← All Episodes
AI for Humans
Spotify Goes AI. People Will Be Furious. Plus, OpenAI Cracks An 80-Year Math Problem.
Thanks to HP & Intel for sponsoring us! More on the Zbook Fury https://bit.ly/4uapNHs Spotify just made AI music official. They cut a deal with Universal Music Group for AI-enhanced and remixed tracks and we all know where this goes. This week on AI For Humans, Spotify officially crossed the line. T
Show Notes
Thanks to HP & Intel for sponsoring us! More on the Zbook Fury https://bit.ly/4uapNHs
Spotify just made AI music official. They cut a deal with Universal Music Group for AI-enhanced and remixed tracks and we all know where this goes.
This week on AI For Humans, Spotify officially crossed the line. They announced a sweeping deal with record labels, distributors, and music publishers to bring AI music tools to the platform, complete with fan-remix features that use real artist voices and tracks. They also rolled out AI-generated personal podcasts and launched Studio by Spotify, a standalone AI creation app. Gavin and Kevin break down why this changes everything for music, who actually signs up, and why the fan backlash is just beginning.
Meanwhile, a Higgsfield AI feature film just premiered at Cannes for $500,000 total cost, but $400,000 of that was pure AI compute, raising big questions about what the new economics of filmmaking actually look like.
Plus, an unreleased OpenAI general LLM model disproved a central conjecture in discrete geometry that had stood for 80 years, the best new Google Omni prompts and practices and a humanoid robot fail that traveled the entire internet
IS IT TIME TO PANIC? NO, SAYS THESE TWO MORONS.
// Show Links //
Spotify's Official Announcement: Artist-First AI Music Collaboration
https://newsroom.spotify.com/2025-10-16/artist-first-ai-music-spotify-collaboration/
Billboard Coverage Of The Spotify AI Music Deal
https://x.com/billboard/status/2057479053649600917?s=20
Spotify's AI-Generated Personal Podcasts (Hollywood Reporter)
Higgsfield Cannes Film: $500K To Make, $400K Was AI Compute
The Impact Of AI In Four Charts
https://x.com/marcportermagee/status/2057000000000000000?s=20
Google Omni Horseback Riding Prompt
https://x.com/AIWarper/status/2057489859615605244?s=20
FoFR Animals With Google Omni
https://x.com/fofrAI/status/2057281646270026097
FoFR Rooster Dog
https://x.com/fofrAI/status/2057290425170616754?s=20
Clean Up The Alley Prompt
https://x.com/nmatares/status/2057143283357569280?s=20
Ogre Walks On Stage
https://x.com/maxescu/status/2057169603001004400?s=20
US Gov Poster Situation
https://x.com/usedgov/status/2056839488152670378?s=20
OpenAI Model Disproves Central Conjecture In Discrete Geometry
https://openai.com/index/model-disproves-discrete-geometry-conjecture/
Noam Brown's Post About The OpenAI Breakthrough
https://x.com/polynoamial/status/2057178198228586824?s=20
The Humanoid Robot Fail That Traveled The Internet
https://x.com/adamcurtisbroll/status/2057050384166764826?s=20
Ben Relles Launches Make Believe (Hollywood Reporter)
Transcript
aiforhumansspotifyaiaudioopenai
===
Kevin Pereira: [00:00:00] Fun fact, not a single person hates AI-generated music. And this is something Spotify knows, which is why they've done a deal to bring AI to their platform.
Gavin Purcell: You'll be able to use certain artists' voices and tracks to create fan remixes. But who will actually sign up for this is a very big question.
Kevin Pereira: Meanwhile, Google Omni, Seedance, and other AI video generators have enabled a $500,000 feature film to premiere at Cannes.
Gavin Purcell: But the crazy thing is that $400,000 of that is just compute costs.
Kevin Pereira: Ugh. All right, fine. I guess traditional media can die. At least AI isn't messing with my sweet foundational mathematics.
Gavin Purcell: Oh, great news, Kevin. OpenAI just solved an 80-year-old mathematic problem with just AI. Uh,
Kevin Pereira: you know what? They're never gonna solve the Riemann hypothesis. That one's hard.
Gavin Purcell: Kevin, that's not what it's called, but this is AI for humans.
Kevin Pereira: What have I been looking up?
Gavin Purcell: Nothing good, Kevin. Nothing good.
Kevin Pereira: We need deeper research.
