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

Description

Chapters

Transcript

0:00 Woo!
0:02 [Music]
0:16 [Music]
0:24 [Music]
0:29 Woo! Oh,
0:31 [Applause]
0:32 [Music]
0:40 [Music]
0:57 yeah.
1:00 [Music]
1:04 [Applause]
1:06 [Music]
1:15 [Applause]
1:16 [Music]
1:18 [Applause]
1:22 Well, I heard there was a zigger
1:28 David played an alarm.
1:32 You don't really care for music, do you?
1:39 With go like this. The fourth, the fifth
1:43 mult
1:45 [Applause]
1:47 king composing.
1:49 Hallelujah.
1:51 [Music]
1:53 Hallelujah.
1:57 Hallelujah.
2:00 Hallelujah.
2:04 Hallelujah.
2:06 [Music]
2:15 Good evening, good people. Happy Monday.
2:18 We are having trouble streaming to
2:20 LinkedIn. This may be an issue with
2:22 LinkedIn. We'll keep trying. Huh. Well,
2:25 look at that.
2:28 Oops.
2:30 Oops. I'm in the wrong live. Hallelujah.
2:37 What are our plans tonight? Um, I have
2:40 an idea for something fun to do with
2:44 Sunno that I experimented with in the
2:46 early days of these music things and I
2:48 haven't tried in a while. Um,
2:52 so we could do that.
2:53 [Music]
3:24 [Music]
3:33 [Music]
3:34 Wonder if is LinkedIn still messed up?
3:37 No, it doesn't look like it. I mean, it
3:40 looks like it's messed up. Just remove
3:42 the destination from the stream and read
3:45 it.
3:47 Remove
3:53 Remove.
3:56 Read it.
4:09 All right. Now, it looks like we're
4:10 streaming on LinkedIn. Have you used
4:14 Superbase OpenAI
4:18 API with lovable yet?
4:22 I've used Superbase with lovable. I
4:24 don't I don't think I've used a specific
4:27 OpenAI API with lovable,
4:32 but I assume I assume that just means
4:34 you can hook into Open AI and do content
4:38 generation and things like that. Is that
4:40 right?
4:47 I'm still trying to get used to my
4:48 microphone. This is my microphone. Oh,
4:51 let me switch these.
4:58 Producer Brandon had a little computer
5:00 meltdown right before we went we went on
5:04 [Music]
5:23 Hello Tobias. What's happening? What's
5:26 shaking good people?
5:31 Wait.
5:35 [Music]
5:49 [Music]
6:00 Wow. Hey,
6:02 [Music]
6:21 [Music]
6:50 [Music]
7:05 Producer Brandon is back. It looks like
7:10 he's at least back on post-it note duty.
7:12 Yeah, it looks like he's back. Okay, we
7:15 have producer Brandon back.
7:18 Uh, do me a favor on Tik Tok. Share the
7:20 live if you'd be so kind. Uh, if you're
7:23 on Twitter or LinkedIn, tell friends.
7:26 Same with YouTube. saw a great breakdown
7:29 of the month
7:32 of
7:33 breakdown of the month of computer
7:36 costs. Everyone has bought into
7:40 reasoning that is
7:44 so hard to generate thousands of times
7:47 tokens even as they get cheaper. Oh,
7:49 compute. Yeah, math. Yeah, I I mean
7:53 listen
7:57 everyone I know has a different take on
7:59 the energy consumption
8:02 and
8:05 it is clear that the training
8:08 the training of these models takes a lot
8:11 of energy but like the training of GPT4
8:16 from what I can understand from like a a
8:19 recent thing I heard it's kind of the
8:21 equivalent of flying a seven a 747
8:25 8 hours
8:27 for one flight to train all of GPT4
8:31 and then inference is even cheaper.
8:34 So
8:36 I don't know like at this point I know
8:38 they need lots and lots of power and
8:40 they need lots and lots of GPUs to serve
8:43 up I I mean Open AAI by the end of this
8:46 year is likely going to have a billion
8:49 weekly users. So, they've got a they've
8:52 got a lot of stuff to do, but you know,
8:54 the whole thing of like it takes a
8:56 gallon of water to do a prompt is all
8:57 [ __ ]
8:59 I think it takes five drops or something
9:01 like that. So, anyway, whatever.
9:05 I I I mean, here's my thoughts on the on
9:08 the old um
9:11 on all the energy stuff is that
9:16 in relatively short order,
9:19 hang on, I'm just repositioning things
9:21 here for for the tick tockers.
9:25 We got a bit of a mess over there on the
9:27 Tik Tok. You can't read the comments.
9:29 It's a mess.
9:31 Um,
9:34 probably within two years,
9:38 Google was will certainly have it. Apple
9:41 may or may not, but we're going to have
9:45 local large language models on our
9:47 laptops and our phones that are probably
9:50 about as good as GPT4, maybe even GPT5,
9:53 maybe even reasoning models, I don't
9:55 know.
9:57 So a lot of the energy consumption on
9:59 inference is going to be offloaded to
10:00 the however much energy your phone
10:03 takes, right? Um and then the bigger
10:08 models are going to get more and more
10:09 efficient.
10:11 So So
10:14 is it an issue right now? It might be. I
10:17 don't know. Um should we be concerned
10:20 about it? I think we should be concerned
10:22 about it if it's threatening our power
10:25 grid, right? If we're prioritizing AI
10:29 training over people having electricity
10:31 in their house, then yeah, that would be
10:33 a problem.
10:36 But, you know,
10:38 I don't think it's doing that yet. So, I
10:40 don't know. I don't know. It's that's
10:43 above my pay grade.
10:45 Not the training. Every model is
10:46 throwing massively more compute at every
10:49 response.
10:51 They're thinking. Oh, they're thinking
10:54 when they don't need to. I agree with
10:55 that. It's killing the costbenefit
10:58 equations.
10:59 Those rapper apps are dying off due to
11:02 expense. Now, that's interesting.
11:06 Grock is spending $5 of compute
11:10 on what cost GPT 5.
11:13 Yeah, but that that's an ego thing.
11:15 That's Elon Musk wanting to have a
11:17 bigger wiener than Sam Alman.
11:25 I mean, listen,
11:28 all of these frontier companies are
11:30 trying to figure out
11:32 what what gets these things to to AGI
11:36 and ASI.
11:39 So, the fact that they're throwing
11:40 massive amounts of compute at it right
11:42 now just because it seems to be working,
11:46 I I don't know. I like
11:49 the where I go with this stuff is
11:52 we're we're in a we're in the
11:54 development
11:56 era of of this generative AI stuff.
11:59 If we were 20 years into this
12:03 and and it were this inefficient, I'd be
12:06 worried. But like nothing's nothing
12:10 stabilized. Nothing is has plateaued
12:12 yet. They're all trying to figure it
12:14 out. So, you know,
12:19 anyway, I don't know. I just again,
12:26 I don't know. I don't know that there's
12:27 clear enough math out there that we can
12:29 make informed decisions as consumers for
12:32 one thing. And then I also think the
12:34 efficiency is going to keep changing.
12:36 just saying that they are screwing
12:38 themselves for the same reason.
12:41 Brains don't think deeply on every idea.
12:44 Yeah. Yeah, I get it. It's fascinating
12:46 to me. Well, in theory, in theory,
12:50 OpenAI's routing model, GPT5, in theory,
12:54 it won't think when it doesn't need to.
12:56 In reality, it seems to be thinking a
12:58 lot more than it needs to. And if you
13:01 look at David Shapiro's recent posts,
13:03 he's he's all pissed off at chat GPT5
13:05 because it's doing all sorts of like
13:07 second-guessing of itself and trying to
13:10 offer up new things and create follow-up
13:12 questions. It's not just answering the
13:15 questions. It's trying to do it's trying
13:17 to be a bit more of your good buddy.
13:25 Oh my god. Um, have you ever used local
13:29 AI image gen models? Um, I used
13:38 not I didn't do it locally, but I I did
13:40 run a Google Collab instance of stable
13:43 diffusion with Dream Dream Booth early
13:46 on in 2022. So, I have I have installed
13:49 and run local image gen models, but I
13:51 did it in a cloud, you know, a cloud
13:54 instance that had a good GPU in it
13:56 because I'm on Mac laptops. I don't have
13:59 um
14:00 I don't I don't have like, you know, a
14:02 4090 or whatever uh to to be doing that.
14:07 It's trying to be politic politically
14:09 correct smarty pants. Yeah, I know.
14:13 Again, I I give I I'll give all of these
14:18 guys grace right now because
14:22 um they're trying to figure this stuff
14:24 out.
14:26 I mean, quite frankly, like the the the
14:29 amount of power
14:31 that each of us individually have, like
14:34 my my co-founder of StoryVine, she let
14:37 her um chat GPT subscription go. So,
14:40 she's just got free chat GBT and she's
14:43 like it's it's good enough for her who
14:46 her to do all the stuff she needs to do
14:48 without without paying for it. Like the
14:50 amount of power that we have, you know,
14:53 in our hands today because of what these
14:55 guys are doing is remarkable. And I
14:58 think us holding them to
15:01 to task or taking them to task for the
15:04 fact that they're changing personalities
15:06 and it's now it's overindexing this way,
15:08 now it's overindexing that way, now
15:10 they're using too much thinking, now
15:12 they're not using enough thinking. Like,
15:14 let them figure it out.
15:18 [Music]
15:26 I got a flash goal on Tik Tok
15:30 [Music]
15:37 about as fast as they fix their fixed
15:39 their naming convention. Well, you know
15:42 what? You know what it is? I I mean, the
15:43 naming convention is a pain in the ass,
15:45 but I think they they knew where they
15:47 were going was not,
15:49 you know, to to a routing model. They
15:51 didn't anticipate the the [ __ ] up with
15:53 the GPT5 launch and and deprecating 40.
15:57 Um, so so we're back to where we started
16:00 with there. Um,
16:02 [Music]
16:04 what I would say is that I I feel like
16:07 the the model companies when they do
16:09 really [ __ ] something up, they they have
16:11 been responding, right? when when uh
16:14 when that one version of GPT40 got
16:17 really sick of Fantic, they they shut
16:19 that thing down within a week. Um
16:25 it it looks like Elon's pulling back
16:28 some of his his tweets about the little
16:31 anime girl that you can talk to, you can
16:33 talk dirty with.
16:39 Wait, who are we supposed to go follow?
16:43 Google. Oh, Google AI Studio. Oh, is
16:47 that is that new? All right, we're
16:50 supposed to go follow Google at Google
16:53 AI Studio on X.
16:56 All right, let me
17:00 share this tab instead.
17:03 search for
17:07 at Google
17:11 AI
17:13 Studio. Um, what what I don't like about
17:16 that, Brandon,
17:21 is that that tells me that the [ __ ] that
17:23 they're going to release tomorrow is
17:25 going to be available in Google AI
17:26 Studio, but not on Gemini.
17:30 And if you don't know the difference,
17:31 Google AI Studio is where the developers
17:33 go hang out and where they they build
17:36 their stuff and access their APIs. So,
17:39 it's not a bad place. It's just not a
17:41 consumer place. The gemini.google.com
17:44 is the consumer place. Oh, by the way,
17:47 today I did I did something that I that
17:51 I'm okay with. I I wouldn't recommend
17:53 you do it. Um,
17:57 but I've got my my son's insurance is
18:00 switching from he turned 26, so he's off
18:02 my company insurance and we've got to
18:04 get him new stuff and they want things
18:07 like what are the specific, you know,
18:10 diagnosis and treatments and things like
18:12 that. So the the place where he gets all
18:15 of his treat treatments, they had 300 no
18:19 400 megabytes of medical records in
18:23 PDFs, like 700 pages and just all these
18:27 medical records and I put them into
18:29 Notebook LM and I said, "Hey, here's the
18:32 situation. Here's what I need." And it
18:34 it wrote up this report that was deadass
18:37 accurate in, you know, 45 seconds. It
18:40 was amazing. just amazing.
