8/25/2025 - Experimenting with AI Music and Spoken Word Art

Transcript

1. 0:00Woo! [Music]
2. 0:16[Music] [Music] Woo! Oh,
3. 0:31[Applause] [Music] [Music]
4. 0:56yeah. [Music] [Applause] [Music]
5. 1:15[Applause] [Music] [Applause] Well, I heard there was a zigger David played an alarm.
6. 1:31You don't really care for music, do you? With go like this. The fourth, the fifth mult [Applause]
7. 1:47king composing. Hallelujah. [Music] Hallelujah. Hallelujah. Hallelujah.
8. 2:03Hallelujah. [Music] Good evening, good people. Happy Monday.
9. 2:18We are having trouble streaming to LinkedIn. This may be an issue with LinkedIn. We'll keep trying. Huh. Well, look at that. Oops.
10. 2:29Oops. I'm in the wrong live. Hallelujah. What are our plans tonight? Um, I have an idea for something fun to do with
11. 2:43Sunno that I experimented with in the early days of these music things and I haven't tried in a while. Um, so we could do that. [Music]
12. 3:24[Music] [Music] Wonder if is LinkedIn still messed up? No, it doesn't look like it. I mean, it
13. 3:39looks like it's messed up. Just remove the destination from the stream and read it. Remove Remove.
14. 3:56Read it. All right. Now, it looks like we're
15. 4:10streaming on LinkedIn. Have you used Superbase OpenAI API with lovable yet? I've used Superbase with lovable. I
16. 4:24don't I don't think I've used a specific OpenAI API with lovable, but I assume I assume that just means
17. 4:33you can hook into Open AI and do content generation and things like that. Is that right? I'm still trying to get used to my
18. 4:48microphone. This is my microphone. Oh, let me switch these. Producer Brandon had a little computer meltdown right before we went we went on
19. 5:03[Music]
20. 5:23Hello Tobias. What's happening? What's shaking good people? Wait. [Music]
21. 5:48[Music] Wow. Hey, [Music]
22. 6:21[Music]
23. 6:50[Music]
24. 7:05Producer Brandon is back. It looks like he's at least back on post-it note duty. Yeah, it looks like he's back. Okay, we
25. 7:14have producer Brandon back. Uh, do me a favor on Tik Tok. Share the live if you'd be so kind. Uh, if you're
26. 7:22on Twitter or LinkedIn, tell friends. Same with YouTube. saw a great breakdown of the month of breakdown of the month of computer
27. 7:36costs. Everyone has bought into reasoning that is so hard to generate thousands of times tokens even as they get cheaper. Oh,
28. 7:49compute. Yeah, math. Yeah, I I mean listen everyone I know has a different take on the energy consumption and
29. 8:04it is clear that the training the training of these models takes a lot of energy but like the training of GPT4
30. 8:16from what I can understand from like a a recent thing I heard it's kind of the equivalent of flying a seven a 747 8 hours
31. 8:27for one flight to train all of GPT4 and then inference is even cheaper. So I don't know like at this point I know
32. 8:37they need lots and lots of power and they need lots and lots of GPUs to serve up I I mean Open AAI by the end of this
33. 8:45year is likely going to have a billion weekly users. So, they've got a they've got a lot of stuff to do, but you know,
34. 8:54the whole thing of like it takes a gallon of water to do a prompt is all [ __ ] I think it takes five drops or something
35. 9:01like that. So, anyway, whatever. I I I mean, here's my thoughts on the on the old um on all the energy stuff is that
36. 9:15in relatively short order, hang on, I'm just repositioning things here for for the tick tockers. We got a bit of a mess over there on the
37. 9:26Tik Tok. You can't read the comments. It's a mess. Um, probably within two years, Google was will certainly have it. Apple
38. 9:40may or may not, but we're going to have local large language models on our laptops and our phones that are probably
39. 9:49about as good as GPT4, maybe even GPT5, maybe even reasoning models, I don't know. So a lot of the energy consumption on
40. 9:58inference is going to be offloaded to the however much energy your phone takes, right? Um and then the bigger
41. 10:07models are going to get more and more efficient. So So is it an issue right now? It might be. I don't know. Um should we be concerned
42. 10:20about it? I think we should be concerned about it if it's threatening our power grid, right? If we're prioritizing AI
43. 10:29training over people having electricity in their house, then yeah, that would be a problem. But, you know,
44. 10:38I don't think it's doing that yet. So, I don't know. I don't know. It's that's above my pay grade. Not the training. Every model is
45. 10:46throwing massively more compute at every response. They're thinking. Oh, they're thinking when they don't need to. I agree with
46. 10:55that. It's killing the costbenefit equations. Those rapper apps are dying off due to expense. Now, that's interesting.
47. 11:06Grock is spending $5 of compute on what cost GPT 5. Yeah, but that that's an ego thing. That's Elon Musk wanting to have a
48. 11:17bigger wiener than Sam Alman. I mean, listen, all of these frontier companies are trying to figure out
49. 11:32what what gets these things to to AGI and ASI. So, the fact that they're throwing massive amounts of compute at it right
50. 11:41now just because it seems to be working, I I don't know. I like the where I go with this stuff is we're we're in a we're in the development
51. 11:55era of of this generative AI stuff. If we were 20 years into this and and it were this inefficient, I'd be
52. 12:06worried. But like nothing's nothing stabilized. Nothing is has plateaued yet. They're all trying to figure it out. So, you know,
53. 12:18anyway, I don't know. I just again, I don't know. I don't know that there's clear enough math out there that we can
54. 12:28make informed decisions as consumers for one thing. And then I also think the efficiency is going to keep changing.
55. 12:36just saying that they are screwing themselves for the same reason. Brains don't think deeply on every idea.
56. 12:43Yeah. Yeah, I get it. It's fascinating to me. Well, in theory, in theory, OpenAI's routing model, GPT5, in theory,
57. 12:54it won't think when it doesn't need to. In reality, it seems to be thinking a lot more than it needs to. And if you
58. 13:00look at David Shapiro's recent posts, he's he's all pissed off at chat GPT5 because it's doing all sorts of like
59. 13:07second-guessing of itself and trying to offer up new things and create follow-up questions. It's not just answering the
60. 13:14questions. It's trying to do it's trying to be a bit more of your good buddy. Oh my god. Um, have you ever used local
61. 13:29AI image gen models? Um, I used not I didn't do it locally, but I I did run a Google Collab instance of stable
62. 13:43diffusion with Dream Dream Booth early on in 2022. So, I have I have installed and run local image gen models, but I
63. 13:51did it in a cloud, you know, a cloud instance that had a good GPU in it because I'm on Mac laptops. I don't have um
64. 14:00I don't I don't have like, you know, a 4090 or whatever uh to to be doing that. It's trying to be politic politically
65. 14:09correct smarty pants. Yeah, I know. Again, I I give I I'll give all of these guys grace right now because
66. 14:21um they're trying to figure this stuff out. I mean, quite frankly, like the the the amount of power that each of us individually have, like
67. 14:33my my co-founder of StoryVine, she let her um chat GPT subscription go. So, she's just got free chat GBT and she's
68. 14:43like it's it's good enough for her who her to do all the stuff she needs to do without without paying for it. Like the
69. 14:50amount of power that we have, you know, in our hands today because of what these guys are doing is remarkable. And I
70. 14:58think us holding them to to task or taking them to task for the fact that they're changing personalities
71. 15:06and it's now it's overindexing this way, now it's overindexing that way, now they're using too much thinking, now
72. 15:11they're not using enough thinking. Like, let them figure it out. [Music]
73. 15:26I got a flash goal on Tik Tok [Music] about as fast as they fix their fixed their naming convention. Well, you know
74. 15:41what? You know what it is? I I mean, the naming convention is a pain in the ass, but I think they they knew where they were going was not,
75. 15:48you know, to to a routing model. They didn't anticipate the the [ __ ] up with the GPT5 launch and and deprecating 40.
76. 15:57Um, so so we're back to where we started with there. Um, [Music] what I would say is that I I feel like the the model companies when they do
77. 16:08really [ __ ] something up, they they have been responding, right? when when uh when that one version of GPT40 got
78. 16:16really sick of Fantic, they they shut that thing down within a week. Um it it looks like Elon's pulling back
79. 16:27some of his his tweets about the little anime girl that you can talk to, you can talk dirty with. Wait, who are we supposed to go follow?
80. 16:43Google. Oh, Google AI Studio. Oh, is that is that new? All right, we're supposed to go follow Google at Google AI Studio on X.