Gavin Purcell: No, we don't. No, we don't.[00:01:00]
Welcome, everybody. This is AI for Humans, your twice-a-week guide to the wonderful world of AI. And Kevin, today we have an unusual story coming out of Spotify of all places. Spotify is making the move. They are pushing forward the AI media conversation- Oh ... significantly. Oh. Yes. How are they? Oh, we haven't, we haven't done a deep dive- Oh
on AI music for a while. Oh,
Kevin Pereira: maybe-
Gavin Purcell: Oh ...
Kevin Pereira: oh, did, did someone discover the AI for Humans podcast from, like, two and a half years ago?
Gavin Purcell: They might have. They might have also seen the growth in Suno, the most popular AI music platform. But let's just get very quickly into the deal. We know that there is a deal that was done with Spotify and UMG, and if you remember, if you've been listening to the show for a while, UMG is the company that got a significant portion of Udio, which was Suno's original kind of, like, competitor, and now is a label half-owned AI music platform.
But Kevin, the deal here is, the biggest deal, it's not out yet, but they are going to allow fan-made remixes of artists' songs. Yeah. And I can imagine there is a lot of people in the world of music who are pretty mad about this. There's probably a lot of people in [00:02:00] the AI space who are interested in this, but in general, I think it's a pretty cool thing.
Kevin Pereira: Uh, we've w- we've been new, as those children say, Gavin. We, uh, uh, uh, we identified that there would be a market for this, that there would be incredible interest for this. You and I were, were making songs on our own- Yep ... using authorized and otherwise tools. We're certainly not alone in that. There are people that have entire accounts and brands dedicated to this stuff, so we knew it was coming.
Certainly someone is sharpening a pitchfork somewhere right now. Yes. At least it's opt-in, so artists literally have to express their interest in being involved with the, the, the program. What I, um, struggle with is that it mentions, like, this could be a great revenue stream for the artist, and of course the labels and the distributors- Sure
which will, um, get a piece as well. But it doesn't mention anything for the creator just yet, other than the creator might have to subscribe to use that tool on top of their Spotify Premium. That-
Gavin Purcell: Yeah, so
Kevin Pereira: that's, that's- ... seems interesting. [00:03:00]
Gavin Purcell: Well, so I think this is a really interesting pathway into what AI media is going to be at large, and maybe we can talk about that, kind of, larger conversation in a second.
We should just very quickly say there are four principles that the Spotify group has said that they're going to, kind of, abide by, and this is not live yet. We'll get into a couple other announcements Spotify made with AI in a second. But the four principles are partnerships with labels, distributors, and publishers.
So this is a business deal, this is something they're doing- Mm-hmm ... as a real business. Choice in participation, which I think we need to spend some time on, which means that the artist themselves has to say, "Check box, I agree to this." Fair compensation and new revenue, and then finally an artist-fan connection, which is the idea that maybe this would drive conne- a closer connection to an artist.
But Kev, I do wanna talk about that artist participation part because- Yeah ... I think this is a really big conversation right now, and we've also talked about the general pushback on AI at large, right? There's a ton of stories out there that say people are not fans of AI in general right now. My question to you is: Do you think actual, like, [00:04:00] living artists, 'cause I think there will be a lot of like, you know, Elvis Presley's estate allows this, but- Yeah
do you think there will be a lot of living artists that agree to do this?
Kevin Pereira: 100%. I absolutely do. Interesting. I think there will be some that, that are violently opposed to it, and some will, um, make very loud claims, uh, that they're against it and, and try to, you know, rile their fan bases accordingly. But I think there were artists, as we know, very prominent ones, that were against digital downloads and streaming.
Yes. Right? Yes. And then, and maybe some of that was because, um, the profit sharing or revenue sharing was nonexistent, let alone not in their favor at the time. Yeah. Like, okay, that I get. Slightly different argument. But there's always been a push and pull with technology. There were artists that would, that said, "I will never allow digital instruments on my records," or, "I'll never sample because that's theft," or whatever.
So, like, over time I think more will adopt it. Um, uh, your thoughts?
Gavin Purcell: Well, I think in general- You're gonna see a lot of, uh, what do they call it? Um, uh, grandstanding in some ways. I, and I don't necessarily- Yeah ... think that's a bad thing. I think a lot of artists are just gonna be like, "My fans are [00:05:00] gonna hate this.
I'm never gonna do this. It seems like a, a mistake for me to do." I think we're in that place right now where it could be, like, a massive PR flub for the first person to do it. Oh, for sure. I do think people will do it. I think, the thing I keep coming back to with AI music involves the '90s and the sampling culture.