18:47 AI Studios come a long way. Yeah, I
18:49 agree with that. Let's let's let's
18:51 actually go there. Let me before I [ __ ]
18:52 all over them. Um
18:56 AI Studio.google.com.
19:03 Share this tab. Autosave is now enabled
19:05 by default.
19:08 Let's go. ++
19:11 oops
19:17 so URL context tool audio to audio
19:21 dialogue try Gemini's neural real time
19:25 dialogue
19:27 with audio and video inputs
19:30 native speech generation
19:33 native image generation
19:36 this is I guess where nano banana banana
19:38 is going to be when they release that.
19:40 If you haven't heard about Nano Banana
19:43 and then here's all the Yeah, this is
19:45 the interface here is much better than
19:46 it used to be. They've clearly gotten
19:49 some people on the team that know what
19:50 they're doing. It looks like It looks
19:52 like Meta after all those hundred
19:54 million dollar hires is losing a bunch
19:56 of people. Looks like they've got a bit
19:58 of a [ __ ] show over there.
20:01 So, I'm not expecting Meta to get any
20:04 any any good anytime soon. I think Rick
20:06 Rubin is the hair sponsor of the day.
20:14 >> What did I just see? Oh, I just Yes.
20:17 >> Oh, no. I just wanted to say that um I
20:19 think that from what I've seen just over
20:22 the weekend is that Google is planning
20:25 to leverage Google AI Studio as its new
20:29 frontier dropper where it will tease new
20:32 models and do a research preview of
20:35 them.
20:36 before it releases them to Gemini, which
20:39 is their more consumer stable friendly
20:42 product.
20:42 >> Yeah, I think that's generally been how
20:44 they've done it, but it's just it's like
20:46 not a lot of people know about it. The
20:48 the the the challenge I have with Google
20:51 is they announce stuff
20:55 as if it's a consumer launch and then
20:58 it's often not. It's often for
21:00 developers only. like I just I just
21:02 don't feel like they have a good sense
21:03 of who their who their actual customer
21:05 is. Listen, it's gotten a lot better in
21:07 the past 6 months and this is now a much
21:10 more usable site even though it's a bit
21:13 cockpit of a 747 still. Um it's much
21:17 better than it was. So yeah, it's not a
21:18 bad thing.
21:19 But I I've seen now two tweets uh since
21:23 the start of the show that have
21:25 suggested that we follow uh Google AI
21:27 Studio on uh on Twitter. Yeah. And it's
21:31 probably the announcements. It looks
21:34 like um OpenAI makes their announcements
21:37 on Thursdays. Um Apple used to make
21:39 theirs on Tuesdays. There was some meth
21:41 there was some like reason why Apple did
21:43 their announcements on Tuesdays. I
21:45 forget what it was, but it could Tuesday
21:47 could be tomorrow could be the day for
21:50 Google. So, we'll see. Um,
21:53 >> yeah.
21:56 Um,
21:59 I wanted to try something
22:02 tonight with Suno. Um, and I think it
22:05 could be fun. So,
22:10 Oh, wait. Maybe I'll
22:13 [Music]
22:28 see
22:30 [Music]
23:06 All right, I got an idea. All right, so
23:09 let's Here's what we're going to do.
23:13 We're going to go to Suno.
23:28 I'm going to go go to create.
23:32 All right. So, if you're if you uh if
23:35 you've never been here before, this is
23:37 sununo.com
23:39 and this is a music making
23:43 uh platform
23:46 and it makes all sorts of cool music.
23:49 So, let's just go grab one here. What do
23:51 we got?
23:58 You think you know me, you don't. You
23:59 can't bone a man. Not cold, not chance.
24:03 I'm
24:04 >> nice.
24:09 I like that.
24:14 [Music]
24:23 >> All right. So, it makes songs. It makes
24:25 songs. It makes songs.
24:29 So about a year year and a half ago on
24:33 UDIO, which is another similar kind of
24:35 thing like this, I did some spoken word
24:37 stuff and it did pretty good. And so I
24:39 wanted to see if Sunno could do good
24:43 with spoken word stuff. So,
24:46 um,
24:49 so actually what I'm going to do,
24:55 I have on my pile of crap here on my
24:58 desk
25:00 a book by one Brandon Tid
25:08 called Fantastical Journeys. So, we're
25:10 just going to pick a random random page
25:13 here.
25:16 So, what we're going to do is we're
25:17 going to go we're going to start with
25:19 audio and we're going to record. All
25:22 right. So, so I got to hang on hang on
25:27 people. This is not easy. All right.
25:32 And this is this is not what we're going
25:33 to use as the final thing. I just want
25:36 to see if it can recognize this a spoken
25:38 word.
25:43 In the depths of the ocean, there lived
25:46 a group of jellyfish who had a special
25:48 talent.
25:50 They could juggle. These jellyfish named
25:54 Gemma, Jack, and Jill put on a juggling
25:57 show every evening under the sea.
26:01 Okay, so let's hear it. In the depths of
26:06 the ocean, there lived a group of
26:08 jellyfish who had a special talent.
26:12 They could juggle these jellyfish named
26:15 Gemma. Right. Good pretentious spoken
26:18 word. New York City, East Village
26:22 Cadence. Right. So, I'm going to save
26:24 that.
26:27 Uhoh. Can't fetch the uploaded audio.
26:31 What? Oh,
26:33 do I have to re-record this? Damn it.
26:42 All right.
26:47 One night, the ring master, a wise old
26:50 alien named Orion, announced a special
26:53 act. The audience gasped as a comet
26:57 zoomed by and out came Luna the lunar.
27:04 All right, save. Please save.
27:08 Oh, come on.
27:14 Let me reload.
27:18 [Laughter]
27:27 record.
27:30 I said to the man, "When you record my
27:34 voice, the worst thing you could do is
27:38 to lose the recording of my voice that
27:41 you record because each performance is
27:44 the most important performance.
27:48 Don't lose it, loser.
27:53 Okay. Save.
27:58 Come on. You can do it.
28:07 Okay, it recorded it. So, now we're
28:10 going to do a cover.
28:13 We're going to continue. Okay, so now
28:14 we've got a recording. So, now we now we
28:17 need lyrics.
28:20 Okay. Oh, wait. Let's see. I said to the
28:23 man when you were. So, so it says
28:25 speech. So, it got that this was spoken
28:27 word.
28:28 So, what we're going to do,
28:31 let's go over to chat jpata.
28:34 Show this tab instead.
28:37 We'll do new chat. And I'm gonna say
28:41 you are an a you are an off Broadway
28:52 spoken word
28:55 artist
28:57 with
29:01 mastery of the
29:05 words you speak.
29:11 I want
29:15 I want you to craft
29:23 a beat
29:25 poem
29:27 spoken word performance.
29:32 No stage directions.
29:39 about the challenges
29:44 and to challenges and toils
29:49 of being
29:52 an AI influencer
29:57 who goes live
30:01 every night.
30:10 The audience on the lower east side
30:15 should appreciate it.
30:21 I'm just here to entertain myself,
30:23 people. All right. Should appreciate it.
30:26 Okay,
30:29 let's get ourselves a script here.
30:31 thinking. I'm using uh chatbt5 thinking
30:34 mini uh because I think it's actually a
30:36 really good writer of of the three
30:38 models, the three main new models.
30:40 That's the one that I like. I turn the
30:42 camera on like a prayer, palms pressed
30:45 against the glass and LED, and the city
30:47 exhales into my lap. A slow subway of
30:50 faces scrolling past. All right, good.
30:53 This This looks This sounds pretentious
30:56 and is long. Um, let me just say tighten
31:00 it up a bit. Tighten.
31:03 No, [ __ ] it. Let's do it. Let's do a
31:05 long one.
31:07 Copy.
31:10 All right. Let's go back to Sunno.
31:14 So, I'm going to completely replace the
31:15 words
31:20 and I'm going to call it um
31:23 let's call it spoken word performance.
31:28 with
31:31 acoustic
31:35 nylon
31:37 guitar bongos
31:42 and
31:44 an appreciative
31:47 audience
31:51 as the performance
31:55 builds.
31:57 All right. Or be build beards.
32:01 Builds builds builds. Dang it. Dang it.
32:07 Shut up.
32:14 Yeah. Put the tag spoken word. spoken
32:17 word performance with acoustic nylon
32:19 guitar, bongos,
32:23 and an appreciative
32:27 audience
32:29 as the performance builds.
32:34 All right. Oh, we almost maxed out our
32:36 words. We're at 48
32:38 57 out of 5,000 words. This is going to
32:42 be long and obnoxious.
32:43 [Laughter]
32:48 Let's see if it's any good. I think it
32:51 will be.
32:56 At least he remembers we're here. Sort
32:58 of. Sort sort of. What's happening?
33:01 Source camp. What's shaking? What's
33:03 going down? Somebody just followed me on
33:06 Tik Tok. Thank you for that.
33:09 Valerie Cox liked the AI learning lab on
33:12 Twitter. Thank you for that.
33:15 Uh, if you if you're on YouTube and you
33:18 haven't subscribed to the YouTube
33:20 channel, please do that. I I don't know
33:23 what's happening with Tik Tok. Part of
33:25 what I've heard is it's all going to go
33:27 away in September. Part of what I've
33:28 heard is it's going to be fine. So, I
33:30 don't know. But um but why we're dual
33:35 simo casting is just at at the point at
33:39 which they just turn that [ __ ] off, you
33:41 know. Be subscribed over on YouTube. All
33:44 right, here we go.
33:48 [Music]
33:57 I turn the camera on like a prayer.
34:00 palms pressed to glass.
34:02 >> That didn't work at all.
34:06 [Music]
34:10 >> I turned the camera on like a
34:11 >> All right. So, that didn't work at all.
34:13 So, let me do let me get rid of spoken
34:15 word performance. Performance probably
34:17 screwed it up.
34:19 Let's just try that.
34:23 I was thinking that maybe performance
34:25 would would make it do something more
34:28 better.
34:31 I should probably give this thing a
34:32 title, too.
34:35 Advanced options. Oh, weirdness.
34:40 Experimental results.
34:42 [Music]
34:47 I turn the camera on like a prayer.
34:50 Palms pressed to glass and LED. And the
34:54 city exhales into my lap. A slow subway
34:57 of faces scrolling past.
35:02 Every night I file myself into a
35:04 thumbnail. The smile that nets, the
35:06 pause that keeps them watching, the joke
35:08 that claps like loose change down a
35:11 stairwell.
35:12 I am a human gift in a world that wants
35:15 me looped.
35:17 Good evening, I say. And it is both
35:19 greeting and invoice. I sell presents in
35:22 micro payments.
35:24 >> Kind of cool. Um, I just changed some
35:26 settings to be more influenced by the
35:29 audio. But let's listen to the other
35:31 one.
35:36 [Music]
35:42 I turn the camera on like a prayer.
35:46 Palms pressed to glass and LED. And the
35:49 city exhales into my lap. A slow subway
35:52 of faces scrolling past. Every night I
35:55 file myself into a thumbnail. The smile
35:58 that nets, the pause that keeps them
36:00 watching, the joke that claps like loose
36:03 change down a stairwell. I am a human
36:06 gift in a world that wants me looped.
36:10 Good evening, I say. And it is both
36:12 greeting and invoice.
36:15 I sell presents in
36:16 >> That's great. Okay, let's see.
36:20 [Music]
36:25 I turned the camera on like a prayer.
36:27 >> I don't know why it's making me a woman.
36:35 [Music]
36:40 >> I turned the camera.
36:41 >> Weird. Um, let's see. Spoken word.
36:47 Let's go.