81. 16:56All right, let me share this tab instead. search for at Google
82. 17:10AI Studio. Um, what what I don't like about that, Brandon, is that that tells me that the [ __ ] that they're going to release tomorrow is
83. 17:24going to be available in Google AI Studio, but not on Gemini. And if you don't know the difference, Google AI Studio is where the developers
84. 17:33go hang out and where they they build their stuff and access their APIs. So, it's not a bad place. It's just not a
85. 17:40consumer place. The gemini.google.com is the consumer place. Oh, by the way, today I did I did something that I that
86. 17:50I'm okay with. I I wouldn't recommend you do it. Um, but I've got my my son's insurance is switching from he turned 26, so he's off
87. 18:02my company insurance and we've got to get him new stuff and they want things like what are the specific, you know,
88. 18:10diagnosis and treatments and things like that. So the the place where he gets all of his treat treatments, they had 300 no
89. 18:18400 megabytes of medical records in PDFs, like 700 pages and just all these medical records and I put them into
90. 18:28Notebook LM and I said, "Hey, here's the situation. Here's what I need." And it it wrote up this report that was deadass
91. 18:37accurate in, you know, 45 seconds. It was amazing. just amazing. AI Studios come a long way. Yeah, I agree with that. Let's let's let's
92. 18:50actually go there. Let me before I [ __ ] all over them. Um AI Studio.google.com. Share this tab. Autosave is now enabled
93. 19:05by default. Let's go. ++ oops so URL context tool audio to audio
94. 19:21dialogue try Gemini's neural real time dialogue with audio and video inputs native speech generation native image generation
95. 19:36this is I guess where nano banana banana is going to be when they release that. If you haven't heard about Nano Banana
96. 19:43and then here's all the Yeah, this is the interface here is much better than it used to be. They've clearly gotten
97. 19:48some people on the team that know what they're doing. It looks like It looks like Meta after all those hundred
98. 19:53million dollar hires is losing a bunch of people. Looks like they've got a bit of a [ __ ] show over there.
99. 20:01So, I'm not expecting Meta to get any any any good anytime soon. I think Rick Rubin is the hair sponsor of the day.
100. 20:13>> What did I just see? Oh, I just Yes. >> Oh, no. I just wanted to say that um I think that from what I've seen just over
101. 20:21the weekend is that Google is planning to leverage Google AI Studio as its new frontier dropper where it will tease new
102. 20:32models and do a research preview of them. before it releases them to Gemini, which is their more consumer stable friendly product.
103. 20:42>> Yeah, I think that's generally been how they've done it, but it's just it's like not a lot of people know about it. The
104. 20:47the the the challenge I have with Google is they announce stuff as if it's a consumer launch and then it's often not. It's often for
105. 20:59developers only. like I just I just don't feel like they have a good sense of who their who their actual customer
106. 21:05is. Listen, it's gotten a lot better in the past 6 months and this is now a much more usable site even though it's a bit
107. 21:13cockpit of a 747 still. Um it's much better than it was. So yeah, it's not a bad thing. But I I've seen now two tweets uh since
108. 21:22the start of the show that have suggested that we follow uh Google AI Studio on uh on Twitter. Yeah. And it's
109. 21:31probably the announcements. It looks like um OpenAI makes their announcements on Thursdays. Um Apple used to make
110. 21:39theirs on Tuesdays. There was some meth there was some like reason why Apple did their announcements on Tuesdays. I
111. 21:44forget what it was, but it could Tuesday could be tomorrow could be the day for Google. So, we'll see. Um, >> yeah. Um,
112. 21:58I wanted to try something tonight with Suno. Um, and I think it could be fun. So, Oh, wait. Maybe I'll
113. 22:12[Music]
114. 22:28see [Music]
115. 23:06All right, I got an idea. All right, so let's Here's what we're going to do. We're going to go to Suno.
116. 23:27I'm going to go go to create. All right. So, if you're if you uh if you've never been here before, this is sununo.com
117. 23:38and this is a music making uh platform and it makes all sorts of cool music. So, let's just go grab one here. What do we got?
118. 23:57You think you know me, you don't. You can't bone a man. Not cold, not chance. I'm >> nice. I like that.
119. 24:13[Music] >> All right. So, it makes songs. It makes songs. It makes songs.
120. 24:28So about a year year and a half ago on UDIO, which is another similar kind of thing like this, I did some spoken word
121. 24:37stuff and it did pretty good. And so I wanted to see if Sunno could do good with spoken word stuff. So, um,
122. 24:49so actually what I'm going to do, I have on my pile of crap here on my desk a book by one Brandon Tid
123. 25:08called Fantastical Journeys. So, we're just going to pick a random random page here. So, what we're going to do is we're
124. 25:17going to go we're going to start with audio and we're going to record. All right. So, so I got to hang on hang on
125. 25:27people. This is not easy. All right. And this is this is not what we're going to use as the final thing. I just want
126. 25:35to see if it can recognize this a spoken word. In the depths of the ocean, there lived a group of jellyfish who had a special talent.
127. 25:50They could juggle. These jellyfish named Gemma, Jack, and Jill put on a juggling show every evening under the sea.
128. 26:01Okay, so let's hear it. In the depths of the ocean, there lived a group of jellyfish who had a special talent.
129. 26:11They could juggle these jellyfish named Gemma. Right. Good pretentious spoken word. New York City, East Village
130. 26:22Cadence. Right. So, I'm going to save that. Uhoh. Can't fetch the uploaded audio. What? Oh, do I have to re-record this? Damn it.
131. 26:41All right. One night, the ring master, a wise old alien named Orion, announced a special act. The audience gasped as a comet
132. 26:57zoomed by and out came Luna the lunar. All right, save. Please save. Oh, come on.
133. 27:14Let me reload. [Laughter] record.
134. 27:29I said to the man, "When you record my voice, the worst thing you could do is to lose the recording of my voice that
135. 27:40you record because each performance is the most important performance. Don't lose it, loser. Okay. Save.
136. 27:58Come on. You can do it. Okay, it recorded it. So, now we're going to do a cover.
137. 28:12We're going to continue. Okay, so now we've got a recording. So, now we now we need lyrics. Okay. Oh, wait. Let's see. I said to the
138. 28:23man when you were. So, so it says speech. So, it got that this was spoken word. So, what we're going to do, let's go over to chat jpata.
139. 28:34Show this tab instead. We'll do new chat. And I'm gonna say you are an a you are an off Broadway
140. 28:52spoken word artist with mastery of the words you speak.
141. 29:10I want I want you to craft a beat poem
142. 29:27spoken word performance. No stage directions. about the challenges
143. 29:43and to challenges and toils of being an AI influencer who goes live
144. 30:00every night. The audience on the lower east side
145. 30:15should appreciate it. I'm just here to entertain myself, people. All right. Should appreciate it. Okay, let's get ourselves a script here.
146. 30:30thinking. I'm using uh chatbt5 thinking mini uh because I think it's actually a really good writer of of the three
147. 30:37models, the three main new models. That's the one that I like. I turn the camera on like a prayer, palms pressed
148. 30:44against the glass and LED, and the city exhales into my lap. A slow subway of faces scrolling past. All right, good.
149. 30:53This This looks This sounds pretentious and is long. Um, let me just say tighten it up a bit. Tighten.
150. 31:03No, [ __ ] it. Let's do it. Let's do a long one. Copy. All right. Let's go back to Sunno. So, I'm going to completely replace the words
151. 31:19and I'm going to call it um let's call it spoken word performance. with acoustic
152. 31:35nylon guitar bongos and an appreciative audience
153. 31:51as the performance builds. All right. Or be build beards. Builds builds builds. Dang it. Dang it.
154. 32:07Shut up. Yeah. Put the tag spoken word. spoken word performance with acoustic nylon guitar, bongos,
155. 32:23and an appreciative audience as the performance builds. All right. Oh, we almost maxed out our words. We're at 48
156. 32:3857 out of 5,000 words. This is going to be long and obnoxious. [Laughter] Let's see if it's any good. I think it will be.
157. 32:56At least he remembers we're here. Sort of. Sort sort of. What's happening? Source camp. What's shaking? What's
158. 33:02going down? Somebody just followed me on Tik Tok. Thank you for that. Valerie Cox liked the AI learning lab on Twitter. Thank you for that.
159. 33:14Uh, if you if you're on YouTube and you haven't subscribed to the YouTube channel, please do that. I I don't know
160. 33:22what's happening with Tik Tok. Part of what I've heard is it's all going to go away in September. Part of what I've
161. 33:28heard is it's going to be fine. So, I don't know. But um but why we're dual simo casting is just at at the point at
162. 33:39which they just turn that [ __ ] off, you know. Be subscribed over on YouTube. All right, here we go. [Music]
163. 33:57I turn the camera on like a prayer. palms pressed to glass. >> That didn't work at all. [Music] >> I turned the camera on like a
164. 34:11>> All right. So, that didn't work at all. So, let me do let me get rid of spoken word performance. Performance probably screwed it up.
165. 34:19Let's just try that. I was thinking that maybe performance would would make it do something more better. I should probably give this thing a
166. 34:32title, too. Advanced options. Oh, weirdness. Experimental results. [Music]
167. 34:47I turn the camera on like a prayer. Palms pressed to glass and LED. And the city exhales into my lap. A slow subway of faces scrolling past.