And I think we've talked about this- Yeah ... on the show a little bit. But I was a giant fan of '90s sampling music. I grew up in the era where, like, hip hop could basically sample everything. Uh, um, one of my favorite hip hop groups of all time is De La Soul, whose music wasn't allowed to be online for a long time because some of the samples weren't cleared, and it's just the amount of money it took to clear a sample was crazy.
In fact, my- Yeah ... favorite story in that world is that, you know, uh, P. Diddy, who now has gone on to an entirely different life i-i-in a different place. But he gave all of the music rights up to Sting for one of his hits because it sampled a big chunk of his song. So- Yeah ... sampling was originally this thing where everybody was mad about it, and then it ended up paying a lot of artists, especially older artists, uh, whose music was sampled for, like, old jazz artists and stuff.
So I do think [00:06:00] there's a world where we can work through this thing. I don't love the label aspect of this, only because, you know, the music industry is famously in fact, going back to old hip hop- Predatory? ... it's been a thing forever that it's predatory. So I'm not sure- Yeah ... that I love the labels doing this, but it does feel like an evolution of what we've been talking about.
Kevin Pereira: Uh, one thing that you said there which is interesting is that, uh, I, I agree with the grandstanding. That's what I mean about artists, like, planting a flag and screaming- Yeah ... from the mountaintops, "Never with my-" Yeah ... "likeness or rights." And I do, but obviously the-
Gavin Purcell: Chaperone I probably guess, Chaperone will not allow this.
People like that who have very active fan bases who I think would get very mad about this stuff.
Kevin Pereira: Well, this is what I think is interesting, is that, uh, certainly there's gonna be some artists, capital A artists, that say- Yeah ... "No, I don't want that." I think the, I think I don't wanna say a majority, but I, there's going to be a large portion of artists that might be really interested and curious in this, but won't do it because of the fans- Yes
explicitly. Yes. Like, they might actually want that, but they're so concerned about the backlash from the fans, which I get. Like, I've been [00:07:00] showing up to commencement speeches, just randomly going from college to college to boo- Is that right?
Gavin Purcell: Oh, lucky you ...
Kevin Pereira: any time there's an AI mentioned. Yeah, it's just, it's fun.
It's fun to be toxic sometimes. But you talk about, like, this could be, again, with the, with maybe with the labels being slightly less greedy somehow in the future. Mm. This could be a big, uh, revenue- Yeah ... opportunity for artists, right? A, a chance for fans to connect more. I wish the fans got a chance to get a taste of that, which is something- Yeah
that I've said that they should do, especially if the fans have to pay for the privilege to remix, which, okay. But not everybody is happy about AI music when it comes to monetization. In fact, Suno is being sued right now, Gavin, by Poseidon Wave Media, which is- Okay ... the entity that, that is behind an indie band that I really love, uh, called The American Dollar.
And they're claiming that they nearly eliminated their licensing revenue. Suno- Wow.
Gavin Purcell: Interesting ...
Kevin Pereira: basically crushed it. They're saying that, that, um, um, let's see. The, the full lawsuit basically says, and this is a, an article out of [00:08:00] musicbusinessworldwide.web, but, um- Oh, love that.
Gavin Purcell: That's my favorite website of all time.
I go to it every
Kevin Pereira: day. It's just so fun to say. It rolls off the tongue. It's how I do vocal warmups for our podcast. But when I go to musicbusinessworldwidewe- web.net Uh, the full thing, it covers 236 sound recordings and compositions across 164 US copyright registrations. They're basically saying that since Suno ingested their copywritten tracks, 80% of the, the duo's licensing revenue has been wiped out.
Like, it's- I- It's a night and day thing.
Gavin Purcell: So I have a question about that, which is, and I think this is gonna be a big conversation across all these things- Sure ... going forward, is that, like, there then that that lawsuit essentially i- is about the training data that went in, and that because- Right ... their training data went in, that they somehow have lost revenue.
I guess the question is, like, the hardest thing for these things always to prove is does the actual output sound like their music, or is it just because there's this kind of flood of new stuff- Right ... that sounds close enough to their music that they use it? That's, I mean, [00:09:00] this is where the, it gets very-
Kevin Pereira: Super valid question.
Gavin Purcell: Yeah, yeah.
Kevin Pereira: Super valid question. Like, I, I have used, personally, I've used ElevenLabs music and sound effects generators- Sure ... instead of going to, like, certain libraries because- Oh, for sure ... there's a specific thing that I want. Yeah. So I know that, you know, it's not because, uh, well, maybe it is because some of my favorite sound libraries have been swept into the ElevenLabs machine.
