Will the Real Velvet Sundown Please Stand Up?

Speaker 1: Okay. I think you're gonna see where I'm going with this, but give this image a look and describe for the listeners what it is you're looking at. I am looking

Speaker 2: at a band playing. I'm 100% sure this is AI generated or at least AI modified, but it looks generated. Yeah.

Speaker 1: There's clues.

Speaker 2: Mostly the composition because, actually, the most the bigger tells Yeah. Aren't present in this image.

Speaker 1: Yeah. Everyone's got the right amount of fingers in this Right amount

Speaker 2: of strings on the guitar. There's no, you know, soft serve swirls anywhere of things going on.

Speaker 1: How many teeth of roughly the right amount of teeth? Yeah. You got it. Yeah.

Speaker 2: Yeah. Same bewildered look on the drummer that I would expect to see on a regular drummer.

Speaker 1: On a regular drummer's face, for sure. We got a keyboardist next to a bassist with a drummer in the distance and a guitarist.

Speaker 2: Like a real bass too? Like a like a kind of a Fender p bass? Like, it all looks real? Like, the instruments look authentic ish?

Speaker 1: Yeah. Stuff gets a little weird at the edges of the strings.

Speaker 2: Mhmm.

Speaker 1: But generally speaking, it looks kind of like a band. Kind of like a band. So in June 2025, this sixties inspired sunny rock pop quartet that you are looking at an alleged image of, calling themselves the Velvet Sundown, shows up on music streaming platforms. Within, like a couple of weeks, they had dropped three call it vintage flavored albums and amassed nearly 1,500,000 monthly listeners. Their most popular song Dust on the Wind, legally distinct, racked up almost 2,000,000 streams. It's this kinda like, you know, moody, sixties, bluesy, soft rock thing. Pretty much immediately as the band racked up all of these listens, people started going, well, this all seems off. Look at the photo. Listen to it. The band's profile reads like it was generated, in my opinion, by a chatbot. The covers of the albums have that kind of smeary, swirly AI quality to them.

Speaker 2: It's hard to differentiate things when write ups are written by chatbots because all write ups are written by chatbots these days.

Speaker 1: Yeah. As a clue goes, it's no longer sufficient. The promo photos also have this kind of, like, overly glossy quality. There was a bunch of them featuring all four members of the band. What are their names? Gabe Farrow, Lenny West, Milo Raines, and Orion Rio Del Mar. Pro tip if you're gonna make up a fake band, they can't all have a cool name. A band might have one Mick Jagger, but there's just gotta be, like, a Steve in there somewhere. You can't give them all Orions and Milos. You're you're giving it away.

Speaker 2: You've gone too far. You've gone

Speaker 1: too far. People start digging into the velvet sundown. Multiple Reddit posts I was able to find flagged that in some of the press photos, you do have those more telltale signs, the missing fingers, importantly, not the same missing finger across different photos. By this point in 2025 the idea that there is potentially AI slop on streaming services, not that weird. The thing that's interesting is that a lot of people seemed to be listening to the Velvet Sundown. Streaming platforms, users found that the group songs were suddenly populating their personalized playlists. Like Velvet Sundown is showing up in a lot of people's Discover weeklies. And the recurring complaint that people started to sort of orbit around is the idea that, like, there's just no mention here that any of this is supposed to be AI. It's all presenting itself as a real band. This is where this all gets interesting to me because an extraordinary amount of press is now reporting on the Velvet Sundown, on this band that's presenting itself to be real that has this growing audience that seems like it might be totally synthetic.

Speaker 2: They're having their breakout.

Speaker 1: They're having their breakout. And that press causes the breakout to accelerate because, you know, it's everyone's reporting on it all of a sudden. At which point, the hoax banned Russian nesting doll starts to, get constructed. On June 29, the quote Velvet Sundown responds to claims they're AI by saying, no, we aren't. And then someone else says, no, we're the real Velvet Sundown. And yes, we are AI, and this is an art project. And someone else goes, no, we're the Velvet Sundown. No, we aren't. We're AI. And it goes around and around and around, and the press doesn't really know which way to go with the story. And the plays keep piling up. So I wanna start here, because a bunch of people have sent this to us with a story of an AI band hoax within a hoax within a hoax. The story of the Velvet Sundown here on Hacked. How you doing, Scott? I'm good. How are you, Jordan? Doing good. I'm doing good. I'm enjoying the dwindling summer. I'm keeping busy. I'm about to go back out on the road again.

Speaker 2: On the road again? I'm gonna get us road again. Copyright striked.

Speaker 1: Yeah. Totally. Well, apparently, those aren't things anymore, so I think we'll be okay.

Speaker 2: The, yes, dwindling summer is here. We have we are now in September, which where I live is just about snow time. So the, leaves are starting to begin to fall. It was beautiful on Saturday. It was, I think, close to 30 degrees, like, 80 some degrees Fahrenheit. And Oh, yeah. We went to the pool, the outdoor pool, and enjoyed the last throes of, of warmth in the sun.

Speaker 1: I haven't been to an outdoor pool in a while. The outdoor pool near our house where my partner swims just closed down. She's very sad about it.

Speaker 2: 'Tis the season outdoor pools close.

Speaker 1: 'Tis the season outdoor pools be closing. Scott, have you tried using, an AI music generation tool like Sunio? Assuming I'm pronouncing that correctly.

Speaker 2: Yes. I've played with a couple of them mostly for fun just to see what level and capacity they were at. But, yeah, I played with them. They're pretty good.

Speaker 1: Yeah. They're impressive.