168. 35:01Every night I file myself into a thumbnail. The smile that nets, the pause that keeps them watching, the joke
169. 35:08that claps like loose change down a stairwell. I am a human gift in a world that wants me looped. Good evening, I say. And it is both
170. 35:19greeting and invoice. I sell presents in micro payments. >> Kind of cool. Um, I just changed some settings to be more influenced by the
171. 35:29audio. But let's listen to the other one. [Music] I turn the camera on like a prayer.
172. 35:46Palms pressed to glass and LED. And the city exhales into my lap. A slow subway of faces scrolling past. Every night I
173. 35:55file myself into a thumbnail. The smile that nets, the pause that keeps them watching, the joke that claps like loose
174. 36:02change down a stairwell. I am a human gift in a world that wants me looped. Good evening, I say. And it is both greeting and invoice.
175. 36:14I sell presents in >> That's great. Okay, let's see. [Music] I turned the camera on like a prayer.
176. 36:27>> I don't know why it's making me a woman. [Music] >> I turned the camera. >> Weird. Um, let's see. Spoken word.
177. 36:46Let's go. spoken word, male voice, acoustic nylon guitar. We'll just say percussion. And then I'm going to lose the audience
178. 37:01because I just have a feeling that's screwing things up. [Music]
179. 37:23I turn the camera on like a prayer. [Music]
180. 37:40I turn the camera on like a prayer. Weird. Anybody know? Cuz here, so here's Did I do this too high? I said to the man, no.
181. 37:58Let's see why. There we go. Oh, voice gender male. There we go. Weirdness. Put that there. Style influence. Put a
182. 38:12little higher. Audio influence. We'll put it higher. Now we'll do this. Oh, song title. Um, pretentious. Pretentious
183. 38:29influencer.
184. 38:49[Music] I turned the camera on like a prayer. [Music] >> Oh, wait. What? >> I >> Why do I have so many of these things?
185. 39:07Well, we'll try the new ones. [Music] This sounds spoken wordy. [Music]
186. 39:28I turn the camera on like a prayer. Palms pressed to glass and LED. [Music] And the city exhales into my lap. A slow subway of faces.
187. 39:40>> That's better. [Music]
188. 40:03I turn the camera on like a prayer. palms pressed to glass and LED and city exhales into my >> producer AI has unlimited prompting in
189. 40:13their chat hole until September 1st. Oh, cool. Okay, I'm going to head over. I want to I was going to try producer next
190. 40:21cuz I want to see if it can do this. >> A slow subway of faces scrolling past. Every night I follow myself into a
191. 40:28thumbnail. The smile that nets the pause that keeps them watching. The joke that claps like loose change down a
192. 40:34stairwell. I am a human gift in a world that wants me looped. Good evening, I say. And it is both greeting and
193. 40:39invoice. I sell presents in micro payments. Attention claps time. My heart is a push notification. My dreams are
194. 40:46buffered. I'm on stage and in the DMs and on every playlist they promise. >> Yeah, I saw that. I saw that, Brandon.
195. 40:52Um, okay. So, let's go. Let's let's flip over to producer.ai. So, that's it. It did okay, but you it's a little finicky with the controls.
196. 41:01Um, so producer.ai if you don't know it is used to be refusion and I obviously just
197. 41:16Oh no, they just their CSS didn't load. There we go. Okay, now can I record something here? plus audio recording.
198. 41:39I wonder. Okay, retry. I wonder as I wander deep down the path, down the archways,
199. 41:55into the woods, out of the woods, the shrubberies, the shrubberies. So well manicured into the woods. Out of the woods. Now, if you're
200. 42:11watching this and you're like, "You sound like an idiot." Thank you. That is the intended That is the intended reaction.
201. 42:23Um, let's see. Did that sound at all good? What's going on? Uhoh. Did I crash it? I might have crashed it. Damn it. Ah.
202. 42:38Wait. Save black bar.
203. 43:02Oh my god. Exit page. Reload. I'm having uh I'm having memory issues
204. 43:19here. Google Chrome. So apparently Perplexity has um said that they want to buy Google Chrome if it if it if they get forced to sell it
205. 43:33um for $34.5 billion. Not that they have that, but they'll go get that. Um I would love for someone to buy it
206. 43:43just to fix the [ __ ] memory leaks. Okay. Google Chrome leaks memory tabs. Tab hoarders hoarding tabs. Close your tabs.
207. 43:57Close your tabs. Don't have too many tabs open. Memory leaks. It won't work. Nothing will work. You are a bad person
208. 44:07if you have more than six tabs. Tab hoarder. Tab hoarder. Okay, let's play this.
209. 44:22Solid. Solid. I like it. Save. Save. Crash your browser. Crash your browser.
210. 44:40Oh, no longer sharing screen. Yeah, well, it doesn't seem to [ __ ] matter. Uh, [Music] retry. Nothing. Dead. Dead.
211. 44:56Damn it. Exit page. Reload. Oh my god. Well, we'll just do one without recording.
212. 45:11Slam poetry night at the salon. That's exactly what it is. All right. So, we're going to go we're going to go um so wait, I'm gonna
213. 45:20let let's go back to chat GBT for a second. Share this tab. Uh let's do one just as pretentious,
214. 45:36but about tab hoarders who keep too many tabs open on Google
215. 45:50Chrome and crash it out and crash it out due to memory leaks.
216. 46:04This should be pro- tab hoarder. Period. A lament on Google's inability to manage memory
217. 46:26like too many humans. Yes. So, we're going to do slam poetry on uh Tab Cathedral. I bow to the altar
218. 46:40of a thousand favicons. Each tiny stained glass promise. Each tab a relic, an argument, an unfinished prayer. We do
219. 46:48not close. We collect. This is good. Okay, great. So, going back to producer, show this tab instead. So, now we're
220. 47:03going to say, um, I want you to create a slam poet slam poet performance
221. 47:22performance with spoken word. a male voice and
222. 47:38a poem called the tab hoarders lament. Here are the lyrics.
223. 47:56I wonder if producer can do spoken word. We're about to find out. Okay, if you don't know Producer, this is the only of the music generating
224. 48:12tools that allows you to do this in an ongoing conversation. So, that I'm pretty excited about. All right, let's see what it says here.
225. 48:25Enjoy free text and voice chat all week. Songs in other generations still use credits. Okay. Love those lyrics. Sharp, relatable, and
226. 48:36ripe for Sam slam poetry performance. Here's how I'd frame the tab hoarders lament as a spoken word piece with a
227. 48:43male voice over leaning heavy on pulse and space, minimal backbeat, and a bit of atmospheric instrumental.
228. 48:52spoken prompt, spoken words, slam poetry performance, male cinemat. Okay. Yeah, this is good. There's your lyrics. Ready
229. 49:00to continue into creation. Um, yes, this is good. and also include um audience
230. 49:17reaction, laughter, laughter and spontaneous applause as appropriate to the performance.
231. 49:38Once you have that added, just make the track. Okay. Uh text on Tik Tok. You need You need to be wearing a
232. 49:54cashmere gray turtleneck. Exactly. I do. Or a pink bow. Yeah. All right, let's Oh, it's making this.
233. 50:08It's making our thing right now. All right, let's see what we get here, people.
234. 50:28We do not close, we collect. We are archaeologists of distraction curators of >> Okay. Um, I want it to sound like a guy from Brooklyn,
235. 50:46not London.
236. 51:03[Music] subtle Brooklyn tone.
237. 51:31I'm going to flip back over to Suno quick and go. I bow to the altar of a thousand favicons. Each a tiny stained glass
238. 51:47promise. Each tab relic, an argument, an unfinished prayer. We do not close. we collect. We are archaeologists of
239. 51:53distraction, curators of half begun obsessions, keepers of breadcrumbed identities. They call us. >> Okay, so that's bad, but we can fix it
240. 52:03because we have we have chatability. So I'm going to go okay the words are okay but there is no acoustic
241. 52:18guitar or percussion or audience. It is all dry and it sucks. I want this to feel
242. 52:37like I am on the Lower East Side, New York City, 1989. Go.
243. 52:53[Laughter] If you if you take your prompting too seriously, lighten up and have some fun with it. Tell your tell your LLM it's a dumb dumb.
244. 53:12If you if if you're not putting down your LLM, you're not doing it right. [Laughter] >> All right. But you got to be careful because when
245. 53:25they get close, we collect We are archaeologists of distraction. Curators of half begun obsessions. Keepers of breadcrumbmed identities.
246. 53:35They call us hoarders like it's a crime. I call us archavist of the incandescent now. Still no background music or
247. 53:50club sounds. Rewrite the prompt. Here's a revised prompt.
248. 54:04Slam poetry, spoken voice, subtle Brooklyn accent, smoky nightclub, loud audience chatter, glasses clinking, laughter,
249. 54:13fingerstyle acoustic guitar, jazzy upright bass, run new track. All right, let's see if it can do it. So, so one of the reasons that I like
250. 54:26doing exercises like this where I have no clue if it can even do it is to figure out a can it do it and b what do
251. 54:34you have to do to get it there? Um, and this kind of problem >> bow to the altar of a thousand favicons. It still didn't
252. 54:47>> a tiny stained glass promise. >> Each tab a relic, an argument, an unfinished prayer. >> We do not close, we collect.