I don't know. I don't know, um, how dirty the data set with Suno is. Right. But it does say, the complaint claims that, um, outputs replicated the rhythmic structure, production, and delay-based temporal architecture of- Okay ... the original. And by the way, if you're like an American dollar nerd- Yes ... like you understand how that describes their music.
I know what I'm saying. Yeah, yeah. I am one of those guys. Yeah. But it says queries referencing Anything You Synthesize, which is a super popular song by the group, everybody should listen to it, um, they, it said it produced, uh, uh, recordings with titles that are near it, and it shared- Okay ... the same structure.
So-
Gavin Purcell: Interesting ... it, y- it's,
Kevin Pereira: it's, yeah- Was it in their secret sauce? I guess we're gonna find out through the lawsuit.
Gavin Purcell: Well, what's so funny to me is we, we, we've talked about these [00:10:00] companies and the lawsuits, and, like, they do seem to, like, some of these are bubbling up now. And I think that now- Mm-hmm
Spotify's doing this deal, this is gonna come to the forefront. A couple of quick things about Spotify's announcements this morning. One, they've also announced personal podcasts. So Kevin, they're coming after us. Yeah. They're coming after us. If you're listening to us, just know- Bring it ... we're here- You can't-
humans for you ... you
Kevin Pereira: can't
Gavin Purcell: replicate this slop. You can't get us. You can't get us personally. Listen here, you mother .
Kevin Pereira: You don't
Gavin Purcell: have- Oh, geez ... the balls. Hey, that's our audience, Kevin. Okay? We don't call them,
Kevin Pereira: call them that in the episode. You can build... No, I'm talk- Spotify. No one at Spotify. Oh, okay, yeah.
Unless Spotify, unless you wanna, like, sponsor this podcast and put us up next to Rogan, which we- Of course, we're always
Gavin Purcell: open
Kevin Pereira: I take it back. We're always open I'll help build a data center so that- Let's- ... you can rip this off.
Gavin Purcell: Let's tell people with personal podcasts is this idea that you could collect things that you wanna hear about, right?
Whether it's an article or stuff like this. Other companies- Yeah ... have been doing this. NotebookLM is kinda doing this. But this is Spotify putting it in your feeds. That's a bigger deal. This also comes out of Studio for Spotify, which is going to be a standalone app which will enable a lot of these AI features, and it's an add-on, so it is not a free add-on.
It's gonna be part of the actual [00:11:00] cost- Mm ... addition to your Spotify- Mm ... ticket. So- Mwah ... another-
Kevin Pereira: Mwah.
Gavin Purcell: Yes, everything sounds good- Oh, I love it ... to you. It sounds great.
Kevin Pereira: Oh, I love it. I can't wait till Spotify builds me custom recipes and helps me automate- Yes ... my smart home. I ca- hate this
Gavin Purcell: Spotify, the
Kevin Pereira: everything app.
They're gonna start
Gavin Purcell: charging...
Kevin Pereira: That's what I'm saying, like enough, enough. Yeah. I don't need direct payments. I don't need a Spotify avatar- Yes ... to wear the latest Spotify threads to give me Spotify achievements. Yes. Enough. Like, they're, uh, um, they already charge too much. When all this slop starts charging $15 more so I can get an extra NotebookLM podcast about whatever, I can't wait.
Give me a reason to like Apple Music. Oh, yeah. Plus that disco ball icon?
Gavin Purcell: Hey, wait, wait. We've
Kevin Pereira: heard
Gavin Purcell: about them as a sponsor. They need to roll. Relax, relax, relax, relax. Remember-
Kevin Pereira: Oh ...
Gavin Purcell: Spotify might eventually wanna answer a, uh- Yeah ... might eventually wanna sponsor us. You know what? So keep that in mind.
Kevin Pereira: That's, that's valid.
Uh, you know what? Their AI DJ is great.
Gavin Purcell: Great. Fantastic. I, and, and the jam session, amazing. Okay, real quick before we move on, there's another big story coming up, but, um, I do [00:12:00] wanna share this chart that's really interesting, and it was from, shared from, uh, Mark Porter McGee, who I think probably- Mark ... stole it from somewhere.
We'll try, we'll try to see if we can find it somewhere. This just shows the impact of AI, it says, in four charts, and the four charts are more books, more self-represented lawsuits, more music- Hm ... more scientific pro- papers. What it basically shows is that after ChatGPT and after this sort of, uh, explosion of AI stuff, that we're just getting more of everything.
We've talked about this before. A huge part of the future of media is going to be more, and how do you kinda discern- Yeah ... what comes to you? Right. And will people care this stuff is AI generated or not? And I think that's a big thing.
