21it'll go over here and fetch it and bring back the location and display the map. So that's what an API is for an
445. 1:16:27application to talk to another application. An MCP is kind of like a simplified version of that that allows an English language um
446. 1:16:40large language model to understand what in in the case of Google Maps what Google Maps can do and what services are
447. 1:16:50available but it does it in a in a like a structured language kind of way that a a large language model can understand.
448. 1:16:58So when you talk to chat GPT or you talk to 11 labs agent, it uses that MCP connection just like an API connection
449. 1:17:06between two apps. This is between a large language model and an app is effectively what it is. So it's just
450. 1:17:12it's just another way to communicate to to pull data from other sources and and pull services from other sources. Okay.
451. 1:17:23So couple of things before I go. If you have not joined the AI salon, um Brandon just put it up there. Go to that URL, the salon.ai,
452. 1:17:36scroll down a little bit, click on join our community, and that is going to take you. [Music] All right, show my screen there,
453. 1:17:49Brandon. that's going to take you to the AI salon community and you should join. And there's a nice little welcome
454. 1:17:59sequence here. We start with a a welcome video from myself and Leah Faston who co-founded the salon with me. We talk
455. 1:18:06about the cycle of AI readiness. Um, play first, mindfully create, generously lead. We talk about that. We talked
456. 1:18:12about the five stages of AI adoption. Um, you can introduce yourself. That's step two. You should definitely go
457. 1:18:18introduce yourself. you can learn our values. Um, these are really important. So, this is this is a community of about
458. 1:18:253,000 people who are curious and playful, have empathy, are brave, and are very very very generous. Um, and so,
459. 1:18:35so there's kind of this little sevenstep welcome process. You can find out when our next events are. So, a week from
460. 1:18:41tonight is is our next uh AI Salon Presents meeting. Um, and and uh and that and then there's things like
461. 1:18:48community corner. There's a news area and then we've got this area called the mastermind. So, we've got a
462. 1:18:53subscription-based area. If you want to step your AI game up to the next level, the AI salon mastermind is a very cheap
463. 1:19:01in my opinion. Um, subscription area of the salon that um you can develop deeper relationships. Um there's uh sort of
464. 1:19:13specific clubs and uh education areas and um educational resources like recording archives and things like that.
465. 1:19:22Um there's a there's a private chat area for mastermind members. So think of the mastermind as just a more focused more
466. 1:19:30intense focused version of the salon. So the salon is free. The mastermind is this more focused spirited uh
467. 1:19:38subscription area. So, if you haven't checked it out, go check that out. And in fact, in the in the first seven
468. 1:19:44welcome things, step seven is join the mastermind. And if you click on that, that'll take you over and explain to you
469. 1:19:50what it is, why it is, you know, who it's for, all that sort of stuff. Okay, fantastic. And then tomorrow,
470. 1:20:00tomorrow, let's say, let me go there. If I go to um aire readiness project.com
471. 1:20:18it takes me to the ar readinesspro.com. Okay. Um so every Wednesday Ann Murphy and I do um this podcast and we're we're kind of
472. 1:20:29in quasi quasi quasi pre-launch. We're somewhere in a transition between pre-launch and launch. Um, we do this thing every week that's an
473. 1:20:39hourlong podcast we have on great guests and things like that. So, if you go to a readiness project.com, you can get the
474. 1:20:45links for the live streams and things like that. Okay. Um, future food, can you please explain that again? Just kidding. MCPs, I know it's
475. 1:20:57funny. Every time I explain MCP, about the time I finish the explanation, someone new comes in and like, "What's an MCP?"
476. 1:21:06All right. So, that is tomorrow at uh 3 p.m. Pacific, 400 p.m. Mountain and 6:00 PM Eastern is is Ann Murphy and my podcast. Okay. Um,
477. 1:21:20if you're if you're over there on the YouTube, follow at learninglab-ai. Learning lab-ai is the URL on YouTube.
478. 1:21:31Uh, if you want to watch recordings of these lives, all of them are on YouTube. Um, please go watch them. Uh, we want to
479. 1:21:38get our view hours up and our subscriptions up. So, if you haven't subscribed over there, shame on you.
480. 1:21:44Shame on you. Well, I just got here and I'm trying to figure out what this channel's all about. Is Is he the
481. 1:21:50professor? Who's doing the teaching here? This doesn't feel very teachy. Ah. All right. Don't go changing everybody. Um I hope
482. 1:22:06you're all well. Hope you have fun tonight. Learned a little bit. Talked a lot. Sometimes how it goes. All right.
483. 1:22:14I'm going to get on out of here and go nurse my headache. Peace out. Have a good one. Bye.