253. 54:58>> We are archaeologists of distraction. >> Bad >> curators of half begun obsessions, keepers of breadcrumbed identities.
254. 55:06Numerous killing that small. Under our little empires, it's meous becomes less accurate. There is an ethics to accumulation. Not greedy, not
255. 55:16manic, but rather tabs or trust jars open wide. Collecting the small currency of pension. [Music] >> Wow. This is It's like it it can't
256. 55:30articulate the spoken word with the music. It's It's [ __ ] up all these words. >> All right, let me You know what I want
257. 55:40to do? I like this. I like this uh I like this prompt. Let's take this prompt back to Suno and let's create a new
258. 55:53a new song or a new a new track. So I'll put the lyrics there style real club throughout.
259. 56:08So we're going to go Oh, wait. So that's lyrics. So lyrics we're going to get from chat GPT. Put those here.
260. 56:28Advanced options. Weirdness. Style influence. Song title tab hoarders lament.
261. 56:50All right. Slam poetry performance. Spoken word. I'm going to start with spoken word because performance seemed to screw it up.
262. 57:03All right, here we go. Tab hoarders lament insuno. I was actually surprised that Refusion did bad on that. What that tells me is
263. 57:14that Refusion was trained more on music than this kind of stuff. And it seems that Sunno's got a broader a more broad pallet to draw from.
264. 57:27[Music] I bow to the altar of a thousand's favicons. Each a tiny stained glass promise. Each
265. 57:44tab a relic and argument and unfinished prayer. We do not close. We collect. >> Much better. >> We are archaeologists of distraction.
266. 57:52Curators of half begun obsessions. Keepers of breadcrumbmed identities. They call us hoarders like it's a crime.
267. 58:03I call us archavist of the incandescent. Now every tab is a shrine to a thought that refused to die. A headphone cable
268. 58:11tangled with yesterday's revolution. A recipe we swore we'd master. A thread we swore we'd answer. >> Interesting.
269. 58:22Interesting. It doesn't know how to say archivist. So, we'll misspell it. Archivist. All right, let's hear the next one.
270. 58:34>> Like this one. It's like slam poetry. Yeah, that's what it that's trying to get. >> I bow to the altar of a thousand favicons. [Music]
271. 58:54Good mongos. >> I bow to the altar of a thousands favors. Each a tiny stained glass promise. Each tab relic, an argument, an unfinished
272. 59:09prayer. We do not close, we collect. We are archaeologists of distraction, curators of half begun obsessions,
273. 59:20keepers of breadcrumbmed identities. They call us hoarders like it's a crime. I call us archeists of the incandescent.
274. 59:28Now every tab is a shrine to a thought that refused to die. It's good. [Music]
275. 59:58A bow to the altar of a thousand favicons. Yeah, I don't know why it's doing that weirdass uh like grally voice, but let's let's
276. 1:00:06change some things here. We'll take weirdness down. Put style influence up. Go to more options, song title, word
277. 1:00:15space. I don't know what that is. Um, all right. Let's rewrite the prompt here because I think it's [ __ ] it up
278. 1:00:24somehow. Spoken word slam poetry performance. Male voice. Let's just do male voice. Smoky Lower East Side Nightclub. We're just going to go.
279. 1:00:40We'll just go New York City nightclub ambiance.
280. 1:00:59We'll do audience chatter, clinking glasses, laughter, applause at big moments, period, live finger style, acoustic guitar,
281. 1:01:10jazzy upright bass. I'm just going to say upright bass. Conga shakers, rhythmic background groove, heavy reverb. Let's lose that.
282. 1:01:21Streetwise energy. Let's lose that. So, I assume Streetwise Energy gave us a guy that sounded kind of black and kind
283. 1:01:33of grally, I guess. Words cut through the analog haze. We can lose that. Raw Bohemian. Let's get rid of raw.
284. 1:01:44Bohemian's fine. Unpolished. Bustling with life. Real club ambiance throughout. Okay, let's try this.
285. 1:02:01Oh, I can probably add audio. Hopefully, it doesn't crash. Well, let's let's see what happened without adding
286. 1:02:11audio. I bow to the altar of a thousand favicons. Each a tiny stained glass promise. Each tab a relic, an argument, an unfinished
287. 1:02:21prayer. We do not close, we collect. We are archaeologists of distraction. [Applause] [Music]
288. 1:02:40I bow to the altar of a thousand favacons. [Music] Each a tiny stained glass promise. Each tab a relic, an argument, an unfinished
289. 1:02:48prayer. We do not close. We collect. We are archaeologists of distraction, curators of half begun obsessions, keepers of breadcrumbmed
290. 1:02:59identities. They call us hoarders like it's a crime. I call us archavists of the incandescent. Now every tab is a shrine
291. 1:03:11to a thought that refused to die. A headphone cable tangled with yesterday's revolution. A recipe we swore we'd
292. 1:03:18master. A thread we swore we'd answer. We keep the world in micro windows like sailors with sex mapping storms of curiosity. [Laughter]
293. 1:03:30Chrome hates us. Chrome groans under our little empires. It's memory like an anxious brain on the subway. Leaking, forgetting,
294. 1:03:43dropping passengers on the line. Memory leak they say as if sin were a bug fixable with a patch. Google with
295. 1:03:51its polished teeth and corporate himnels cannot fathom the sanctity of our open questions. It trims it prunes. It
296. 1:03:58gaslights our tabs into sleeping bags. But we are stubborn, gentle, glorious, unruly. Let the browsers call it inefficiency.
297. 1:04:09We call it fidelity. We are a million tiny vows. I will return. I will return. and I will return. [Music]
298. 1:04:22Each tab is a bookmark for an argument with the future. We keep versions of ourselves and duplicate the person we
299. 1:04:29were at 2 a.m. The person who clicked that headline in a cafe. The person who loved a sentence. Close one and you risk
300. 1:04:36killing that small particular grief. Close one and the map of you becomes less accurate. There is an ethics to accumul.
301. 1:04:50>> All right. It's a little loud, but I like it a lot. It's a little aggressive, but I like it. So, we're going to give that one a We're
302. 1:04:59going to give that one a hearty heart thumbs up so we can come back to that one. Um, let me let me blast out a couple of more
303. 1:05:15renders here and then I'm going to record some audio because I have a feeling this could crash it out. Check out Gil Scott Heron
304. 1:05:25though. Okay, cool. Interesting how it's it's a million tiny valves. Yeah. Okay. Record.
305. 1:05:42A thousand tiny valves. Vows and tabs and tabs and valves. Chrome guarding our lives as if they give a damn. They don't.
306. 1:05:56A thousand tiny valves. Vows and tabs and tabs and vows. Chrome guarding our live. Okay. Save. Don't crash. Please don't crash.
307. 1:06:11Crashed. I'll try one more. I don't think it's going to work.
308. 1:06:25I don't think it's going to work. I record. I record. I hit upload. It says no. It says no. Can't fetch the file.
309. 1:06:34Can't fetch the file. Where did you put it? Loser. User. User error. Loser error. All right.
310. 1:06:53Did we already mention Meta is partnering with Midjourney? He did not. Success. Okay, that uploaded. So, we're
311. 1:07:00going to do cover continue. Did it lose Did I lose all of my things? Yes, I lost my lyrics, but that's okay. We'll just put them back. Oops.
312. 1:07:13Let me go get them. Copy. I like these lyrics. I think they're good. Paste. And then let's go down. I think we have new options now.
313. 1:07:25Advanced options, new exclude styles, audio influence, we'll put that a little higher. We'll bring weirdness up a little bit. Song title
314. 1:07:41tab hoarders unite. This one's going to be called unite exclamation point. All right. So, what we're going to do here is I I
315. 1:07:53think we're gonna we're going to go make a midjourney we're going to go make a midjourney avatar. Then we're going to go into
316. 1:07:59Hedra and we're going to make a spoken word performance and we're going to we're going to upload it on on Twitter and then Twitter's
317. 1:08:08going to show it to like two people cuz they're [ __ ] like this. [Music]
318. 1:08:23I bow to the altar of a thousand favicons. Each a tiny stained glass promise. Each tab relic an argument and unfinished
319. 1:08:33prayer. We do not close. We collect. We are archaeologists of distraction. Curators of half-bun obsessions. Keepers
320. 1:08:42of breadcrumbmed identities. They call us hoarders like it's a crime. I call us archivists of the incandescent
321. 1:08:50now. Every tab is a shrine to a thought that refused to die. A headphone cable tangled with yesterday's revolution.
322. 1:08:58A recipe we swore we'd master. A thread we swore we'd answer. We keep the world in micro windows. Like sailors with sex,
323. 1:09:07mapping storms of curiosity. Chrome hates us. Chrome groans under our little empires. It's memory like an anxious
324. 1:09:13brain on the subway. Leaking, forgetting, dropping passengers on the line. memory leak. They say as if sin
325. 1:09:18were a bug fixable with a patch. Google with its polished teeth and corporate himnels cannot fathom the sanctity of
326. 1:09:24our open questions. It trims, it prunes, it gaslights our tabs into sleeping bags. But we are stubborn, gentle, glorious, unruly.