36:48 spoken word,
36:51 male voice,
36:53 acoustic nylon guitar.
36:56 We'll just say percussion.
37:00 And then I'm going to lose the audience
37:01 because I just have a feeling that's
37:04 screwing things up.
37:13 [Music]
37:24 I turn the camera on like a prayer.
37:28 [Music]
37:40 I turn the camera on like a prayer.
37:44 Weird.
37:46 Anybody know? Cuz here, so here's Did I
37:49 do this too high? I said to the man, no.
37:59 Let's see
38:02 why. There we go. Oh, voice gender male.
38:05 There we go. Weirdness.
38:10 Put that there. Style influence. Put a
38:13 little higher. Audio influence. We'll
38:15 put it higher. Now we'll do this.
38:20 Oh, song title. Um, pretentious.
38:25 Pretentious
38:30 influencer.
38:50 [Music]
38:53 I turned the camera on like a prayer.
38:56 [Music]
39:00 >> Oh, wait. What?
39:02 >> I
39:03 >> Why do I have so many of these things?
39:07 Well, we'll try the new ones.
39:13 [Music]
39:16 This sounds spoken wordy.
39:19 [Music]
39:28 I turn the camera on like a prayer.
39:32 Palms pressed to glass and LED.
39:34 [Music]
39:36 And the city exhales into my lap. A slow
39:38 subway of faces.
39:40 >> That's better.
39:42 [Music]
40:04 I turn the camera on like a prayer.
40:06 palms pressed to glass and LED and city
40:10 exhales into my
40:11 >> producer AI has unlimited prompting in
40:13 their chat hole until September 1st. Oh,
40:17 cool. Okay, I'm going to head over. I
40:18 want to I was going to try producer next
40:21 cuz I want to see if it can do this.
40:23 >> A slow subway of faces scrolling past.
40:26 Every night I follow myself into a
40:28 thumbnail. The smile that nets the pause
40:30 that keeps them watching. The joke that
40:32 claps like loose change down a
40:34 stairwell. I am a human gift in a world
40:36 that wants me looped. Good evening, I
40:38 say. And it is both greeting and
40:40 invoice. I sell presents in micro
40:42 payments. Attention claps time. My heart
40:44 is a push notification. My dreams are
40:46 buffered. I'm on stage and in the DMs
40:49 and on every playlist they promise.
40:50 >> Yeah, I saw that. I saw that, Brandon.
40:52 Um, okay. So, let's go. Let's let's flip
40:55 over to producer.ai. So, that's it. It
40:58 did okay, but you it's a little finicky
41:00 with the controls.
41:02 Um,
41:04 so
41:06 producer.ai if you don't know it is used
41:09 to be refusion
41:12 and I obviously just
41:16 Oh no, they just their CSS didn't load.
41:19 There we go. Okay,
41:23 now can I record something here? plus
41:26 audio
41:29 recording.
41:39 I wonder.
41:43 Okay, retry.
41:48 I wonder as I wander deep down the path,
41:53 down the archways,
41:55 into the woods, out of the woods, the
41:59 shrubberies,
42:01 the shrubberies.
42:03 So well manicured
42:06 into the woods.
42:08 Out of the woods. Now, if you're
42:11 watching this and you're like, "You
42:13 sound like an idiot." Thank you. That is
42:16 the intended
42:19 That is the intended reaction.
42:23 Um, let's see. Did that sound at all
42:25 good? What's going on? Uhoh. Did I crash
42:29 it? I might have crashed it. Damn it.
42:35 Ah.
42:38 Wait.
42:40 Save
42:51 black bar.
43:02 Oh my god.
43:10 Exit page.
43:14 Reload.
43:16 I'm having uh I'm having memory issues
43:19 here. Google Chrome. So apparently
43:22 Perplexity has um said that they want to
43:28 buy Google Chrome if it if it if they
43:30 get forced to sell it
43:33 um for $34.5 billion.
43:38 Not that they have that, but they'll go
43:40 get that.
43:41 Um I would love for someone to buy it
43:44 just to fix the [ __ ] memory leaks.
43:46 Okay.
43:49 Google Chrome leaks memory tabs. Tab
43:54 hoarders hoarding tabs. Close your tabs.
43:57 Close your tabs. Don't have too many
44:00 tabs open. Memory leaks. It won't work.
44:04 Nothing will work. You are a bad person
44:07 if you have more than six tabs. Tab
44:10 hoarder. Tab hoarder.
44:14 Okay, let's play this.
44:22 Solid. Solid. I like it. Save.
44:28 Save.
44:31 Crash your browser. Crash your browser.
44:40 Oh, no longer sharing screen. Yeah,
44:43 well, it doesn't seem to [ __ ] matter.
44:46 Uh,
44:49 [Music]
44:52 retry. Nothing. Dead. Dead.
44:57 Damn it.
44:59 Exit page. Reload. Oh my god.
45:05 Well, we'll just do one without
45:07 recording.
45:12 Slam poetry night at the salon. That's
45:14 exactly what it is. All right. So, we're
45:16 going to go
45:18 we're going to go um so wait, I'm gonna
45:21 let let's go back to chat GBT for a
45:23 second. Share this tab. Uh
45:27 let's do one just as pretentious,
45:36 but about
45:40 tab hoarders
45:44 who keep too many tabs open on Google
45:50 Chrome
45:52 and crash it out
45:58 and crash it out
46:00 due to memory leaks.
46:05 This should be
46:08 pro- tab hoarder.
46:13 Period. A lament
46:16 on Google's inability to manage memory
46:26 like too many humans.
46:33 Yes. So, we're going to do slam poetry
46:36 on uh Tab Cathedral. I bow to the altar
46:41 of a thousand favicons. Each tiny
46:43 stained glass promise. Each tab a relic,
46:46 an argument, an unfinished prayer. We do
46:49 not close. We collect. This is good.
46:57 Okay, great. So, going back to producer,
47:01 show this tab instead. So, now we're
47:03 going to say, um,
47:06 I want you to create a
47:11 slam
47:13 poet slam
47:17 poet performance
47:22 performance
47:25 with spoken word.
47:33 a male voice
47:36 and
47:39 a poem called
47:42 the tab hoarders
47:46 lament.
47:50 Here are
47:52 the lyrics.
47:56 I wonder if producer can do spoken word.
47:59 We're about to find out.
48:04 Okay, if you don't know Producer, this
48:08 is the only of the music generating
48:12 tools that allows you to do this in an
48:15 ongoing conversation.
48:19 So, that I'm pretty excited about. All
48:22 right, let's see what it says here.
48:25 Enjoy free text and voice chat all week.
48:29 Songs in other generations still use
48:31 credits. Okay.
48:34 Love those lyrics. Sharp, relatable, and
48:36 ripe for Sam slam poetry performance.
48:39 Here's how I'd frame the tab hoarders
48:41 lament as a spoken word piece with a
48:44 male voice over leaning heavy on pulse
48:46 and space, minimal backbeat, and a bit
48:49 of atmospheric instrumental.
48:52 spoken prompt, spoken words, slam poetry
48:55 performance, male cinemat. Okay. Yeah,
48:58 this is good. There's your lyrics. Ready
49:01 to continue into creation. Um, yes, this
49:06 is good.
49:09 and also include
49:14 um audience
49:18 reaction,
49:20 laughter,
49:22 laughter and spontaneous
49:26 applause
49:29 as appropriate
49:31 to the performance.
49:39 Once you have that added,
49:44 just make the track.
49:47 Okay. Uh text on Tik Tok.
49:52 You need You need to be wearing a
49:54 cashmere gray turtleneck. Exactly. I do.
49:57 Or a pink bow.
50:03 Yeah.
50:07 All right, let's Oh, it's making this.
50:08 It's making our thing right now. All
50:10 right, let's see what we get here,
50:12 people.
50:28 We do not close,
50:30 we collect.
50:32 We are archaeologists of distraction
50:34 curators of
50:36 >> Okay. Um, I want it
50:40 to sound like a guy from Brooklyn,
50:47 not London.
51:03 [Music]
51:09 subtle Brooklyn tone.
51:31 I'm going to flip back over to Suno
51:33 quick and
51:39 go.
51:43 I bow to the altar of a thousand
51:45 favicons. Each a tiny stained glass
51:47 promise. Each tab relic, an argument, an
51:49 unfinished prayer. We do not close. we
51:52 collect. We are archaeologists of
51:54 distraction, curators of half begun
51:56 obsessions, keepers of breadcrumbed
51:57 identities. They call us.
51:59 >> Okay, so that's bad, but we can fix it
52:03 because we have we have chatability. So
52:06 I'm going to go
52:09 okay the words are okay but there is no
52:17 acoustic
52:19 guitar or percussion
52:22 or audience.
52:25 It is all dry and it sucks.
52:33 I want this to feel
52:37 like I am on the Lower East
52:44 Side, New York City,
52:47 1989.
52:50 Go.
52:53 [Laughter]
52:59 If you if you take your prompting too
53:01 seriously, lighten up and have some fun
53:03 with it.
53:05 Tell your tell your LLM it's a dumb
53:07 dumb.
53:12 If you if if you're not putting down
53:14 your LLM, you're not doing it right.
53:17 [Laughter]
53:22 >> All right.
53:24 But you got to be careful because when
53:26 they get close, we collect We are
53:29 archaeologists of distraction.
53:31 Curators of half begun obsessions.
53:33 Keepers of breadcrumbmed identities.
53:36 They call us hoarders like it's a crime.
53:38 I call us archavist of the incandescent
53:40 now.
53:43 Still no background music or
53:50 club sounds.
53:55 Rewrite
53:57 the prompt.
54:01 Here's a revised prompt.
54:04 Slam poetry, spoken voice, subtle
54:07 Brooklyn accent, smoky nightclub, loud
54:10 audience chatter, glasses clinking,
54:12 laughter,
54:14 fingerstyle acoustic guitar, jazzy
54:16 upright bass, run new track. All right,
54:18 let's see if it can do it.
54:24 So, so one of the reasons that I like
54:27 doing exercises like this where I have
54:29 no clue if it can even do it is to
54:32 figure out a can it do it and b what do
54:35 you have to do to get it there? Um,
54:42 and this kind of problem
54:43 >> bow to the altar of a thousand favicons.
54:46 It still didn't
54:47 >> a tiny stained glass promise.
54:50 >> Each tab a relic, an argument, an
54:53 unfinished prayer.
54:56 >> We do not close, we collect.
54:58 >> We are archaeologists of distraction.
55:01 >> Bad
55:01 >> curators of half begun obsessions,
55:04 keepers of breadcrumbed identities.
55:06 Numerous killing that small. Under our
55:08 little empires, it's meous
55:11 becomes less accurate. There is an
55:14 ethics to accumulation. Not greedy, not
55:16 manic, but rather tabs or trust jars
55:19 open wide. Collecting the small currency
55:22 of pension.
55:23 [Music]
55:29 >> Wow. This is It's like it it can't
55:31 articulate the spoken word with the
55:33 music. It's It's [ __ ] up all these
55:35 words.
55:39 >> All right, let me You know what I want
55:41 to do? I like this. I like this uh
55:45 I like this prompt. Let's take this
55:47 prompt back to Suno
55:50 and let's create a new
55:53 a new song or a new a new track. So I'll
55:57 put the lyrics there style
56:04 real club throughout.
56:08 So we're going to go
56:11 Oh, wait. So that's lyrics. So lyrics
56:14 we're going to get from chat GPT.
56:22 Put those here.
56:28 Advanced options. Weirdness.
56:32 Style influence. Song title
56:36 tab
56:38 hoarders
56:41 lament.
56:50 All right. Slam poetry performance.
56:53 Spoken word. I'm going to start with
56:54 spoken word because performance seemed
56:57 to screw it up.
57:04 All right, here we go. Tab hoarders
57:07 lament insuno.