Kevin Pereira: And, and more doesn't necessarily mean better.
Gavin Purcell: No, it can mean way
Kevin Pereira: worse.
By, by any stretch. But also- Yeah, I think it's proving way worse- Yeah ... with, like, books and whatnot. Yeah. But it'd be interesting. Can we also get a chart where we, um, see where, uh, Brian Johnson's, like, nocturnal erections- ... have grown? I'm curious if the release of ChatGPT- You think it's around ChatGPT as well?
has affected all... Look, data is everywhere, Gavin. I bet we can get it. It's true. I bet we can get it.
Gavin Purcell: [00:13:00] Another thing that kinda points to this kind of crazy media space is this announcement from Higgsfield. Now, Higgsfield we've talked about on this show for a while. They're a company that provides a lot of AI models.
They have brought a full-length movie to Cannes this year, and if you're not familiar, Cannes is a big film festival. They brought a movie that is called, uh, what's-
Kevin Pereira: Hellgrind, Gavin
Gavin Purcell: Yeah,
Kevin Pereira: okay. Hellgrind is- Yes ... a 95-minute fully AI generated film
Gavin Purcell: Yeah, so the interesting thing here is- I thought there would be a cooler
Kevin Pereira: description of it in the title of The Wall Street Journal article, but there's not
Gavin Purcell: so Hell grind is this film that, like, Higgsfield kind of, like, organized and put together with a bunch of AI artists and they brought this thing. It's 95 minutes. It's all generated by AI. The interesting thing here is that Higgsfield is very good at making PR, and that's clearly what this is. The Wall Street Journal wrote a long piece about it.
But of that $500,000, $400,000 was on compute, so that's the cost to pay out to, like, say, Seedance or to, to an Omni model or, you know, all this other stuff. Kevin, I think that's really interesting to think about what that means. If this is a feature length movie, and say, I don't think we're... I'm... No one knows if this is good or not.
[00:14:00] I, we haven't seen it. But, like, imagine a world where you can start making feature length movies for $500,000. Now, uh, there's all sorts of, like, guild and all that other sort of issues, but we're also gonna get a lot more movies, right? We're gonna get a lot more movies, and a lot of probably bad movies.
But as you and I were both raised on the kind of direct-to-video or the VHS model where there's all these movies- Yeah ... that, like, maybe didn't do well in the movie theater, but you saw them on TV or you see them in, in a movie store. It will be really interesting to see what movies get made in this world and, like, which ones people will respond to, I think.
Kevin Pereira: Yeah, I, I, I mean, are you gonna watch it? I mean, that's
Gavin Purcell: the- Well, I'll definitely watch it I- I'm so curious The question is, will I pay for it? That's the question, right? Like, and I think that's the big question everybody's gonna have is, would I actually pay to go get a ticket for this? Would I... Would s- would Netflix or any of the streaming services pay for it?
Because I think that's where- Right ... the question will really be. Sure, I'll... I mean, by the way, I'll watch it as a curious person who's interested in AI stuff, but, like, if it's good, I think I would watch it. I mean, you know me, I have very low toler- I have a very low tolerance [00:15:00] for AI slop at large. Like, I enjoy watching sloppy stuff.
But if this is good and it's interesting, then of course I think people will watch it, but again, it goes back to that thing of, like, who's paying for it? Who's making the money? These are the kind of, like, swirling questions around AI media that I don't think we have solved or are we even coming kind of close to solving yet.
Kevin Pereira: I'm gonna send you this here, Gavin. Because, h- you know, when people talk- Yes ... about AI slop- Yes, yes. ... and how the machines will never- Make- ... human ingenuity- Yes ... and it's just, it's, it's nonsense, and we would never have that. Do you wanna, do you wanna sh- share with the audience the link that you just got sent by?
Gavin Purcell: Kevin just sent me the link for the movie that's called Balls Up, and it stars Mark Wahlberg, you may have heard about this, on Amazon. The tagline for this is, "In this raunchy, over-the-top comedy, marketing executives Brad, Mark Wahlberg, and Elijah, Paul Walter Hauser, go balls out and pitch a full coverage condom sponsorship with the World Cup."
So yes, Kevin, this is kind of a sloppy movie on its own for, no pun intended there, but, like, it is definitely something that feels like it could have come out [00:16:00] of Charlie Curran's, uh, uh, brain. I- Which we've seen him do those amazing AI videos. But here's the thing, like,
Kevin Pereira: e- e- yeah, exactly. Exactly. You could have made the trailer for this with Mark Wahlberg, Sacha Bar- Borat is in the movie, Gavin.