327. 1:09:34Let the browsers call it inefficiency. We call it fidelity. We are a million tiny vows. I will return. I will return.
328. 1:09:42I will return. Each tab is a bookmark for an argument with the future. We keep versions of ourselves and
329. 1:09:49duplicate. The person we were at 2 a.m. The person who clicked that headline in a cafe. The person who loved a sentence.
330. 1:09:55Close one and you risk killing that. >> This is all your fault. >> You never >> What? What the [ __ ] are those? Oh, similar.
331. 1:10:06Oh, cuz I'm in here. I get it. Back here.
332. 1:10:21Wait, where's about to the altar of a thousand favicons, each a tiny stained glass? fathom the
333. 1:10:36sanctity of our open questions as keepers of breadcrumb thought that refused to die. Chrome hates us world and micro.
334. 1:10:43>> That's pretty cool. All right, let's try this again. Tic tac pen. Xai is coming up with an
335. 1:10:59operating system that is purely AI. Oh, macro hard. Yeah, so that's thanks for that reminder, Joker.
336. 1:11:05So Elon Musk, you know, trying to be the master troller that he is, is going after Microsoft with a new company
337. 1:11:13called Macro hard, which is, you know, a division or it's probably right now a joke company spin-off or spin up inside XAI. But the
338. 1:11:26basic idea is he thinks that he can fully replicate Microsoft by basically turning all of its workers
339. 1:11:35including its executive management into AI bots. So So Macroheart is going to be a fully AI company that tries to replicate
340. 1:11:46everything that Microsoft is doing, which listen I I mean, first of all, like the guy's got balls. like he he
341. 1:11:54just he he just trolls consistently and relentlessly, right? He just just all he does. Um and he's smart. I mean, if you
342. 1:12:05look at OpenAI's um levels of AGI, the fifth level of AGI was organization, right? You had you had chat bots and
343. 1:12:14then you had uh reasoning and then you had agents and then you had innovators. So, we're somewhere right now between
344. 1:12:22agents and innovators. And then the fifth level was organizations. And an organization would be a collection of agents that go off and
345. 1:12:31innovate and and do your bidding. And so, it's a relatively straightforward process to map all of the functions within a company
346. 1:12:43like Microsoft and assign AI to them and then and then optimize for that. So, it's going to be fascinating to see.
347. 1:12:51Okay, go to Higsfield AI on X. Okay, I will. But we got to taborters unite. We've got to get We got to get the right sound here, kids.
348. 1:13:00[Applause] [Music] I like the I like the applause to start here. >> I bow to the altar of a thousand favors.
349. 1:13:10>> Each a tiny stained glass promise. >> It went back to that weird [Music] I bow to the altar of a thousand
350. 1:13:22favicons. Each a tiny stained glass promise. Each tab relic and argument and unfinished prayer. We do not close, we collect.
351. 1:13:32We are archaeologists of distraction, curators of half-bun obsessions, keepers of breadcrumbmed identities.
352. 1:13:39They call us hoarders like it's a crime. I call us archivists of the incandescent now. are chests. I bow to the altar of a thousand
353. 1:13:49favicons. No, >> I bow to the altar of a thousand favicons. Each a tiny stained glass promise. Each tab a relic, an argument,
354. 1:13:57an unfinished prayer we do not close. We collect. We are archaeologists of distraction. Corders like it's a
355. 1:14:04headphone, a thread we swirl. As chrome groans under say as if. >> All right, that one has no. So, let's do style higher.
356. 1:14:13This is driving me crazy. male um spoken word slam poetry.
357. 1:14:52Remix remix. All right, let's go to Higsfield AI on X at Higsfield.
358. 1:15:09Higsfield's an interesting company. They seem to have good technology, but they market they they market the snot out of
359. 1:15:16it, and it's kind of annoying. Ah, introducing Higsfield Records. Interesting. The first AI record label.
360. 1:15:23I I called this six months ago. I'm telling you, every one of these content generation companies is probably going to become a studio
361. 1:15:33or or a label. Multi-million dollar contracts already in motion. You can become the next global idol with us. It's easy. No
362. 1:15:41talent required. We reach retweet to get the details in your DM. Uh just adding in here, if you look at
363. 1:15:52the first like comment underneath it, uh there's a form that you can fill out to be on the wait list, but they do ask for
364. 1:15:59your Twitter handle, so they're probably looking for you to retweet it. >> Yeah. Well, they said retweet it.
365. 1:16:05>> No, but say it's not just the retweet. There's also a form that you fill out. >> Got it. Okay. All right. Cool. Um but let's retweet it
366. 1:16:16anyway. quote. Um, I called this six months back um, media
367. 1:16:31generation platforms will be getting into the content game.
368. 1:16:46The opportunity opportunity to get let's see to be an artist of the future
369. 1:17:04is now. Let the haters hate. This is the new game. Rock on.
370. 1:17:21Tag Scoilizer in there. Post. All right, let's go back to Suno. See if we got a
371. 1:17:35See if we got some action here. We need something that doesn't suck. This sounds good. A bow to the altar of a thousand
372. 1:17:51favicons. Each a tiny stained glass promise. Each tabar relic an argument and unfinished prayer. We do not close, we collect.
373. 1:18:03We are archaeologists of distraction, curators of half begun obsessions, keepers of breadcrumbmed identities.
374. 1:18:10They call us hoarders like it's a crime. I call us archivists of the incandescent. Now every tab is a shrine
375. 1:18:17to a thought that refused to die. [Music]
376. 1:18:33Great question. Who who's asking this? Wolf man Clint. I don't know why or get it pronouncing pronouncing of that word
377. 1:18:40is it not in its vocabulary. So in its vocabulary is a really interesting choice there in in pre-ai systems you would actually
378. 1:18:51have pronunciation for all the words. You would have a vocabulary and you would have pronunciations of it.
379. 1:18:57Generative AI is not a database. It's not it's not um uh you know logic based computing. It's it's making these words
380. 1:19:07in real time. So, um it it it obviously doesn't have enough training around certain words that it that it's just
381. 1:19:16mucking them up when it when it's trying to to generate that sound. It's just not doing a good job of it. But it's a really good question.
382. 1:19:26>> About to the altar of a thousand. >> No, [Music] baby. [Music]
383. 1:19:42I bow to the altar of a thousand favicons. Each a tiny stained glass promise. Each tab relic an argument and unfinished
384. 1:19:50prayer. We do not close. We collect. We are archaeologists of distraction. Curators of half begun obsessions.
385. 1:19:58Keepers of breadcrumbmed identities. They call us hoarders like it's a crime. I call us archist. Archists. [Music]
386. 1:20:16I spelled it. No, I spelled it with a K. I think it it even has a K in it now. Oh, no. I I replaced it. Hang on. Um,
387. 1:20:26I'll do it with a K. R. I'll I'll do CK. R CK. Archivists. We'll see if it does it now. Remix.
388. 1:20:48I like this. Yeah, that's how it's spelled. Ryan. >> Yeah, it's going to keep mispronouncing it a different way.
389. 1:20:57>> To the altar of a thousand favor guns. Each a tiny stained glass promise. Each tab have a relic, an argument, an
390. 1:21:03unfinished prayer. We do not close. [Music]
391. 1:21:20I bow to the altar of a thousand's favicons. Each a tiny stained glass promise, each tab a relic, an argument, an unfinished
392. 1:21:28prayer. We do not close, we collect. We are archaeologists of distraction. Curators of half begun obsessions,
393. 1:21:36keepers of breadcrumbmed identities. They call us hoarders like it's a crime. I call us archivists of the incandescent. Now,
394. 1:21:49>> damn it. Archivists. Let's go to the last two. Here's where we changed the spelling of it. [Music]
395. 1:22:17A bow to the altar of a thousand favicons. Each a tiny stained glass promise. Each tab a relic and argument
396. 1:22:23and unfinished prayer we do not close. We collect. We are archaeologists of distraction, creators of half-bun obsessions, keepers
397. 1:22:33of breadcrumbmed identities. [Music] Oh, I know what I'm going to do. I call us keepers. I call us keepers of the incandescent.
398. 1:22:46Now, let's see. I call us I call us um dark keepers of the incandescent now remix. That's good. That's better. Dark keepers
399. 1:23:02archivists or archivists. Guys, the reason it's it's pronouncing the words different each time is the way
400. 1:23:10diffusion models work is they're they're combining probably thousands of different kind of random samples
401. 1:23:21sample tokens. Like each token is like this little snippet of like, you know, a thousand layers deep. And it's so
402. 1:23:29depending on how it weights those things, that's why you're getting different things each time. It's It's
403. 1:23:35not nearly as predictive as we'd like it to be. Sea slug. Yeah, the I saw the one with the eyeball in it, but I just
404. 1:23:40rewrote the line al together now. >> Like the harmonic >> to the altar of a thousand favicons.