57:10 I was actually surprised that Refusion
57:12 did bad on that. What that tells me is
57:14 that Refusion was trained more on music
57:17 than this kind of stuff.
57:19 And it seems that Sunno's got a broader
57:23 a more broad pallet to draw from.
57:28 [Music]
57:38 I bow to the altar of a thousand's
57:40 favicons.
57:41 Each a tiny stained glass promise. Each
57:44 tab a relic and argument and unfinished
57:46 prayer. We do not close. We collect.
57:49 >> Much better.
57:50 >> We are archaeologists of distraction.
57:53 Curators of half begun obsessions.
57:55 Keepers of breadcrumbmed identities.
58:00 They call us hoarders like it's a crime.
58:04 I call us archavist of the incandescent.
58:06 Now every tab is a shrine to a thought
58:09 that refused to die. A headphone cable
58:11 tangled with yesterday's revolution.
58:14 A recipe we swore we'd master. A thread
58:17 we swore we'd answer.
58:20 >> Interesting.
58:22 Interesting. It doesn't know how to say
58:24 archivist.
58:26 So, we'll misspell it. Archivist.
58:32 All right, let's hear the next one.
58:35 >> Like this one. It's like slam poetry.
58:38 Yeah, that's what it that's trying to
58:39 get.
58:41 >> I bow to the altar of a thousand
58:42 favicons.
58:47 [Music]
58:55 Good mongos.
58:57 >> I bow to the altar
58:59 of a thousands favors.
59:03 Each a tiny stained glass promise. Each
59:06 tab relic, an argument, an unfinished
59:10 prayer.
59:12 We do not close, we collect.
59:14 We are archaeologists of distraction,
59:17 curators of half begun obsessions,
59:20 keepers of breadcrumbmed identities.
59:23 They call us hoarders like it's a crime.
59:26 I call us archeists of the incandescent.
59:28 Now every tab is a shrine to a thought
59:31 that refused to die.
59:36 It's good.
59:42 [Music]
59:58 A bow to the altar of a thousand
1:00:00 favicons. Yeah, I don't know why it's
1:00:02 doing that weirdass
1:00:04 uh like grally voice, but let's let's
1:00:06 change some things here. We'll take
1:00:08 weirdness down. Put style influence up.
1:00:13 Go to more options, song title, word
1:00:15 space. I don't know what that is.
1:00:18 Um, all right. Let's rewrite the prompt
1:00:21 here
1:00:23 because I think it's [ __ ] it up
1:00:24 somehow. Spoken word slam poetry
1:00:26 performance. Male voice. Let's just do
1:00:29 male voice. Smoky
1:00:33 Lower East Side Nightclub.
1:00:36 We're just going to go.
1:00:40 We'll just go
1:00:46 New York City nightclub
1:00:53 ambiance.
1:00:59 We'll do audience chatter, clinking
1:01:02 glasses, laughter, applause at big
1:01:04 moments,
1:01:07 period,
1:01:09 live finger style, acoustic guitar,
1:01:11 jazzy upright bass. I'm just going to
1:01:13 say upright bass.
1:01:15 Conga shakers, rhythmic background
1:01:17 groove, heavy reverb. Let's lose that.
1:01:22 Streetwise energy. Let's lose that.
1:01:29 So, I assume Streetwise Energy gave us a
1:01:32 guy that sounded kind of black and kind
1:01:34 of grally, I guess. Words cut through
1:01:37 the analog haze. We can lose that.
1:01:41 Raw Bohemian. Let's get rid of raw.
1:01:44 Bohemian's fine. Unpolished. Bustling
1:01:47 with life. Real club ambiance
1:01:49 throughout. Okay,
1:01:52 let's try this.
1:02:01 Oh, I can probably add audio.
1:02:08 Hopefully, it doesn't crash. Well, let's
1:02:10 let's see what happened without adding
1:02:12 audio. I bow to the altar of a thousand
1:02:15 favicons.
1:02:17 Each a tiny stained glass promise. Each
1:02:19 tab a relic, an argument, an unfinished
1:02:22 prayer. We do not close, we collect.
1:02:26 We are archaeologists of distraction.
1:02:30 [Applause]
1:02:31 [Music]
1:02:40 I bow to the altar of a thousand
1:02:42 favacons.
1:02:43 [Music]
1:02:44 Each a tiny stained glass promise. Each
1:02:47 tab a relic, an argument, an unfinished
1:02:49 prayer. We do not close.
1:02:52 We collect. We are archaeologists of
1:02:55 distraction, curators of half begun
1:02:58 obsessions, keepers of breadcrumbmed
1:03:00 identities.
1:03:03 They call us hoarders like it's a crime.
1:03:07 I call us archavists of the
1:03:08 incandescent. Now every tab is a shrine
1:03:11 to a thought that refused to die. A
1:03:13 headphone cable tangled with yesterday's
1:03:16 revolution. A recipe we swore we'd
1:03:18 master. A thread we swore we'd answer.
1:03:21 We keep the world in micro windows like
1:03:23 sailors with sex mapping storms of
1:03:26 curiosity.
1:03:28 [Laughter]
1:03:31 Chrome hates us. Chrome groans under our
1:03:34 little empires.
1:03:36 It's memory like an anxious brain on the
1:03:38 subway. Leaking,
1:03:42 forgetting,
1:03:43 dropping passengers on the line.
1:03:47 Memory leak they say as if sin were a
1:03:49 bug fixable with a patch. Google with
1:03:52 its polished teeth and corporate himnels
1:03:54 cannot fathom the sanctity of our open
1:03:56 questions. It trims it prunes. It
1:03:58 gaslights our tabs into sleeping bags.
1:04:01 But we are stubborn, gentle, glorious,
1:04:04 unruly.
1:04:06 Let the browsers call it inefficiency.
1:04:10 We call it fidelity. We are a million
1:04:13 tiny vows. I will return. I will return.
1:04:16 and I will return.
1:04:19 [Music]
1:04:22 Each tab is a bookmark for an argument
1:04:25 with the future. We keep versions of
1:04:27 ourselves and duplicate the person we
1:04:29 were at 2 a.m. The person who clicked
1:04:32 that headline in a cafe. The person who
1:04:34 loved a sentence. Close one and you risk
1:04:37 killing that small particular grief.
1:04:40 Close one and the map of you becomes
1:04:42 less accurate. There is an ethics to
1:04:45 accumul.
1:04:50 >> All right. It's a little loud, but I
1:04:52 like it a lot.
1:04:55 It's a little aggressive, but I like it.
1:04:57 So, we're going to give that one a We're
1:04:59 going to give that one a hearty heart
1:05:01 thumbs up so we can come back to that
1:05:03 one. Um,
1:05:11 let me let me blast out a couple of more
1:05:16 renders here
1:05:18 and then I'm going to record some audio
1:05:20 because I have a feeling this could
1:05:22 crash it out. Check out Gil Scott Heron
1:05:25 though. Okay, cool.
1:05:28 Interesting how it's it's a million tiny
1:05:30 valves. Yeah.
1:05:37 Okay. Record.
1:05:42 A thousand tiny valves. Vows and tabs
1:05:46 and tabs and valves. Chrome guarding our
1:05:50 lives as if they give a damn. They
1:05:52 don't.
1:05:57 A thousand tiny valves. Vows and tabs
1:06:00 and tabs and vows. Chrome guarding our
1:06:04 live. Okay. Save.
1:06:07 Don't crash. Please don't crash.
1:06:11 Crashed.
1:06:21 I'll try one more. I don't think it's
1:06:23 going to work.
1:06:25 I don't think it's going to work. I
1:06:27 record. I record. I hit upload. It says
1:06:31 no. It says no. Can't fetch the file.
1:06:34 Can't fetch the file. Where did you put
1:06:36 it? Loser. User. User error. Loser
1:06:40 error.
1:06:42 All right.
1:06:53 Did we already mention Meta is
1:06:56 partnering with Midjourney? He did not.
1:06:58 Success. Okay, that uploaded. So, we're
1:07:00 going to do cover continue. Did it lose
1:07:04 Did I lose all of my things? Yes, I lost
1:07:07 my lyrics, but that's okay. We'll just
1:07:10 put them back. Oops.
1:07:13 Let me go get them. Copy. I like these
1:07:17 lyrics. I think they're good.
1:07:19 Paste.
1:07:21 And then let's go down. I think we have
1:07:23 new options now.
1:07:25 Advanced options, new exclude styles,
1:07:30 audio influence, we'll put that a little
1:07:32 higher.
1:07:34 We'll bring weirdness up a little bit.
1:07:39 Song title
1:07:41 tab hoarders unite. This one's going to
1:07:45 be called unite
1:07:47 exclamation point.
1:07:49 All right.
1:07:51 So, what we're going to do here is I I
1:07:53 think we're gonna we're going to go make
1:07:54 a midjourney
1:07:57 we're going to go make a midjourney
1:07:58 avatar. Then we're going to go into
1:08:00 Hedra and we're going to make a spoken
1:08:01 word performance
1:08:04 and we're going to we're going to upload
1:08:05 it on on Twitter and then Twitter's
1:08:08 going to show it to like two people cuz
1:08:10 they're [ __ ]
1:08:17 like this.
1:08:21 [Music]
1:08:24 I bow to the altar of a thousand
1:08:26 favicons.
1:08:28 Each a tiny stained glass promise. Each
1:08:31 tab relic an argument and unfinished
1:08:34 prayer. We do not close. We collect.
1:08:38 We are archaeologists of distraction.
1:08:40 Curators of half-bun obsessions. Keepers
1:08:43 of breadcrumbmed identities.
1:08:46 They call us hoarders like it's a crime.
1:08:48 I call us archivists of the incandescent
1:08:51 now. Every tab is a shrine to a thought
1:08:53 that refused to die. A headphone cable
1:08:56 tangled with yesterday's revolution.
1:08:59 A recipe we swore we'd master. A thread
1:09:02 we swore we'd answer. We keep the world
1:09:04 in micro windows. Like sailors with sex,
1:09:07 mapping storms of curiosity. Chrome
1:09:10 hates us. Chrome groans under our little
1:09:11 empires. It's memory like an anxious
1:09:13 brain on the subway. Leaking,
1:09:15 forgetting, dropping passengers on the
1:09:16 line. memory leak. They say as if sin
1:09:19 were a bug fixable with a patch. Google
1:09:21 with its polished teeth and corporate
1:09:22 himnels cannot fathom the sanctity of
1:09:25 our open questions. It trims, it prunes,
1:09:27 it gaslights our tabs into sleeping
1:09:29 bags. But we are stubborn, gentle,
1:09:32 glorious, unruly.
1:09:35 Let the browsers call it inefficiency.
1:09:37 We call it fidelity. We are a million
1:09:40 tiny vows. I will return. I will return.
1:09:42 I will return. Each tab is a bookmark
1:09:45 for an argument with the future.
1:09:48 We keep versions of ourselves and
1:09:49 duplicate. The person we were at 2 a.m.
1:09:52 The person who clicked that headline in
1:09:54 a cafe. The person who loved a sentence.
1:09:56 Close one and you risk killing that.
1:10:00 >> This is all your fault.
1:10:02 >> You never
1:10:03 >> What? What the [ __ ] are those? Oh,
1:10:05 similar.
1:10:07 Oh, cuz I'm in here. I get it.
1:10:10 Back here.
1:10:21 Wait, where's
1:10:31 about to the altar of a thousand
1:10:32 favicons,
1:10:34 each a tiny stained glass? fathom the
1:10:36 sanctity of our open questions as
1:10:38 keepers of breadcrumb
1:10:40 thought that refused to die. Chrome
1:10:42 hates us world and micro.
1:10:44 >> That's pretty cool. All right, let's try
1:10:46 this again.