Yes, Borat. You could have released it on X, you could have seen the way people recoiled, and you could have saved a lot of time and money.
Gavin Purcell: So there you go, AI slop is, uh, alive, and regular slop is also alive. I do wanna quickly shout out before we, uh, move on from AI video stuff, the Omni Model, the Google Omni Model, which we covered on- Yeah
the show on Wednesday, is fantastic. It's super fun. There's been a lot of people who've been kind of up and down on it. I will say, personally, I don't think it tops the C Dance 2 model just from a pure video model. But Kevin, the coolest thing I've continued to do with this, and people are out there are doing with it, is editing videos.
And I do just wanna share a couple really interesting examples of the editing prowess of what this can do. Uh, AI Warper, who's been doing some incredible videos lately, basically added his avatar or her, her avatar, I'm not really sure what, uh, where AI, uh, Warper puts themselves on these, on the spectrum but, [00:17:00] uh, there's a female avatar, and you see the female avatar riding a horse that they added directly to this video of a horse riding around, which is very cool.
Yeah. Ethel FR, who works at Google now, continues to do interesting videos with this. He made a panda dog, which I really loved, and a bunch of other animals. Are,
Kevin Pereira: did you see, I don't know if you saw rooster, uh, rooster dog? Um,
Gavin Purcell: oh no, I don't know if I did see Rooster Dog ...
Kevin Pereira: so if you haven't seen Rooster Dog, it's worth a special shout-out.
The way that it gets the tail wag and motion of the dog walk combined with the rooster's- ... sort of jutting head movements, that's, uh, that's, that's actually pretty impressive.
Gavin Purcell: That is amazing. That's im- amazing to see. A couple other quick ones here. Nick Patrice actually showed off what it can do in terms of, like, just taking stuff out of a scene.
He cleaned up an alley, which was really a cool thing to see because basically he got rid of a couple pallets and a bunch of garbage in the alley, and this is- Mm-hmm ... a really useful thing if you're gonna be doing edits, if you're gonna ... Even, by the way, if you're not generating all of it with AI video, you can imagine a world where a quick shot replacement thing [00:18:00] is a really big deal here, like, for normal- Yeah
video editors. The final one I wanna shout is this, is this video from Alex Petroski, which is kind of a follow-up on some of the stuff I saw the Google team sharing, where you're actually adding something to a real-life scene, and this is four people on stage at Google I/O, and an ogre walks out behind, and just seeing the shadows that the ogre casts on the wall.
Yeah. You can just get a sense of, like, we are very close to this stuff feeling like things you could see in a movie or you could watch a whole movie of, so it's a pretty big deal.
Kevin Pereira: 100%, and then make it, you know, near real time, 'cause that's going to come, right? Yeah. Then put it in glasses, and now suddenly the magic leap demos that we loved- Yeah
what, 15 years ago- They actually are real. Yeah ... of they could actually be real. So I think that's, that's really impressive. But on the other end of the impressive spectrum, Gavin, we've got the US Department of Education. Did you see this poster?
Gavin Purcell: I saw this. Yes. I don't know why the government, with all of its ability to do things, doesn't at least kinda turn up the knob on the thinking mode a little bit, but this is a poster-
that was released as [00:19:00] a t- way to get people to kind of try to do plumbing jobs, and it's the worst possible version of what plumbers both look like but also what they're working on. It's so bad. Like this, I would not hire these plumbers because they have put too many pipes on this sink. There's way too many things going on.
Kevin Pereira: Yes. All the tools are completely broken. I mean, at least they have an appropriate number of fingers, I suppose, but outside of that- Yeah ... everything is just pure slop. That is ... It was gross and embarrassing, Gavin. That's all.
Gavin Purcell: You know what? It's also okay to try a bunch of different things, and one of the greatest ways, Kevin, to try a bunch of different things, especially locally, is to have a great computer that can do all sorts of stuff locally for you, just like this.
Most laptops make you choose. You can edit video, you can run an AI agent, or you can have 40 tabs open. The HP ZBook Fury has a completely different idea, which is, why are you choosing at all? Just do everything. Huge thanks to HP and Intel for sponsoring AI for Humans this week and lending us this absurdly powerful machine.
Look how big it is. So here's what I tried. On one session, on one laptop, I cut an episode of AI for Humans in my video [00:20:00] editor, I had Claude Code running a side project in another window, I ran a local Qwen3 8B model in the background for research, I generated thumbnail options in Comfy UI, and I had a Blender window open prototyping a 3D set, and nothing stuttered.