405. 1:23:55Each a tiny stained glass promise. Each tab a relic, an argument, an unfinished prayer. We do not close, we collect.
406. 1:24:04We are archaeologists of distraction, curators of half-bun obsessions, keepers of breadcrumbmed identities.
407. 1:24:14They call us hoarders like it's a crime. I call us dark keepers of the incandescent. Now every tab is a shrine
408. 1:24:22to a thought that refused to die. A headphone cable tank. >> Wait, what's that? They are digital twinning the application. ask me for my
409. 1:24:33height and weight. Oh, they want to make you the superstar. So, should I go to my email? Oh, no. I should go to my DMs, right?
410. 1:24:43I don't have a DM from them. Doesn't look like >> No, the it was on the type form that I uh in the next um >> Oh, in the next thing
411. 1:24:52>> down, but yeah, it's really interesting. if you um once I finished the application, it said um if you retweet,
412. 1:24:59fill out the application, we're going to make you the superstar. So like this person, this chai person is apparently
413. 1:25:08>> somebody that a likeness that they've trained off of somebody a real human and they've made them into an AI superstar.
414. 1:25:16So Kyle, that could be you floating in that murky water there. I I I think if anyone is to be an AI superstar, I think I think it's me.
415. 1:25:27Here comes the twerking. Finally. I just thought that was unique because it it's taking some real grounded stuff
416. 1:25:40and converting them into a AI persona. >> Uh what's your ex? So at >> Shannon >> tab tab.
417. 1:25:56>> Oh, am I not tabbed? Oh. Oh, because it opens a new tab. Your other social media. Tick tock.comarninglab. [Music]
418. 1:26:10Uh, what else? That's good. Okay. Age. I'm [ __ ] old. I'm old, man. Share your physical data. Um,
419. 1:26:285 foot8 in 200. I'm a fat [ __ ] How did you discover Twitter? Thank you for your application. All
420. 1:26:43right, cool. There you go. Beautiful. All right, back to Sunno. Share this tab instead. >> So, that one's pretty good. Let's hear this one.
421. 1:26:56[Music] Like this. [Music] About to the altar of a thousand
422. 1:27:10favicons. Each a tiny stained glass promise. Each tab a relic, an argument, an unfinished prayer. We do not close,
423. 1:27:17we collect. We are archaeologists of distraction, curators of half-bun obsessions, keepers of breadcrumbmed identities.
424. 1:27:27They call us hoarders like it's a crime. I call us dark keepers of the incandescent now. Every tab is a shrine
425. 1:27:33to a thought that refused to die. A headphone cable tangled with yesterday's revolution. A recipe we swore we'd
426. 1:27:39master. A thread we swore we'd answer. We keep the world in micro windows like sailors with sextants mapping storms of curiosity.
427. 1:27:49Chrome hates us. Chrome groans under our little empires. Its memory like an anxious brain on the subway leaking
428. 1:27:55forgetting dropping passengers on the line. Memory leak. They say as if sin were a bug fixable with a patch. Google
429. 1:28:01with its polished teeth and corporate himnels. Cannot fathom the sanctity of our open questions. It trims. It prunes.
430. 1:28:07It gaslights our tabs into sleeping bags. But we are stubborn, gentle, glorious, unruly. Let the browsers call it in efficiency.
431. 1:28:16We call it fidelity. We are a million tiny vows. I will return. I will return. I will return. [Music] Each tab is a bookmark for an argument
432. 1:28:26with the future. We keep versions of ourselves in duplicate. The person we were at 2 a.m. The person who clicked
433. 1:28:31that headline in a cafe. The person who loved a sentence. Close one and you risk killing that small particular grief.
434. 1:28:37close one in the map of you becomes less accurate. [Music] >> There is an ethics to accumulation. Not greedy, not manic, but reverent.
435. 1:28:47>> There's an ethics to tab hoarding. Not greedy, not manic, but reverent. Tabs are trust jars. We build the
436. 1:28:56altars. Meanwhile, the engineers whisper in fluorescent rooms, wielding heristics like brooms, saying, "Optimize, compress, discard."
437. 1:29:04They they mistake tidiness for intelligence. They forgot memory is not just storage, it's testimony. A memory
438. 1:29:11leak is an in an app is a moral leak in the machine. That's really good. Um, we refuse so and so we refuse efficiency
439. 1:29:21romantic in our rebellion. Okay, I'm going to end it right there. Um, we relish the moment. So, let's make this one its own stanza.
440. 1:29:32For every crash there's revolution. If Google cannot hold us, that failing is ours. Raise a tab then another. Let's get rid of that.
441. 1:29:56We will not let optimization rewrite the archive of our desire. That's a good That's a good ending line. Okay. So, let's
442. 1:30:09let's make that a new stanza. Let's make that a new stanza.
443. 1:30:23Okay. Remix. So, this is going to be shorter now. So, it depends on the genre. It It
444. 1:30:38matches the the words to the lyrics quite well in a lot of cases. Um, I'm purposefully doing spoken word here. So,
445. 1:30:46spoken word is is generally not very lyric friendly. I mean, very like it doesn't match the the percussion behind it. [Music]
446. 1:31:03Wait. I I bow to the altar of a thousand favicons. That's kind of a cool starting. Each tiny. We do not close. We
447. 1:31:13collect. We are archivists of disgra curators who have begun obsessions. Oh, we got keepers in here twice. Keepers. Um, let's see. Um,
448. 1:31:38we hoard the tabs that you Oh, wait. We
449. 1:31:55preserve the tabs that you discard. They call us hoarders like it's they call us tab hoarders like it's a crime.
450. 1:32:35Keepers of the incandescent. Now every tab is a shrine. Okay, remix. All right, we got some new lyrics here. People,
451. 1:32:41calm down, everybody. Calm down. Gonna run it again. Yeah, I'm going to run it again. Keep running it. Thank
452. 1:33:03[Music]
453. 1:33:20you. I bow to the altar of a thousand favicons. Each a tiny stained glass promise. Each tab a relic and argument
454. 1:33:30and unfinished prayer. We do not close. We collect. We are archaeologists of distraction. >> Damn it. It's close but not there.
455. 1:33:38>> Bow to the altar of a thousand favors. Damn.
456. 1:34:03[Music] That sounds like Ani DeFranco. [Music] I bow to the altar of a thousand
457. 1:34:17favicans. Each a tiny stained glass promise. Each tabar relic and argument an unfinished prayer. We do not close. We collect.
458. 1:34:28We are archaeologists of distraction, creators of half begun obsessions, keepers of bread. >> I like that one a lot actually.
459. 1:34:36Give that a thumbs up. Um, I like that. It's funky. I bow to the altar of a thousand
460. 1:34:50favicons. Each a tiny stained glass promise. Each tab a relic, an argument, an unfinished prayer. We do not close,
461. 1:34:57we collect. We are archaeologists of distraction, curators of half begun obsessions. We preserve the tabs that you discard.
462. 1:35:08They call us tab hoarders like it's a crime. I call us inspired keepers of the incandescent now. Every tab is a shrine
463. 1:35:14to a thought that refused to die. A headphone cable tangled with yesterday's revolution. A recipe we swore we'd
464. 1:35:21master. A thread we swore we'd answer. We keep the world in micro windows like sailors with sextants mapping storms of
465. 1:35:27curiosity. Chrome hates us. Chrome groans under our little empires. It's memory like an anxious brain on the
466. 1:35:33subway. Leaking, forgetting, dropping passengers on the line. Memory leak, they say as if sin were a bug fixable with a patch. Google with
467. 1:35:42its polished teeth and corporate himnels cannot fathom the sanctity of our open questions. It trims, it prunes, it
468. 1:35:48gaslights our tabs into sleeping bags. But we are stubborn, gentle, glorious, unruly. [Music] Let the browsers call it inefficiency.
469. 1:36:00We call it fidelity. We are a million tiny vows. I will return. I will return. I will return. Each tab is a bookmark
470. 1:36:07for an argument with the future. We keep versions of ourselves and duplicate. The person we were at 2 a.m. The person who
471. 1:36:13clicked that headline in a cafe. The person who loved a sentence. Close one and you risk killing that small
472. 1:36:19particular grief. Close one and the map of you becomes less accurate. There is an ethics to tab hoarding. Not
473. 1:36:26greedy, not manic, but reverent. [Music] Tabs are trust jars. Open wide. >> This one's close. [Music]
474. 1:36:46I bow to the altar of a thousand favic kinds. Each a tiny stained glass promise. Each tabar relic an argument.
475. 1:36:54An unfinished prayer we do not close. We collect. We are archaeologists of distraction. Curators of half begun obsessions. We
476. 1:37:03preserve the tabs that you discard. They call us tab hoarders like it's a crime. I call us inspired keepers of the incandescent. Now
477. 1:37:15every tab is a shrine to a thought that refused to die. A headphone cable tangled with yesterday's revolution. A
478. 1:37:22recipe we swore we'd master. A thread we swore we'd answer. We keep the world in micro windows like sailors with sex
479. 1:37:30mapping storms of curiosity. >> All right, that's the one. All right, so now we're going to go. And I know it's
480. 1:37:38late, so if you got to go, I get it. But we're going to go turn this into something [ __ ] cool. Um, midjourney because
481. 1:37:50if I don't do it now, I won't come back to it and then it will be lost. And us tab hoarders got to stick together.