1:10:55 Tic tac pen. Xai is coming up with an
1:10:59 operating system that is purely AI. Oh,
1:11:02 macro hard. Yeah, so that's thanks for
1:11:04 that reminder, Joker.
1:11:06 So Elon Musk, you know, trying to be the
1:11:09 master troller that he is, is going
1:11:11 after Microsoft with a new company
1:11:13 called Macro hard, which is, you know, a
1:11:17 division or it's probably right now a
1:11:20 joke company
1:11:23 spin-off or spin up inside XAI. But the
1:11:27 basic idea is he thinks that he can
1:11:30 fully replicate Microsoft
1:11:33 by basically turning all of its workers
1:11:36 including its executive management into
1:11:39 AI bots.
1:11:41 So So Macroheart is going to be a fully
1:11:43 AI company that tries to replicate
1:11:47 everything that Microsoft is doing,
1:11:49 which listen I I mean, first of all,
1:11:53 like the guy's got balls. like he he
1:11:55 just he he just trolls consistently and
1:11:58 relentlessly, right? He just just all he
1:12:01 does. Um and he's smart. I mean, if you
1:12:05 look at OpenAI's um levels of AGI, the
1:12:10 fifth level of AGI was organization,
1:12:12 right? You had you had chat bots and
1:12:15 then you had uh reasoning and then you
1:12:18 had agents and then you had innovators.
1:12:21 So, we're somewhere right now between
1:12:23 agents and innovators. And then the
1:12:25 fifth level was organizations.
1:12:28 And an organization would be a
1:12:29 collection of agents that go off and
1:12:31 innovate and and do your bidding. And
1:12:35 so, it's a relatively straightforward
1:12:38 process to map
1:12:41 all of the functions within a company
1:12:43 like Microsoft and assign AI to them and
1:12:46 then and then optimize for that. So,
1:12:49 it's going to be fascinating to see.
1:12:51 Okay, go to Higsfield AI on X. Okay, I
1:12:54 will. But we got to taborters unite.
1:12:56 We've got to get We got to get the right
1:12:58 sound here, kids.
1:13:01 [Applause]
1:13:01 [Music]
1:13:02 I like the I like the applause to start
1:13:05 here.
1:13:06 >> I bow to the altar of a thousand favors.
1:13:10 >> Each a tiny stained glass promise.
1:13:13 >> It went back to that weird
1:13:15 [Music]
1:13:20 I bow to the altar of a thousand
1:13:22 favicons. Each a tiny stained glass
1:13:25 promise. Each tab relic and argument and
1:13:28 unfinished prayer. We do not close, we
1:13:30 collect.
1:13:32 We are archaeologists of distraction,
1:13:35 curators of half-bun obsessions, keepers
1:13:37 of breadcrumbmed identities.
1:13:39 They call us hoarders like it's a crime.
1:13:42 I call us archivists of the incandescent
1:13:45 now. are chests.
1:13:48 I bow to the altar of a thousand
1:13:49 favicons. No,
1:13:51 >> I bow to the altar of a thousand
1:13:52 favicons. Each a tiny stained glass
1:13:55 promise. Each tab a relic, an argument,
1:13:57 an unfinished prayer we do not close. We
1:14:00 collect. We are archaeologists of
1:14:02 distraction. Corders like it's a
1:14:04 headphone, a thread we swirl. As chrome
1:14:06 groans under say as if.
1:14:08 >> All right, that one has no. So, let's do
1:14:11 style higher.
1:14:13 This is driving me crazy. male
1:14:17 um spoken word slam poetry.
1:14:53 Remix
1:14:55 remix.
1:14:57 All right, let's go to Higsfield AI on X
1:15:05 at Higsfield.
1:15:10 Higsfield's an interesting company. They
1:15:12 seem to have good technology, but they
1:15:14 market they they market the snot out of
1:15:16 it, and it's kind of annoying. Ah,
1:15:18 introducing Higsfield Records.
1:15:20 Interesting. The first AI record label.
1:15:23 I I called this six months ago. I'm
1:15:26 telling you, every one of these content
1:15:29 generation companies is probably going
1:15:31 to become a studio
1:15:33 or or a label.
1:15:35 Multi-million dollar contracts already
1:15:37 in motion. You can become the next
1:15:39 global idol with us. It's easy. No
1:15:42 talent required. We reach retweet to get
1:15:46 the details in your DM.
1:15:51 Uh just adding in here, if you look at
1:15:52 the first like comment underneath it, uh
1:15:56 there's a form that you can fill out to
1:15:58 be on the wait list, but they do ask for
1:15:59 your Twitter handle, so they're probably
1:16:01 looking for you to retweet it.
1:16:03 >> Yeah. Well, they said retweet it.
1:16:06 >> No, but say it's not just the retweet.
1:16:08 There's also a form that you fill out.
1:16:12 >> Got it. Okay.
1:16:14 All right. Cool. Um but let's retweet it
1:16:17 anyway. quote.
1:16:19 Um, I called this
1:16:23 six months back
1:16:26 um,
1:16:28 media
1:16:31 generation
1:16:33 platforms
1:16:38 will be getting into the content game.
1:16:47 The opportunity
1:16:52 opportunity
1:16:55 to get
1:16:59 let's see to be an artist of the future
1:17:04 is now.
1:17:07 Let the haters hate. This is
1:17:13 the new game.
1:17:16 Rock on.
1:17:21 Tag Scoilizer in there.
1:17:28 Post. All right,
1:17:31 let's go back to Suno. See if we got a
1:17:36 See if we got some action here. We need
1:17:38 something that doesn't suck.
1:17:41 This sounds good.
1:17:49 A bow to the altar of a thousand
1:17:51 favicons.
1:17:53 Each a tiny stained glass promise. Each
1:17:56 tabar relic an argument and unfinished
1:17:59 prayer. We do not close, we collect.
1:18:04 We are archaeologists of distraction,
1:18:06 curators of half begun obsessions,
1:18:08 keepers of breadcrumbmed identities.
1:18:10 They call us hoarders like it's a crime.
1:18:13 I call us archivists of the
1:18:15 incandescent. Now every tab is a shrine
1:18:18 to a thought that refused to die.
1:18:22 [Music]
1:18:33 Great question. Who who's asking this?
1:18:35 Wolf man Clint. I don't know why or get
1:18:37 it pronouncing pronouncing of that word
1:18:40 is it not in its vocabulary. So in its
1:18:44 vocabulary is a really interesting
1:18:46 choice there
1:18:48 in in pre-ai systems you would actually
1:18:52 have pronunciation for all the words.
1:18:54 You would have a vocabulary and you
1:18:56 would have pronunciations of it.
1:18:58 Generative AI is not a database. It's
1:19:00 not it's not um uh you know logic based
1:19:05 computing. It's it's making these words
1:19:08 in real time. So, um it it it obviously
1:19:12 doesn't have enough training around
1:19:14 certain words that it that it's just
1:19:17 mucking them up when it when it's trying
1:19:19 to to generate that sound. It's just not
1:19:21 doing a good job of it.
1:19:24 But it's a really good question.
1:19:26 >> About to the altar of a thousand.
1:19:28 >> No,
1:19:29 [Music]
1:19:31 baby.
1:19:35 [Music]
1:19:43 I bow to the altar of a thousand
1:19:44 favicons.
1:19:46 Each a tiny stained glass promise. Each
1:19:48 tab relic an argument and unfinished
1:19:51 prayer. We do not close. We collect.
1:19:54 We are archaeologists of distraction.
1:19:56 Curators of half begun obsessions.
1:19:58 Keepers of breadcrumbmed identities.
1:20:02 They call us hoarders like it's a crime.
1:20:05 I call us archist. Archists.
1:20:11 [Music]
1:20:17 I spelled it. No, I spelled it with a K.
1:20:19 I think it it even has a K in it now.
1:20:21 Oh, no. I I replaced it. Hang on.
1:20:25 Um,
1:20:27 I'll do it with a K. R. I'll I'll do CK.
1:20:31 R
1:20:32 CK. Archivists.
1:20:37 We'll see if it does it now. Remix.
1:20:48 I like this.
1:20:51 Yeah, that's how it's spelled. Ryan.
1:20:55 >> Yeah, it's going to keep mispronouncing
1:20:57 it a different way.
1:20:58 >> To the altar of a thousand favor guns.
1:21:00 Each a tiny stained glass promise. Each
1:21:02 tab have a relic, an argument, an
1:21:04 unfinished prayer. We do not close.
1:21:08 [Music]
1:21:21 I bow to the altar of a thousand's
1:21:22 favicons.
1:21:24 Each a tiny stained glass promise, each
1:21:26 tab a relic, an argument, an unfinished
1:21:29 prayer. We do not close, we collect.
1:21:33 We are archaeologists of distraction.
1:21:35 Curators of half begun obsessions,
1:21:37 keepers of breadcrumbmed identities.
1:21:41 They call us hoarders like it's a crime.
1:21:45 I call us archivists of the
1:21:47 incandescent. Now,
1:21:49 >> damn it. Archivists.
1:21:52 Let's go to the last two. Here's where
1:21:54 we changed the spelling of it.
1:21:58 [Music]
1:22:17 A bow to the altar of a thousand
1:22:19 favicons. Each a tiny stained glass
1:22:22 promise. Each tab a relic and argument
1:22:24 and unfinished prayer we do not close.
1:22:26 We collect.
1:22:28 We are archaeologists of distraction,
1:22:31 creators of half-bun obsessions, keepers
1:22:33 of breadcrumbmed identities.
1:22:37 [Music]
1:22:40 Oh, I know what I'm going to do. I call
1:22:42 us keepers.
1:22:44 I call us keepers of the incandescent.
1:22:47 Now, let's see. I call us I call us
1:22:52 um dark keepers
1:22:55 of the incandescent now remix.
1:22:59 That's good. That's better. Dark keepers
1:23:03 archivists or archivists.
1:23:06 Guys, the reason it's it's pronouncing
1:23:08 the words different each time is the way
1:23:10 diffusion models work is they're they're
1:23:14 combining probably thousands of
1:23:17 different
1:23:18 kind of random samples
1:23:21 sample tokens. Like each token is like
1:23:24 this little snippet of like, you know, a
1:23:28 thousand layers deep. And it's so
1:23:30 depending on how it weights those
1:23:31 things, that's why you're getting
1:23:33 different things each time. It's It's
1:23:35 not nearly as predictive as we'd like it
1:23:37 to be. Sea slug. Yeah, the I saw the one
1:23:39 with the eyeball in it, but I just
1:23:41 rewrote the line al together now.
1:23:52 >> Like the harmonic
1:23:53 >> to the altar of a thousand favicons.
1:23:56 Each a tiny stained glass promise. Each
1:23:59 tab a relic, an argument, an unfinished
1:24:01 prayer. We do not close, we collect.
1:24:05 We are archaeologists of distraction,
1:24:08 curators of half-bun obsessions, keepers
1:24:10 of breadcrumbmed identities.
1:24:14 They call us hoarders like it's a crime.
1:24:18 I call us dark keepers of the
1:24:19 incandescent. Now every tab is a shrine
1:24:23 to a thought that refused to die.
1:24:26 A headphone cable tank.
1:24:28 >> Wait, what's that? They are digital
1:24:30 twinning the application. ask me for my
1:24:34 height and weight.
1:24:36 Oh, they want to make you the superstar.
1:24:38 So, should I go to my email? Oh, no. I
1:24:40 should go to my DMs, right?
1:24:44 I don't have a DM from them. Doesn't
1:24:46 look like
1:24:46 >> No, the it was on the type form that I
1:24:49 uh in the next um
1:24:51 >> Oh, in the next thing
1:24:52 >> down, but yeah, it's really interesting.