The Intel Core Ultra 9 VPro processor is doing the real heavy lifting here. It's the reason that multiple AI workflows can be running side by side without the machine throwing a tantrum. And the NVIDIA RTX Pro Blackwell GPU eats 3D rendering and ImageGen for breakfast. And with the 256 gigabytes of RAM, that means I can just keep piling stuff on, and the laptop just kinda shrugs, which is fantastic because it's freed up enough time for me to generate what I've been most excited about, which is a virtual raccoon watching app.
It tracks virtual raccoons, it rates their behaviors, it alerts me when one of them is doing something especially troubling. I built it while doing literally everything else that I mentioned. That is the point. When a machine is this capable, you stop pruning your ideas to fit the hardware. There are some ideas that should be pruned.
That's the ZBook Fury. It can do whatever you were going to do today, plus [00:21:00] all the dumb side quests you actually want to do at the same time. Check out the link in our description to spec out your own ZBook Fury, and thank you to HP and Intel for sponsoring AI for Humans. All right, Kevin, the other thing we need people to do is like and subscribe.
We had a really fun show last week. People hyped us up, so keep hyping us. We actually, w- Did we hit the hype charts? Is that what I heard? That was, like, a thing that, that somebody told us, right?
Kevin Pereira: I, we literally ... Well, we got a text from, uh, Will. Shout out Will, who's cutting this right now. Beep, Will. You can do a drop for that if you want, by the way.
Will, I don't know. By the way, we don't call you Big Will in text threads behind your back I think we're
Gavin Purcell: going to now
Kevin Pereira: But we're gonna start. So, uh, shout out to Big Will, who said, "Hey, your video was actually hyped into my timeline this morning." So y- y'all are just a delightful, uh, legion, and we appreciate your support because we're, we're, we're winning hearts and eyeballs and taking over screens one device at a time.
So thank you for the comments, the likes, the five-star reviews, and that sweet algo juice.
Gavin Purcell: That's right. Juice us up, baby. All right, Kevin, there is a giant story that is very nerdy [00:22:00] that I do think is really important to make sure that everybody listens to us, but also you can spread out further because we actually have had...
You know, all these models get released, and we're always talking about the benchmark boys and all that stuff, and like what it can do. But at the edge, at the frontier, what these models are starting to be able to do is pretty incredible, and this is a story that O- OpenAI dropped yesterday. My favorite thing is there's a, there's a video that is, that kind of went along with this of like these three- Yeah
mathematicians who work at OpenAI just being like, "I can't believe it did this." But it basically solved an 80-year-old math problem that previously had not been solved. Um, this is specifically I'm gonna say the headline of their blog post 'cause it's very nerdy. A OpenAI model has disproved a central conjecture in discrete geometry, and this is one of those Erdos problems that people have talked about for a while that some of these are getting- Mm-hmm
better at. But the big thing here is that in the past it has kind of helped the scientists or the mathematicians get to an answer. The big thing here is that it got to the answer on its own. And Kevin, I, I know that this [00:23:00] like, we're a show that tries to disseminate this world for the masses and kind of help people understand.
How would you kind of like tell the normal person why this is a big deal?
Kevin Pereira: Well, if you let, if you let P be a finite set of different points in a two-dimensional plane- No, no, no. No, I, no ... then you define VP- No math, no math ... as the number of unordered pairs of points- Just give me a no math answer. Oh, um, AI, LLMs, large language models- Yeah
your chatbot, uh, people have screamed it's a stochastic parrot. Yes. That means it just takes statistics, rummages around, and spits out words. It just parrots things. It doesn't actually know what it is doing or solving. Well, this is kind of one of those signals where it is, it is, maybe it is a stochastic parrot, but it's, um, it hasn't heard Polly wanting this specific cracker before.
Yes. That's exactly-
Gavin Purcell: This is
Kevin Pereira: new That's very good. That's a good metaphor. Like this is using- Yes ... logic to solve something. Um, and again, if means P is a set of points in [00:24:00] that flat 2D plane, then the means pick any two different points from P and don't counter order. So- Parenthetically, PQ is the same as P comma space
Gavin Purcell: Q Okay.
Uh, I'm not gonna check that math, but I'm just gonna assume you're right. Let's just very quickly hear from Noam Brown, who was the guy, uh, who's at OpenAI that's working on some of this stuff. I wanna just shout out his tweet. He said, "Today we're sharing a general purpose internal OpenAI model achieved a breakthrough with one of the best known combinatorial geometry problems.