482. 1:38:07Um, [Music] let's see.
483. 1:38:23So, now I'm in mid journey.
484. 1:38:45Ending frame.
485. 1:39:06Do I have to do this in draft? Oops. Get rid of that there.
486. 1:39:22Why is this I'm confused. This happened to me once before. I'm trying to add a an image as a style reference
487. 1:39:33or omni reference as they call it. Oh, I think it Oh, it's because I'm in draft mode. Omni reference. There we go. Okay.
488. 1:39:45So, let's pop him into Omni Reference. And then we're going to go um hip New York City performer
489. 1:40:00in a Let's see. How do we want this? We want this vertical 2x3, 3x4, 3x4.
490. 1:40:152 by3 3x4 Okay, in a moody New York City
491. 1:40:30club in a spotlight sitting on a stool.
492. 1:40:49looking at the camera and then we're going to go we'll go weirdness at I don't know 500 stylization at 400 and then we'll go
493. 1:41:04dash dash R for repeat space 4 and that's going to give us a whole [ __ ] pile of these images to play with or to pickank
494. 1:41:21dank. All right. Thoughts, questions? Anybody have any questions? I just finished four tracks, lyrics, and music today. Every
495. 1:41:30vocal key. Wait, every vocal key, music key, etc. Cool. Needs more cowbell. Absolutely. Oh, these are cool.
496. 1:41:48would love to be able to see him. We're still looking at some of the lyrics. >> Damn it. Sorry about that. Well,
497. 1:42:06amic pentaber. Pentaber. I like it. That That one's actually pretty cool. That looks really realistic.
498. 1:42:32I think that first one's the one. Yeah, let's let's we're going to upscale this. Let's see. Upscale subtle upscale
499. 1:42:47creative [Music] dunk.
500. 1:43:03Okay. Then we're going to go I'm going to go over here and I'm going to go to Hedra. HD R A.AI. I'm going to log in.
501. 1:43:22So, we got to go back to Sunno. Is that thing done? How's that looking? Whoops.
502. 1:43:39That looks fake. But this is one thing I've noticed in MidJourney is sometimes when you upscale
503. 1:43:54things, it it gives it more of that plastic look. This one's not bad. He's far enough away. I think I think this could be okay.
504. 1:44:10Okay, we're going to download him and then we're going to go back to Suno. We're going to download this one. We'll download this as an MP3.
505. 1:44:25And then we're going to also say generate video because we'll just upload the video on its own. And now we're going to go to Hedra
506. 1:44:38chat with a Hedra realtime avatar now. Oh, that's cool. All right, let's see. Image video character three. Okay.
507. 1:44:54All right. Are you all seeing this? No, because I didn't switch tabs. Now you are. So now I'm at Hedra. So I'm gonna upload an audio script,
508. 1:45:06which is our little song we just made or a little [Music]
509. 1:45:21I bow to the altar of a thousand favorite kinds. Each a tiny stained glass promise. Each tab relic. [Music]
510. 1:45:41There's our dude. I think we'll just go. Bang. Error. Failed to fetch. Try it again.
511. 1:46:03That failed to fetch thing is probably a memory leak error, which is what this spoken word poetry is all about.
512. 1:46:14It's Dave Stewart. It does look like Dave Stewart. tink.
513. 1:46:31A memory leak error is when you have too many tabs open and the application can't manage the memory from all of the different tabs
514. 1:46:43and it just basically runs out of memory and then craps. [Laughter] I like picking on Google. I've been
515. 1:46:56pissed off at them ever since they got rid of their let's do no evil as one of their corporate headlines. Fuckersink.
516. 1:47:15All right, that's processing. So, I don't know what what we can do now. We can just go chat now. All right.
517. 1:47:23So, I don't know. What were the uh what were the takeaways from tonight? Go try to make [ __ ] happen that you
518. 1:47:33think can't happen is not a bad exercise. Um, AI can't pronounce things. But it's also not that it hasn't been
519. 1:47:46trained on that word. It hasn't been trained on any word. It's just that it doesn't have enough training data to get
520. 1:47:52that word right, which is a weirdly different thing. Has the inshitification of the internet happened to you, too?
521. 1:47:59Oh, that happened years ago. You don't know how to use the black bar. I don't. Um, it can't match lyrics to music. Uh, it
522. 1:48:11it it's pretty [ __ ] good at it. It can write music. Yeah, it can write music. Have you never played with Sunno?
523. 1:48:20It can write music. It can write lyrics. Here, let's While that's processing, let's go back to Sunno. We'll just go We'll go create
524. 1:48:29uh somewhere at the bottom here. I think I can clear everything, right? Yeah. clear all.
525. 1:48:43But it it matches it matches lyrics to music. It does it pretty good. Let's do rap. Melada. No, we won't do
526. 1:48:50rap because that's more spoken word. We'll do uh we'll do voice cracks. Voice cracks. Soaring. What's that? Russian thing tab.
527. 1:49:10lyrics. Write the lyrics. Auto. Describe your lyrics. Uh song about matching lyrics to the music.
528. 1:49:25We'll get a little meta here. Uh let's see. Voice crack. Uh, let's do um let's do uh funk. Um, funk driving.
529. 1:49:41Uh, electric guitar, voice cracks, creative harmonies,
530. 1:49:56and synopation. And we'll go style influence. And we'll go we'll do weirdless low. We'll do
531. 1:50:15match the lyrics.
532. 1:50:31What is syncupation exactly? Timing with the drums. No, syncopation is when you're when it's you've got beats in between the main beats.
533. 1:50:49[Music] words are dancing on the page tonight.
534. 1:51:08[Music] Tiptoeing around the nose just right. [Music] Do they swing or do they crawl
535. 1:51:22or do they eat? >> Yeah, not bad, but in the neighborhood. >> It had some good syncupation. The all those beats that were off the main
536. 1:51:31beats, that's syncupation. [Music]
537. 1:51:47Yeah, it's like an offbeat. Let's have ChatchBT define it for us. Um, define syncopation simply
538. 1:52:05dunk. Uh simple definition, putting the musical accent where you don't expect it of the week's beats on the weak beats or the ands between
539. 1:52:16beats. A quick clap clap example. One and two and three and four and clap on the ends. The offbeats. One
540. 1:52:26one two three four. Yeah, there you go. That offbeat emphasis equals syncopation. That's cool. [Music]
541. 1:52:46No, not yet.
542. 1:53:01Oh, that's good. All right, let's try that. Uh, let's do 68 time signature.
543. 1:53:35[Music] words like a dark. >> That's pretty cool. Like that beat. All
544. 1:53:50right, here's the 68 ones. [Music] I've got a line that homes in my head. >> No. [Music]
545. 1:54:40Whatever. All right, let's see. Our video is at 60%. All right, let's let's make another song. There's a there's a
546. 1:54:47cool thing we can do on Suno. If you click on the explore tab, you get this giant wheel of of uh style mashups.
547. 1:54:59Africanner, dreamy acoustic rock, acid techno, avancgard jazz, Spanish pop, alt country dubstep, saxophone, basosanova,
548. 1:55:16mandarin drums, Egyptian flamnco, slushwave, new jack swing, [Music]
549. 1:55:48[Music] Let's do that one.
550. 1:56:03All right. Did it pull the style in? Did not. That sucks.
551. 1:56:21[Music] Coral Afro jazz. Coral Afro jazz.
552. 1:56:39Auto lyrics. Um, lyrics about machines being more creative than
553. 1:57:00we ever expected in our lifetimes. Create create.
554. 1:57:18All right, the video is at 80%. We're almost done, people. Shank dank shank dank. [Music]
555. 1:57:46It started with sparks in a silent cold. Oh, [Music]
556. 1:58:16hey. [Music]
557. 1:58:32It started with sparks in a mental brain. A whisper of thought in a coded vein.
558. 1:58:48[Music] Not a soul, not a heart, just wires and steel. But it [Applause] >> was kind of cool. [Music]
559. 1:59:14It started with sparks. works in a metal brain. A whisper of thought in a coated vein.
560. 1:59:30Not a soul, not a heart, just wires and steel. But it painted a world that made us feel
561. 1:59:45>> the machine alive. It dreams in life. Colors we've never seen. Ignite the night.
562. 2:00:00It's more than we imagine, more than we could know. [Music] The creative machine steals the show.
563. 2:00:17It wrote us a story with no end inside. A melody spinning from black and white.
564. 2:00:32Now, that one's pretty cool. That gives a thumbs up. All right. Um, let's go watch our video. Are we done? Are we done? Oh, we're not
565. 2:00:42done. I thought we'd be done by now. Come on. I just want to see a guy in a club doing spoken poetry about tab hoarding.