1:24:54 if you um once I finished the
1:24:57 application, it said um if you retweet,
1:25:00 fill out the application, we're going to
1:25:02 make you the superstar. So like this
1:25:05 person, this chai person is apparently
1:25:08 >> somebody that a likeness that they've
1:25:10 trained off of somebody a real human and
1:25:13 they've made them into an AI superstar.
1:25:16 So Kyle, that could be you floating in
1:25:19 that murky water there. I I I think if
1:25:22 anyone is to be an AI superstar, I think
1:25:25 I think it's me.
1:25:28 Here comes the twerking. Finally.
1:25:34 I just thought that was unique because
1:25:36 it it's taking some real grounded stuff
1:25:41 and converting them into a AI persona.
1:25:47 >> Uh what's your ex? So at
1:25:52 >> Shannon
1:25:53 >> tab tab.
1:25:56 >> Oh, am I not tabbed? Oh.
1:25:59 Oh, because it opens a new tab. Your
1:26:01 other social media. Tick
1:26:04 tock.comarninglab.
1:26:05 [Music]
1:26:11 Uh, what else? That's good.
1:26:15 Okay.
1:26:17 Age. I'm [ __ ] old. I'm old, man.
1:26:24 Share your physical data. Um,
1:26:29 5 foot8 in
1:26:32 200.
1:26:34 I'm a fat [ __ ]
1:26:38 How did you discover Twitter?
1:26:42 Thank you for your application. All
1:26:43 right, cool. There you go. Beautiful.
1:26:46 All right, back to Sunno.
1:26:50 Share this tab instead.
1:26:53 >> So, that one's pretty good. Let's hear
1:26:54 this one.
1:26:57 [Music]
1:27:02 Like this.
1:27:05 [Music]
1:27:10 About to the altar of a thousand
1:27:11 favicons. Each a tiny stained glass
1:27:14 promise. Each tab a relic, an argument,
1:27:16 an unfinished prayer. We do not close,
1:27:18 we collect. We are archaeologists of
1:27:20 distraction, curators of half-bun
1:27:22 obsessions, keepers of breadcrumbmed
1:27:24 identities.
1:27:27 They call us hoarders like it's a crime.
1:27:29 I call us dark keepers of the
1:27:31 incandescent now. Every tab is a shrine
1:27:33 to a thought that refused to die. A
1:27:35 headphone cable tangled with yesterday's
1:27:37 revolution. A recipe we swore we'd
1:27:40 master. A thread we swore we'd answer.
1:27:42 We keep the world in micro windows like
1:27:44 sailors with sextants mapping storms of
1:27:46 curiosity.
1:27:49 Chrome hates us. Chrome groans under our
1:27:52 little empires. Its memory like an
1:27:53 anxious brain on the subway leaking
1:27:55 forgetting dropping passengers on the
1:27:57 line. Memory leak. They say as if sin
1:27:59 were a bug fixable with a patch. Google
1:28:02 with its polished teeth and corporate
1:28:03 himnels. Cannot fathom the sanctity of
1:28:05 our open questions. It trims. It prunes.
1:28:07 It gaslights our tabs into sleeping
1:28:09 bags. But we are stubborn, gentle,
1:28:11 glorious, unruly.
1:28:14 Let the browsers call it in efficiency.
1:28:17 We call it fidelity. We are a million
1:28:19 tiny vows. I will return. I will return.
1:28:22 I will return.
1:28:24 [Music]
1:28:25 Each tab is a bookmark for an argument
1:28:27 with the future. We keep versions of
1:28:29 ourselves in duplicate. The person we
1:28:30 were at 2 a.m. The person who clicked
1:28:32 that headline in a cafe. The person who
1:28:34 loved a sentence. Close one and you risk
1:28:36 killing that small particular grief.
1:28:38 close one in the map of you becomes less
1:28:39 accurate.
1:28:41 [Music]
1:28:43 >> There is an ethics to accumulation. Not
1:28:45 greedy, not manic, but reverent.
1:28:47 >> There's an ethics to tab hoarding.
1:28:52 Not greedy, not manic, but reverent.
1:28:54 Tabs are trust jars. We build the
1:28:56 altars. Meanwhile, the engineers whisper
1:28:58 in fluorescent rooms, wielding heristics
1:29:01 like brooms, saying, "Optimize,
1:29:03 compress, discard."
1:29:05 They they mistake tidiness for
1:29:06 intelligence. They forgot memory is not
1:29:09 just storage, it's testimony. A memory
1:29:11 leak is an in an app is a moral leak in
1:29:15 the machine. That's really good. Um, we
1:29:19 refuse so and so we refuse efficiency
1:29:21 romantic in our rebellion. Okay, I'm
1:29:23 going to end it right there. Um, we
1:29:26 relish the moment. So, let's make this
1:29:29 one its own stanza.
1:29:33 For every crash there's revolution.
1:29:35 If Google cannot hold us, that failing
1:29:39 is ours.
1:29:42 Raise a tab then another. Let's get rid
1:29:46 of that.
1:29:56 We will not let optimization rewrite the
1:29:58 archive of our desire. That's a good
1:30:00 That's a good ending line.
1:30:03 Okay.
1:30:04 So, let's
1:30:09 let's make that a new stanza.
1:30:13 Let's make that a new stanza.
1:30:24 Okay. Remix. So, this is going to be
1:30:27 shorter now.
1:30:36 So, it depends on the genre. It It
1:30:38 matches the the words to the lyrics
1:30:40 quite well in a lot of cases. Um, I'm
1:30:43 purposefully doing spoken word here. So,
1:30:47 spoken word is is generally not very
1:30:50 lyric friendly. I mean, very like it
1:30:53 doesn't match the the percussion behind
1:30:56 it.
1:31:00 [Music]
1:31:04 Wait. I I bow to the altar of a thousand
1:31:07 favicons. That's kind of a cool
1:31:10 starting. Each tiny. We do not close. We
1:31:13 collect. We are archivists of disgra
1:31:15 curators who have begun obsessions. Oh,
1:31:19 we got keepers in here twice. Keepers.
1:31:24 Um, let's see. Um,
1:31:38 we hoard
1:31:41 the tabs
1:31:43 that you
1:31:47 Oh, wait.
1:31:49 We
1:31:55 preserve
1:31:59 the tabs that you
1:32:04 discard.
1:32:06 They call us hoarders like it's they
1:32:08 call us tab hoarders like it's a crime.
1:32:36 Keepers of the incandescent. Now every
1:32:37 tab is a shrine. Okay, remix. All right,
1:32:40 we got some new lyrics here. People,
1:32:42 calm down, everybody. Calm down.
1:32:50 Gonna run it again. Yeah, I'm going to
1:32:51 run it again. Keep running it.
1:32:54 Thank
1:33:03 [Music]
1:33:20 you.
1:33:24 I bow to the altar of a thousand
1:33:26 favicons. Each a tiny stained glass
1:33:29 promise. Each tab a relic and argument
1:33:31 and unfinished prayer. We do not close.
1:33:33 We collect.
1:33:35 We are archaeologists of distraction.
1:33:36 >> Damn it. It's close but not there.
1:33:38 >> Bow to the altar of a thousand favors.
1:33:40 Damn.
1:34:03 [Music]
1:34:07 That sounds like Ani DeFranco.
1:34:10 [Music]
1:34:15 I bow to the altar of a thousand
1:34:17 favicans.
1:34:19 Each a tiny stained glass promise. Each
1:34:21 tabar relic and argument an unfinished
1:34:24 prayer. We do not close. We collect.
1:34:28 We are archaeologists of distraction,
1:34:30 creators of half begun obsessions,
1:34:32 keepers of bread.
1:34:33 >> I like that one a lot actually.
1:34:36 Give that a thumbs up.
1:34:39 Um,
1:34:44 I like that. It's funky.
1:34:49 I bow to the altar of a thousand
1:34:51 favicons. Each a tiny stained glass
1:34:53 promise. Each tab a relic, an argument,
1:34:55 an unfinished prayer. We do not close,
1:34:58 we collect. We are archaeologists of
1:35:01 distraction, curators of half begun
1:35:03 obsessions. We preserve the tabs that
1:35:05 you discard.
1:35:09 They call us tab hoarders like it's a
1:35:10 crime. I call us inspired keepers of the
1:35:13 incandescent now. Every tab is a shrine
1:35:15 to a thought that refused to die. A
1:35:17 headphone cable tangled with yesterday's
1:35:19 revolution. A recipe we swore we'd
1:35:21 master. A thread we swore we'd answer.
1:35:23 We keep the world in micro windows like
1:35:25 sailors with sextants mapping storms of
1:35:27 curiosity. Chrome hates us. Chrome
1:35:29 groans under our little empires. It's
1:35:31 memory like an anxious brain on the
1:35:33 subway. Leaking, forgetting, dropping
1:35:35 passengers on the line.
1:35:38 Memory leak, they say as if sin were a
1:35:40 bug fixable with a patch. Google with
1:35:42 its polished teeth and corporate himnels
1:35:44 cannot fathom the sanctity of our open
1:35:46 questions. It trims, it prunes, it
1:35:48 gaslights our tabs into sleeping bags.
1:35:50 But we are stubborn, gentle, glorious,
1:35:53 unruly.
1:35:55 [Music]
1:35:58 Let the browsers call it inefficiency.
1:36:00 We call it fidelity. We are a million
1:36:02 tiny vows. I will return. I will return.
1:36:05 I will return. Each tab is a bookmark
1:36:08 for an argument with the future. We keep
1:36:10 versions of ourselves and duplicate. The
1:36:12 person we were at 2 a.m. The person who
1:36:14 clicked that headline in a cafe. The
1:36:16 person who loved a sentence. Close one
1:36:18 and you risk killing that small
1:36:19 particular grief. Close one and the map
1:36:21 of you becomes less accurate.
1:36:24 There is an ethics to tab hoarding. Not
1:36:27 greedy, not manic, but reverent.
1:36:30 [Music]
1:36:32 Tabs are trust jars. Open wide.
1:36:34 >> This one's close.
1:36:39 [Music]
1:36:47 I bow to the altar of a thousand favic
1:36:49 kinds. Each a tiny stained glass
1:36:51 promise. Each tabar relic an argument.
1:36:55 An unfinished prayer we do not close. We
1:36:57 collect.
1:36:59 We are archaeologists of distraction.
1:37:02 Curators of half begun obsessions. We
1:37:04 preserve the tabs that you discard.
1:37:08 They call us tab hoarders like it's a
1:37:10 crime. I call us inspired keepers of the
1:37:13 incandescent. Now
1:37:15 every tab is a shrine to a thought that
1:37:17 refused to die. A headphone cable
1:37:19 tangled with yesterday's revolution. A
1:37:22 recipe we swore we'd master. A thread we
1:37:25 swore we'd answer. We keep the world in
1:37:27 micro windows like sailors with sex
1:37:30 mapping storms of curiosity.
1:37:34 >> All right, that's the one. All right, so
1:37:36 now we're going to go. And I know it's
1:37:39 late, so if you got to go, I get it. But
1:37:44 we're going to go turn this into
1:37:45 something [ __ ] cool. Um, midjourney
1:37:49 because
1:37:51 if I don't do it now, I won't come back
1:37:53 to it and then it will be lost.
1:37:57 And us tab hoarders got to stick
1:37:59 together.
1:38:08 Um,
1:38:09 [Music]
1:38:19 let's see.
1:38:24 So, now I'm in mid journey.
1:38:46 Ending frame.
1:39:06 Do I have to do this in draft?
1:39:16 Oops.
1:39:18 Get rid of that there.
1:39:22 Why is this
1:39:25 I'm confused. This happened to me once
1:39:27 before. I'm trying to add a
1:39:30 an image as a style reference
1:39:34 or omni reference as they call it. Oh, I
1:39:36 think it Oh, it's because I'm in draft
1:39:39 mode.
1:39:41 Omni reference. There we go. Okay.