Less than one year ago, frontier AI models were IMO gold-level performance." That's a big kinda test to see how they can do. "I expect this pace to continue." And then he continued on and said, "This is a general purpose LLM. It wasn't targeted at this problem or even at mathematics. It was not a scaffold." So the important thing to know about this is they did not create a model specifically to solve this problem.
So one of the things that people always talk about with, with AI is, like, what Demis Hassabis did when they... AlphaGo was a very specific model designed to play Go, and it was still crazy that it was able to make, I think [00:25:00] it was move 52. There was a specific move where it took this big leap and it showed off, oh, maybe it can do this stuff.
Mm-hmm. This is the kind of model that eventually you will get access to. Some people are saying this is GPT 5.6. It might be a bigger model than that, but- Yeah ... this is the kind of opening of the door to AI solving the bigger issues. And when I say bigger issues- Yes ... it's things like fusion, cancer, all of these- Yeah
things that are the large promises that AI has had for a while. We are finally starting to see glimpses of that coming through, which is very exciting- I know ... and also kind of scary, right? It's kinda scary because it just means that, like, this is gonna change. Everything is gonna change. I've had a couple existential moments after reading this.
I was like I don't know if we're ready for what this c- what comes from this Mm Like, I don't know what it means. So that's the, that's the thing I'm coming out of this with. I don't know what it means for the world for the next five to 10 years.
Kevin Pereira: But let's go. Yeah. Let's go. I mean, honestly, like- Yeah, yeah, sure
a- a- as, a- as exciting as it is to discuss, like, lawsuits over the tsunami of reggaeton [00:26:00] Taylor Swift covers that are gonna be- Yes ... you know, unleashed on Spotify, and that, oh, now you can, you generate bears riding skateboards, but it's gonna destroy Hollywood. Like, yes, there are serious questions and whatnot, but, like, the promised future was, was maglev vehicles and teleportation- Yes.
And we are nowhere close ... and the, and the eradication of all human disease- Yes ... and extending our lifelines. Like, let's go. Like, I do agree with Demis, who was like, "I wish we would've solved cancer before we did, like, chatbots."
Gavin Purcell: Yeah.
Kevin Pereira: This, this is the kinda signal that we need. And I know, u- unfortunately, human beings, fallible as they are and imperfect as they are, they will wield these capabilities to do a lot of really terrible things.
But I really, I really hope the signal to the noise is a lot of positive stuff, and solving complex math, and helping us out with our science. And I'm sorry, you got... Are you hearing the Taylor Swift reggaeton cover right now? 'Cause it is so juicy. No, but I'm sure it's, I,
Gavin Purcell: I'm sure it's amazing. I- Oh, we would be
Kevin Pereira: copyright flagged anyways.
Yeah, we'd be copyright flagged.
Gavin Purcell: I do wanna shout out before we go, if you want something on the other side of this platform, there's an amazing video [00:27:00] that we'll just show here that's been traveling around the internet, it's gotten bigger than everywhere, of a Unitree robot attempting very much to try to dance to Billie Jean.
We won't also play that. But please find this video if you haven't seen it already. It's another one of our favorite videos of just, you know, we're not there yet, Kevin. The robot is there trying to dance. It falls down, goes on its side, and it's just, like, another good example of the long gap that we have to get to this amazing future we've all been promised.
Before we leave, I wanna shout out Ben Relles, a friend of ours. Uh, he's got a new company that's launching. I'm doing a little bit of work with him on it, but also very excited for him. He's launching a new company called Make Believe. Ooh. And we will see you all next week.
Kevin Pereira: Gavin, wait, hold on. You can't just say that.
Insider trade me, daddy. How do I play this on Polymarket? You got your-
Gavin Purcell: There is no Polymarket, Kevin ... what's going on with the company? It is a, it is a startup around AI media and AI a- interactive avatars really is what Ben's working on- Oh ... more than anything else. But really cool, interesting stuff he's doing.
Uh, he's been working with Reid Hoffman for a long time, and worked on a thing called Reid AI, which Reid was a very specific interactive avatar. Yeah. Anyway, it's gonna be a cool thing, I think, going [00:28:00] forward, to have, uh, more people that are smart and good and have experience in the AI media space. So look for more from Ben, uh, as it comes along.
Kevin Pereira: Okay, so let me connect my wallet. So is it, like- It's not a, it's not a crypto token ...
Gavin Purcell: they're gonna release the first movie by...
Kevin Pereira: Ah, all right. Bye, everyone. All
Gavin Purcell: right, bye, everybody.