566. 2:00:55Is that so much to ask? Ah, what do you think, Pate M? We're just screwing around tonight. There's
567. 2:01:10nothing resembling productivity on the uh learning lab tonight. If you learn something tonight, that's
568. 2:01:18on you. I I claim no responsibility for anything resembling education. Um, I have a feeling this is done, but the No, it's not.
569. 2:01:39You know what? Let's go do something. While that's doing that, um, remix, edit, open an editor.
570. 2:01:53Oh, I got to switch tabs. So, I'm in. Click and drag the song lyrics to select them in the timeline.
571. 2:02:09Use new lyrics to enter new lyrics. Okay. It started with sparks. And let's change metal to silicon.
572. 2:02:24And a silicon. Oh, wait. I've got to I've got to add new lyrics. Oh, wait. So, I'm going to copy that. I'm going to put it here.
573. 2:02:40And then I'm going to go silicon. It started with sparks in a silicon brain. A whisper of thought in a coated vein. I
574. 2:02:52like that. Not a soul, not a heart, just wires. Here, let's grab all this. Let's grab all this.
575. 2:03:16in a coded vein. Not a soul, not a heart.
576. 2:03:46Let's see.
577. 2:04:04electronic zeal. Not a soul, not a heart. Electronic zeal, but it painted a word world that made us feel. All right,
578. 2:04:11let's try that. So, we're going to replace those lyrics with those lyrics. Replace
579. 2:04:32And I think what it does is it'll give us two alternates. Our other video is still at 80%. So,
580. 2:04:57I need coasters. I'm getting a lot of sweating on my drinks. All
581. 2:05:19right. with sparks in a silicon brain. A whisper of thought in a coated vein,
582. 2:05:35[Music] not a soul, not a heart. >> Oh, nice. >> Yeah, I kind of missed that. But it painted a world that made us
583. 2:05:58[Music] >> It started with sparks in a silicon brain. A whisper of thought in a cold
584. 2:06:15[Music] It started with sparks in a silic >> like that. A whisper of thought
585. 2:06:31in a coded vein. Not a soul, not a heart. It electron ex. But it painted a world
586. 2:06:45that made us feel alive. [Music] All right. So, that's in there. New new lyrics are in there. This thing is
587. 2:07:00really cool. If you haven't played with this this new um the new editor, you can move these sections around. You can
588. 2:07:07repeat them. You can skip them. All right. So, let's save as a new song. You can also split it into stems. Go to song.
589. 2:07:27The creative machine edit. Come on. Why you not playing? Reload.
590. 2:07:53Why is it not playing? Um, create. [Music]
591. 2:08:17It started with sparks in a silicon brain. A whisper of thought in a coded vein.
592. 2:08:33Not a soul, not a heart. It had trauma, [Music] but it painted a world that made us feel
593. 2:08:48the machines alive. It dreams in light. Colors we've never seen. Ignite the night.
594. 2:09:03It's more than we imagine. more than we could know. The creative machine
595. 2:09:18that that that weird thing where it goes up like that, that's like a pseudo tick. I've noticed that it it'll do these
596. 2:09:26really unexpected jumps. That's something I would go edit. I I would not change the lyrics there, but I would
597. 2:09:34change the the music there. Fascinating. All right, we have a video. Let's go look at our video. Uh download.
598. 2:09:52All right, here we go. We're going to open this here. I'm going to change my sharing to
599. 2:10:07[Music] I hope this doesn't suck. Uhoh. Where'd that go? [Music]
600. 2:10:28I bow to the altar of a thousand favor kinds. Each a tiny stained glass promise. Each tab relic an argument. An
601. 2:10:36unfinished prayer. We do not close. We collect. We are archaeologists of distraction. Curators of half begun obsessions. We
602. 2:10:45preserve the tabs that you discard. They call us tab hoarders like it's a crime. I call us inspired keepers of the incandescent. Now
603. 2:10:56every tab is a shrine to a thought that refused to die. A headphone cable tangled with yesterday's revolution. A
604. 2:11:03recipe we swore we'd master. A thread we swore we'd answer. We keep the world in micro windows like sailors with sex
605. 2:11:11mapping storms of curiosity. Chrome hates us. Drone groans under our little empires. It's memory like an
606. 2:11:20anxious brain on the subway. Leaking, forgetting, dropping passengers on the line. Memory leak, they say as if sin were a
607. 2:11:29bug fixable with a patch. Google with its polished teeth and corporate himnels cannot fathom the sanctity of our open questions. It
608. 2:11:39trims, it prunes, it gaslights our tabs into sleeping bags. But we are stubborn, gentle, glorious, unruly. [Music]
609. 2:11:52Let the browsers call it in efficiency. We call it fidelity. We are a million tiny vows. I will return. I will return.
610. 2:12:01I will return. Each tab is a bookmark. All right, we're going to publish this. I'm such an idiot.
611. 2:12:17[Laughter] All right, here we go. Oh my god, y'all are [ __ ] weird. You're still here. How many people? 16.
612. 2:12:28Yeah, I knew we blew blew a bunch out tonight. All right. People like, "Fuck it. I'm going to bed. He's not going to talk to us." All
613. 2:12:44right. There's that. Um
614. 2:14:14[Laughter] Tabo orders unite. A spoken word plea for Google to fix the goo to fix Google Chrome to be more friendly to those of
615. 2:14:29us who generously retain tabs over days, weeks, and even months. We're not the problem. We're the solution. and apps. What's apps?
616. 2:14:47What's apps? >> Google apps. No, the different apps that you've used like you for your >> CHGPO
617. 2:15:03at chatgpt at OpenAI at Hedra.
618. 2:15:20Um, I think that's it. We'll do scobilizer. See if we can get a little publicity out of this.
619. 2:16:20Your memory leaks are killing us. We're not the problem. We're the solution.
620. 2:16:36All right. So, everybody, as as as I always request, did this upload? No, of course not. Um, well, we'll try to post it. See if it uploads.
621. 2:16:541% 0% 1%. This is not going to upload. Oh my god. Elon, [ __ ] fix your video upload. God damn it.
622. 2:17:17All right, let me go. We're already in QuickTime here, so let me export this as 780p or 720p. Um, let's have hoarders
623. 2:17:32unite error. Close that. We'll get rid of
624. 2:17:47this. We'll re-upload it.
625. 2:18:11Come on. post.
626. 2:18:31The [ __ ] is intended to be broken. Well, it is. And it pisses me off. It's like, you know,
627. 2:18:45you aren't supposed to be able to post what you want. I know. I know. The free speech platform doesn't allow you to post what you want.
628. 2:18:59All right. Well, this will be up there at some point. 8% 13% looks like it's going. It looks like we might have snuck
629. 2:19:09a a socket in there. Oh, back down to 30%. Come on, 31. You can do it. Oh. Oh, no. God damn it.
630. 2:19:30inspo for your next music video. Elon, fix your [ __ ] upload. Hey, you can send a rocket to space and
631. 2:19:39catch it with a pair of chopsticks, but could you fix your [ __ ] upload service on Twitter, please? You're the
632. 2:19:46one claiming this is going to be a video forward site. Um, one of the things that's required for that to be true is to be able to
633. 2:19:55upload video. Elon, anybody wants to trim that little part of the video out and send it to him? I'm fine with that.
634. 2:20:07Someone want to do a little clipping tonight? Good luck uploading it. Oh my god. Let me try uploading it. Let
635. 2:20:20me let me pop it in there again. See if that does something. It's just so [ __ ] random when it works or not. Yeah, it's not going to
636. 2:20:29work. It goes up, then comes down, goes up, goes down, goes up, goes down. Does Does anyone have any technical
637. 2:20:39chops that understands why this is happening? In the past, you've copied and refreshed. Oh, yeah, that's right.
638. 2:20:45Okay, so let me try that. So, I'm going to copy all this. I'm going to refresh the browser. I'm going to paste a new thing in here.
639. 2:21:01And then I'm going to put the video in. It could be some local Actually, you know what it could be? It could be local
640. 2:21:13memory issues with Chrome. Yeah. See, it didn't go. It didn't go at all that time
641. 2:21:27post. All right, listen. I will do this afterwards. I don't know what's going on, but it's driving me [ __ ] crazy.
642. 2:21:43Let me try let me try the other format one more time. We'll try one more time. Downloads this 0%.
643. 2:22:09Yeah. Post. All right. It's not going to go. All right, people. Uh, hope you had fun tonight. Tomorrow night's regular time.
644. 2:22:18Uh, there's no salon tomorrow night. Um, we'll see if this week it looks like we're going to get some new stuff from Google that is
645. 2:22:31worth playing with it. It looks like this nano banana um image model. Um people are saying it is actually the next big thing for
646. 2:22:42images anyway. Um so we'll see. We'll play with that. And then it looks like they're going to launch something else
647. 2:22:47whether it's Gemini 3 or some new fancy version of 2.5 2.6 something like that. It'll be interesting to see. Um
648. 2:22:57but yeah, I'll see you tomorrow. All right. Peace out everybody.