1:39:46 So, let's pop him into Omni Reference.
1:39:50 And then we're going to go
1:39:53 um
1:39:56 hip New York City performer
1:40:00 in a
1:40:03 Let's see. How do we want this? We want
1:40:05 this
1:40:09 vertical 2x3, 3x4,
1:40:12 3x4.
1:40:16 2 by3
1:40:19 3x4 Okay,
1:40:23 in a moody
1:40:27 New York City
1:40:30 club
1:40:33 in a spotlight
1:40:42 sitting on a stool.
1:40:50 looking at the camera
1:40:54 and then we're going to go
1:40:58 we'll go weirdness at I don't know 500
1:41:01 stylization at 400 and then we'll go
1:41:04 dash dash R for repeat space 4
1:41:09 and that's going to give us a whole [ __ ]
1:41:12 pile of these images to play with or to
1:41:15 pickank
1:41:21 dank.
1:41:23 All right. Thoughts, questions? Anybody
1:41:25 have any questions? I just finished four
1:41:27 tracks, lyrics, and music today. Every
1:41:30 vocal key. Wait, every vocal key, music
1:41:34 key, etc. Cool.
1:41:37 Needs more cowbell. Absolutely. Oh,
1:41:40 these are cool.
1:41:48 would love to be able to see him. We're
1:41:50 still looking at some of the lyrics.
1:41:52 >> Damn it.
1:41:54 Sorry about that.
1:41:58 Well,
1:42:07 amic
1:42:09 pentaber. Pentaber. I like it. That That
1:42:13 one's actually pretty cool. That looks
1:42:14 really realistic.
1:42:32 I think that first one's the one.
1:42:40 Yeah, let's let's we're going to upscale
1:42:42 this.
1:42:45 Let's see. Upscale subtle upscale
1:42:48 creative
1:42:54 [Music]
1:42:58 dunk.
1:43:03 Okay. Then we're going to go I'm going
1:43:05 to go over here and I'm going to go to
1:43:07 Hedra. HD R A.AI.
1:43:12 I'm going to log in.
1:43:22 So, we got to go back to Sunno. Is that
1:43:24 thing done? How's that looking?
1:43:32 Whoops.
1:43:39 That looks fake.
1:43:49 But this is one thing I've noticed in
1:43:52 MidJourney is sometimes when you upscale
1:43:54 things, it it gives it more of that
1:43:56 plastic look.
1:44:00 This one's not bad.
1:44:02 He's far enough away. I think I think
1:44:04 this could be okay.
1:44:11 Okay, we're going to download him and
1:44:13 then we're going to go back to Suno.
1:44:16 We're going to download
1:44:19 this one.
1:44:21 We'll download this as an MP3.
1:44:25 And then we're going to also say
1:44:27 generate video
1:44:30 because we'll just upload the video on
1:44:32 its own. And now we're going to go to
1:44:34 Hedra
1:44:38 chat with a Hedra realtime avatar now.
1:44:42 Oh, that's cool.
1:44:45 All right, let's see. Image video
1:44:50 character three. Okay.
1:44:55 All right. Are you all seeing this? No,
1:44:56 because I didn't switch tabs. Now you
1:44:58 are. So now I'm at Hedra.
1:45:02 So I'm gonna upload an audio script,
1:45:07 which is our little song we just made or
1:45:09 a little
1:45:18 [Music]
1:45:21 I bow to the altar of a thousand
1:45:23 favorite kinds. Each a tiny stained
1:45:25 glass promise. Each tab relic.
1:45:31 [Music]
1:45:42 There's our dude.
1:45:48 I think we'll just go. Bang.
1:45:53 Error. Failed to fetch. Try it again.
1:46:04 That failed to fetch thing is probably a
1:46:06 memory leak error, which is what this
1:46:08 spoken word poetry is all about.
1:46:15 It's Dave Stewart. It does look like
1:46:17 Dave Stewart.
1:46:24 tink.
1:46:32 A memory leak error is when you have too
1:46:35 many tabs open and the application can't
1:46:38 manage
1:46:40 the memory from all of the different
1:46:41 tabs
1:46:43 and it just basically runs out of memory
1:46:45 and then craps.
1:46:49 [Laughter]
1:46:54 I like picking on Google. I've been
1:46:57 pissed off at them ever since they got
1:46:59 rid of their let's do no evil as one of
1:47:01 their corporate headlines.
1:47:04 Fuckersink.
1:47:15 All right, that's processing. So, I
1:47:17 don't know what what we can do now. We
1:47:18 can just go chat now.
1:47:22 All right.
1:47:23 So, I don't know. What were the uh what
1:47:27 were the takeaways from tonight?
1:47:31 Go try to make [ __ ] happen that you
1:47:33 think can't happen is not a bad
1:47:35 exercise.
1:47:37 Um,
1:47:40 AI can't pronounce things.
1:47:45 But it's also not that it hasn't been
1:47:47 trained on that word. It hasn't been
1:47:49 trained on any word. It's just that it
1:47:51 doesn't have enough training data to get
1:47:53 that word right, which is a weirdly
1:47:55 different thing. Has the inshitification
1:47:58 of the internet happened to you, too?
1:48:00 Oh, that happened years ago.
1:48:03 You don't know how to use the black bar.
1:48:04 I don't. Um,
1:48:08 it can't match lyrics to music. Uh, it
1:48:11 it it's pretty [ __ ] good at it.
1:48:15 It can write music. Yeah, it can write
1:48:16 music. Have you never played with Sunno?
1:48:20 It can write music. It can write lyrics.
1:48:22 Here, let's While that's processing,
1:48:24 let's go back to Sunno. We'll just go
1:48:27 We'll go create
1:48:29 uh somewhere at the bottom here. I think
1:48:31 I can clear everything, right? Yeah.
1:48:34 clear all.
1:48:44 But it it matches it matches lyrics to
1:48:46 music. It does it pretty good.
1:48:49 Let's do rap. Melada. No, we won't do
1:48:50 rap because that's more spoken word.
1:48:52 We'll do uh we'll do voice cracks. Voice
1:48:56 cracks.
1:48:58 Soaring.
1:49:00 What's that? Russian
1:49:02 thing tab.
1:49:10 lyrics. Write the lyrics. Auto.
1:49:14 Describe your lyrics.
1:49:16 Uh song
1:49:19 about matching lyrics to the music.
1:49:26 We'll get a little meta here. Uh let's
1:49:28 see. Voice crack. Uh, let's do um let's
1:49:32 do uh funk.
1:49:36 Um,
1:49:38 funk driving.
1:49:42 Uh, electric
1:49:45 guitar,
1:49:48 voice cracks,
1:49:53 creative harmonies,
1:49:57 and synopation.
1:50:03 And we'll go style influence. And we'll
1:50:06 go
1:50:08 we'll do weirdless low. We'll do
1:50:16 match the lyrics.
1:50:32 What is syncupation exactly?
1:50:35 Timing with the drums. No, syncopation
1:50:37 is when you're when it's you've got
1:50:41 beats in between the main beats.
1:50:50 [Music]
1:51:02 words are dancing on the page tonight.
1:51:09 [Music]
1:51:10 Tiptoeing around the nose just right.
1:51:15 [Music]
1:51:19 Do they swing or do they crawl
1:51:23 or do they eat?
1:51:24 >> Yeah, not bad, but in the neighborhood.
1:51:27 >> It had some good syncupation. The all
1:51:29 those beats that were off the main
1:51:32 beats, that's syncupation.
1:51:40 [Music]
1:51:48 Yeah, it's like an offbeat. Let's have
1:51:50 ChatchBT define it for us. Um,
1:51:54 define
1:51:57 syncopation simply
1:52:05 dunk. Uh simple definition, putting the
1:52:08 musical accent where you don't expect it
1:52:12 of the week's beats
1:52:14 on the weak beats or the ands between
1:52:17 beats. A quick clap clap example. One
1:52:20 and two and three and four and clap on
1:52:23 the ends. The offbeats. One
1:52:26 one two three four. Yeah, there you go.
1:52:30 That offbeat emphasis equals
1:52:32 syncopation.
1:52:34 That's cool.
1:52:36 [Music]
1:52:46 No, not yet.
1:53:02 Oh, that's good. All right, let's try
1:53:03 that. Uh, let's do
1:53:11 68 time signature.
1:53:36 [Music]
1:53:45 words like a dark.
1:53:49 >> That's pretty cool. Like that beat. All
1:53:51 right, here's the 68 ones.
1:53:54 [Music]
1:53:58 I've got a line that homes in my head.
1:54:02 >> No.
1:54:04 [Music]
1:54:40 Whatever. All right, let's see. Our
1:54:42 video is at 60%. All right, let's let's
1:54:45 make another song. There's a there's a
1:54:48 cool thing we can do on Suno. If you
1:54:50 click on the explore tab, you get this
1:54:53 giant wheel of of uh
1:54:57 style mashups.
1:55:00 Africanner, dreamy acoustic rock,
1:55:05 acid techno, avancgard jazz,
1:55:09 Spanish pop, alt country dubstep,
1:55:13 saxophone, basosanova,
1:55:16 mandarin drums, Egyptian flamnco,
1:55:21 slushwave, new jack swing,
1:55:24 [Music]
1:55:48 [Music]
1:56:01 Let's do that one.
1:56:03 All right. Did it pull the style in?
1:56:08 Did not.
1:56:13 That sucks.
1:56:21 [Music]
1:56:28 Coral Afro jazz.
1:56:35 Coral Afro jazz.
1:56:40 Auto lyrics. Um,
1:56:43 lyrics about machines
1:56:48 being more creative than
1:57:01 we ever expected in our lifetimes.
1:57:11 Create
1:57:13 create.
1:57:19 All right, the video is at 80%. We're
1:57:21 almost done, people.
1:57:25 Shank dank
1:57:27 shank dank.
1:57:31 [Music]
1:57:47 It started with sparks in a silent cold.
1:57:51 Oh,
1:57:52 [Music]
1:58:17 hey.
1:58:21 [Music]
1:58:33 It started with sparks in a mental
1:58:36 brain.
1:58:43 A whisper of thought in a coded vein.
1:58:49 [Music]
1:58:51 Not a soul, not a heart, just wires and
1:58:55 steel. But it
1:58:58 [Applause]
1:59:00 >> was kind of cool.
1:59:02 [Music]
1:59:15 It started with sparks. works in a metal
1:59:20 brain.
1:59:22 A whisper of thought
1:59:26 in a coated vein.
1:59:31 Not a soul, not a heart, just wires and
1:59:35 steel.
1:59:38 But it painted a world
1:59:42 that made us feel
1:59:46 >> the machine alive.
1:59:50 It dreams in life.
1:59:54 Colors we've never seen. Ignite the
1:59:59 night.
2:00:01 It's more than we imagine,
2:00:05 more than we could know.
2:00:08 [Music]
2:00:10 The creative machine steals the show.
2:00:17 It wrote us a story
2:00:21 with no end inside.
2:00:25 A melody spinning
2:00:29 from black and white.
2:00:33 Now, that one's pretty cool. That gives
2:00:35 a thumbs up. All right. Um, let's go
2:00:37 watch our video.
2:00:39 Are we done? Are we done? Oh, we're not
2:00:43 done. I thought we'd be done by now.
2:00:46 Come on.
2:00:49 I just want to see a guy in a club doing
2:00:52 spoken poetry about tab hoarding.
2:00:56 Is that so much to ask?
2:01:05 Ah, what do you think, Pate M? We're
2:01:08 just screwing around tonight. There's
2:01:11 nothing resembling productivity on the
2:01:13 uh learning lab tonight.
2:01:16 If you learn something tonight, that's
2:01:18 on you. I I claim no responsibility for
2:01:22 anything resembling education.
2:01:24 Um, I have a feeling
2:01:29 this is done, but the No, it's not.
2:01:39 You know what? Let's go do something.
2:01:41 While that's doing that,
2:01:44 um,
2:01:48 remix, edit, open an editor.
2:01:53 Oh, I got to switch tabs.
2:01:58 So, I'm in.
2:02:04 Click and drag the song lyrics to select
2:02:06 them in the timeline.
2:02:09 Use new lyrics to enter new lyrics.
2:02:12 Okay.
2:02:16 It started with sparks.
2:02:22 And let's change metal to silicon.
2:02:25 And a silicon. Oh, wait. I've got to
2:02:28 I've got to add new lyrics. Oh, wait.
2:02:37 So, I'm going to copy that. I'm going to
2:02:39 put it here.
2:02:40 And then I'm going to go silicon.
2:02:44 It started with sparks in a silicon
2:02:47 brain.
2:02:50 A whisper of thought in a coated vein. I
2:02:52 like that. Not a soul, not a heart, just
2:02:55 wires.
2:02:57 Here, let's grab all this.
2:03:01 Let's grab all this.
2:03:16 in a coded vein. Not a soul, not a
2:03:19 heart.
2:03:46 Let's see.
2:04:05 electronic zeal. Not a soul, not a
2:04:07 heart. Electronic zeal, but it painted a
2:04:10 word world that made us feel. All right,
2:04:12 let's try that. So, we're going to
2:04:14 replace
2:04:16 those lyrics with those lyrics. Replace
2:04:33 And I think what it does is it'll give
2:04:34 us two alternates.
2:04:38 Our other video is still at 80%. So,
2:04:58 I need coasters. I'm getting a lot of
2:05:00 sweating on my drinks.
2:05:06 All
2:05:19 right.
2:05:22 with sparks
2:05:25 in a silicon brain.
2:05:29 A whisper of thought
2:05:33 in a coated vein,
2:05:36 [Music]
2:05:37 not a soul,
2:05:41 not a heart.
2:05:43 >> Oh, nice.
2:05:47 >> Yeah, I kind of missed that. But it
2:05:49 painted a world that made us
2:05:58 [Music]
2:06:01 >> It started with sparks
2:06:04 in a silicon brain.
2:06:08 A whisper of thought
2:06:12 in a cold
2:06:16 [Music]
2:06:19 It started with sparks
2:06:23 in a silic
2:06:26 >> like that. A whisper of thought
2:06:31 in a coded vein.
2:06:36 Not a soul, not a heart. It electron ex.
2:06:43 But it painted a world
2:06:46 that made us feel
2:06:52 alive.
2:06:53 [Music]
2:06:57 All right. So, that's in there. New new
2:06:58 lyrics are in there. This thing is
2:07:00 really cool. If you haven't played with
2:07:02 this this new um the new editor, you can
2:07:06 move these sections around. You can
2:07:08 repeat them. You can skip them. All
2:07:10 right. So, let's save as a new song. You
2:07:14 can also split it into stems.
2:07:19 Go to song.
2:07:28 The creative machine edit.
2:07:36 Come on. Why you not playing?
2:07:41 Reload.
2:07:53 Why is it not playing?
2:07:56 Um, create.
2:08:04 [Music]
2:08:17 It started with sparks
2:08:21 in a silicon brain.
2:08:25 A whisper of thought
2:08:29 in a coded vein.
2:08:33 Not a soul, not a heart. It had trauma,
2:08:38 [Music]
2:08:40 but it painted a world
2:08:43 that made us feel
2:08:48 the machines alive.
2:08:52 It dreams in light.
2:08:57 Colors we've never seen. Ignite the
2:09:01 night.
2:09:03 It's more than we imagine.
2:09:07 more than we could know.
2:09:13 The creative machine
2:09:19 that that that weird thing where it goes
2:09:22 up like that, that's like a pseudo tick.
2:09:25 I've noticed that it it'll do these
2:09:27 really unexpected jumps. That's
2:09:30 something I would go edit. I I would not
2:09:33 change the lyrics there, but I would
2:09:35 change the the music there. Fascinating.
2:09:40 All right, we have a video.
2:09:42 Let's go look at our video.
2:09:46 Uh download.
2:09:52 All right, here we go. We're going to
2:09:54 open this here. I'm going to change my
2:09:56 sharing
2:09:59 to
2:10:07 [Music]
2:10:13 I hope this doesn't suck. Uhoh. Where'd
2:10:15 that go?
2:10:20 [Music]
2:10:28 I bow to the altar of a thousand favor
2:10:30 kinds. Each a tiny stained glass
2:10:33 promise. Each tab relic an argument. An
2:10:37 unfinished prayer. We do not close. We
2:10:39 collect.
2:10:41 We are archaeologists of distraction.
2:10:43 Curators of half begun obsessions. We
2:10:45 preserve the tabs that you discard.
2:10:49 They call us tab hoarders like it's a
2:10:51 crime. I call us inspired keepers of the
2:10:54 incandescent. Now
2:10:56 every tab is a shrine to a thought that
2:10:59 refused to die. A headphone cable
2:11:01 tangled with yesterday's revolution. A
2:11:04 recipe we swore we'd master. A thread we
2:11:06 swore we'd answer. We keep the world in
2:11:09 micro windows like sailors with sex
2:11:12 mapping storms of curiosity.
2:11:16 Chrome hates us. Drone groans under our
2:11:18 little empires. It's memory like an
2:11:20 anxious brain on the subway. Leaking,
2:11:22 forgetting, dropping passengers on the
2:11:25 line.
2:11:27 Memory leak, they say as if sin were a
2:11:30 bug fixable with a patch.
2:11:32 Google with its polished teeth and
2:11:35 corporate himnels cannot fathom the
2:11:37 sanctity of our open questions. It
2:11:39 trims, it prunes, it gaslights our tabs
2:11:42 into sleeping bags. But we are stubborn,
2:11:46 gentle, glorious, unruly.
2:11:49 [Music]
2:11:53 Let the browsers call it in efficiency.
2:11:55 We call it fidelity. We are a million
2:11:58 tiny vows. I will return. I will return.
2:12:01 I will return. Each tab is a bookmark.
2:12:07 All right, we're going to publish this.
2:12:13 I'm such an idiot.
2:12:17 [Laughter]
2:12:20 All right, here we go.
2:12:24 Oh my god, y'all are [ __ ] weird.
2:12:26 You're still here. How many people? 16.
2:12:29 Yeah, I knew we blew blew a bunch out
2:12:31 tonight.
2:12:37 All right. People like, "Fuck it. I'm
2:12:40 going to bed.
2:12:42 He's not going to talk to us." All
2:12:44 right. There's that. Um
2:14:14 [Laughter]
2:14:22 Tabo orders unite. A spoken word plea
2:14:24 for Google to fix the goo to fix Google
2:14:27 Chrome to be more friendly to those of
2:14:30 us who generously retain tabs over days,
2:14:33 weeks, and even months. We're not the
2:14:35 problem. We're the solution.
2:14:43 and apps. What's apps?
2:14:48 What's apps?
2:14:50 >> Google apps. No, the different apps that
2:14:53 you've used like you for your
2:14:55 >> CHGPO
2:15:03 at chatgpt
2:15:09 at OpenAI
2:15:14 at Hedra.
2:15:20 Um,
2:15:24 I think that's it.
2:15:26 We'll do scobilizer. See if we can get a
2:15:28 little publicity out of this.
2:16:20 Your memory leaks
2:16:25 are killing us.
2:16:28 We're not the problem. We're the
2:16:30 solution.
2:16:36 All right. So, everybody, as as as I
2:16:39 always request, did this upload? No, of
2:16:42 course not.
2:16:44 Um, well, we'll try to post it. See if
2:16:46 it uploads.
2:16:55 1% 0% 1%. This is not going to upload.
2:17:00 Oh my god.
2:17:03 Elon,
2:17:05 [ __ ] fix your video upload.
2:17:08 God damn it.
2:17:18 All right, let me go. We're already in
2:17:20 QuickTime here, so let me export this as
2:17:22 780p or 720p.
2:17:28 Um,
2:17:30 let's have hoarders
2:17:32 unite
2:17:45 error. Close that. We'll get rid of
2:17:47 this.
2:17:49 We'll re-upload it.
2:18:12 Come on.
2:18:24 post.
2:18:31 The [ __ ] is intended to be broken. Well,
2:18:34 it is. And it pisses me off.
2:18:38 It's like, you know,
2:18:45 you aren't supposed to be able to post
2:18:47 what you want. I know. I know. The free
2:18:51 speech platform doesn't allow you to
2:18:52 post what you want.
2:19:00 All right. Well, this will be up there
2:19:03 at some point. 8% 13% looks like it's
2:19:05 going. It looks like we might have snuck
2:19:09 a a socket in there. Oh, back down to
2:19:12 30%. Come on, 31. You can do it. Oh. Oh,
2:19:16 no. God damn it.
2:19:30 inspo for your next music video. Elon,
2:19:33 fix your [ __ ] upload.
2:19:37 Hey, you can send a rocket to space and
2:19:39 catch it with a pair of chopsticks, but
2:19:41 could you fix your [ __ ] upload
2:19:44 service on Twitter, please? You're the
2:19:47 one claiming this is going to be a video
2:19:49 forward site.
2:19:50 Um, one of the things that's required
2:19:53 for that to be true is to be able to
2:19:55 upload video.
2:19:58 Elon,
2:20:00 anybody wants to trim that little part
2:20:02 of the video out and send it to him? I'm
2:20:04 fine with that.
2:20:07 Someone want to do a little clipping
2:20:09 tonight? Good luck uploading it.
2:20:18 Oh my god. Let me try uploading it. Let
2:20:20 me let me pop it in there again. See if
2:20:23 that does something.
2:20:26 It's just so [ __ ] random when it
2:20:28 works or not. Yeah, it's not going to
2:20:30 work. It goes up, then comes down, goes
2:20:33 up, goes down, goes up, goes down.
2:20:37 Does Does anyone have any technical
2:20:39 chops that understands why this is
2:20:41 happening? In the past, you've copied
2:20:44 and refreshed. Oh, yeah, that's right.
2:20:46 Okay, so let me try that. So, I'm going
2:20:48 to copy all this.
2:20:52 I'm going to refresh the browser.
2:20:57 I'm going to paste a new thing in here.
2:21:01 And then I'm going to put
2:21:05 the video in.
2:21:08 It could be some local Actually, you
2:21:11 know what it could be? It could be local
2:21:13 memory issues with Chrome. Yeah. See, it
2:21:16 didn't go.
2:21:18 It didn't go at all that time
2:21:28 post.
2:21:38 All right, listen. I will do this
2:21:40 afterwards. I don't know what's going
2:21:41 on, but it's driving me [ __ ] crazy.
2:21:44 Let me try let me try the other format
2:21:46 one more time. We'll try one more time.
2:21:50 Downloads this
2:21:55 0%.
2:22:10 Yeah.
2:22:12 Post. All right. It's not going to go.
2:22:15 All right, people. Uh, hope you had fun
2:22:17 tonight. Tomorrow night's regular time.
2:22:19 Uh, there's no salon tomorrow night. Um,
2:22:25 we'll see if
2:22:28 this week it looks like we're going to
2:22:29 get some new stuff from Google that is
2:22:31 worth playing with it. It looks like
2:22:33 this nano banana
2:22:36 um image model. Um people are saying it
2:22:40 is actually the next big thing for
2:22:42 images anyway. Um so we'll see. We'll
2:22:45 play with that. And then it looks like
2:22:46 they're going to launch something else
2:22:47 whether it's Gemini 3 or some new fancy
2:22:50 version of 2.5 2.6 something like that.
2:22:54 It'll be interesting to see. Um
2:22:57 but yeah, I'll see you tomorrow. All
2:23:00 right. Peace out everybody.