Speaker 2: Yeah. The the thing for me, like, Jordan is a music music maker, and I dabble in music making stuff. And music is very sciency. You know, there's beat signatures and timings, and, you know, composition is very, you know, structured often. Structure of a song is very you know, pop songs have a a formulaic structure. So all things that computers can easily do.

Speaker 1: Yeah. Especially if you feed them massive swaths of, reference material. It's interesting. And on that note, I know there's gonna be a story here for years to come about just how legal all of this actually is. And I'm guessing the record label not wanting to get caught in another Napster LimeWire situation is being very, they're employing some very good lawyers as we speak to get to the bottom of all this.

Speaker 2: For sure.

Speaker 1: And unless they suddenly get very cool about a whole bunch of stuff to do with IP, there's probably gonna be some big movements on that front.

Speaker 2: Well, there's been some it's been an interesting conversation. Like, we talked about it on the show before, but

Speaker 1: Yeah.

Speaker 2: So much of music. You know, there can only be so many chord progressions. There can only be so many, you know, rhythm signatures and drum beats. If you reuse one that was used 38 ago in another song and all of a sudden you're getting sued for, you you know, copyright violations. It's like, wow. It's like to me, this this space, just given the finite of artistic like, enjoyable artistic output, like, there's only so many things that work in music. Like, you have all of these different chords. You have all these different tones, and compositions are very similar. Like, lots of modern music of this type are very similar. So it's, yeah. So anyway, to me, it's always been a fascinating industry to look at from the outside looking in because there is such finite boundaries to like what is acceptable and enjoyable to most listeners is a very finite subset of what's possible. And it's also very formulaic. So

Speaker 1: Yeah. If you wanna to me, there's a spectrum. And if you wanna hear about the far side of the spectrum where it's like you really can't own music compositions in the sense that it's just math, you can listen to our how to brute force music episode. We do it. An interview that's totally worth checking out. On this side of it, it's the fact that you can very much own a recording, and these models don't work unless you feed recordings in, which is different from songwriting. No one can own a g major to a c minor change. That'd be a weird change, but no one can own that and sing in a d over it or whatever.

Speaker 2: Totally.

Speaker 1: But the guitar that you recorded that day mixed with the drums, mixed with the vocal performance, that can be owned. And a model trained on that is a different thing than a model that knows to go from one chord to the other. And it's there that we're just gonna be living in where the the exact legal boundary is between those things. And as you said, as a person who likes really loves making music, I definitely have the response, and this is my bias, of, like, oh, yay. Thank you so much for automating the menial, trivial task of self expression through music. Now I can get back to doing my dishes. Thank God someone automated that for me. I can

Speaker 2: go back to these boring things that I dislike.

Speaker 1: Totally. Because

Speaker 2: you've taken away the one thing

Speaker 1: I do like. The one one of the things that really brings me joy. Thank god. Yeah. Which brings us nicely, I think, to the velvet sundown. So like I said, a bunch of people sent us this. I'm sure you saw these come in. Like, this was a big story. We kinda were waiting for a little bit more to happen than just there's an AI band getting listens, and a little bit more happened. It seems worth chatting about just to kick the episode off. So pick up the story from the intro. News outlets picked up the story of the Velvet sundown in, like, the final week of June. At this point, they have over half a million listeners. It's, like, pleasant enough if you've listened to it. It's quite, it it sounds like if an a an algorithm stitched together soft rock cliches based on a bunch of soft rock in a training data set. An industry a and r veteran quoted anonymously in Rolling Stone, Rolling Stone comes up later in the story. Noted that, like, the band's buzz came from this, like, murmur that they're not real, not because the music's necessarily exceptional. So faced with these suspicions that we're starting to kinda fester, the Velvet Sundown's purported members or someone on their behalf hits back publicly. June 29, a Twitter account saying, like, we're the Velvet Sundown unleashes this, like, very angry tirade denying that they have anything to do with artificial intelligence. Quote, absolutely crazy that so called journalists keep pushing the lazy, baseless theory that the velvet sundown is AI generated with zero evidence. Quote, this is not a joke. This is our music written in long, sweaty nights in a cramped bungalow in California with real instruments, real minds, and a real soul. To my mind, that denouncement, to your earlier point, does sound AI generated, but that's neither here nor there.

Speaker 2: Yeah. It's I feel like they probably watched the Anderson Paak Bruno Mars Silk Sonic success because it kinda has the same hues and color energy.

Speaker 1: Like, I

Speaker 2: have a lot of, the aesthetics of it. Great rep. And they they kind of were like, what if we did this?

Speaker 1: Yes. You

Speaker 2: know, what if we made a kind of a not surf rock, but kinda I don't I don't know. Like, a seventies inspired acoustic rock band that kinda has this energy. And we've seen that a few times over the years. Like, there's been a lot of artists that have come out and replicated Dolly Parton, kind of taken energy and inspiration from them.

Speaker 1: Trends and cycles. Trends

Speaker 2: and cycles. Totally. Everything bold

Speaker 1: is new again. Not a lot of new ideas under the sun.

Speaker 2: We're back in the nineties.

Speaker 1: Yeah. They went back to the sixties and seventies

Speaker 2: Totally.

Speaker 1: Like many before them. The words we and they in all of these characterizations are doing or let's use those loosely. It's like, I don't know, like, what when we say they did that it's like, my theory is a prompt did that. Yes, exactly.

Speaker 2: Yes. They prompted.

Speaker 1: Yes. Whoever prompted this did that. So this band looks like it's AI. No. We're not AI. The that protest is, I would say, pretty easily debunked by the evidence at hand. Like we said, a lot of this imagery feels very clearly AI generated. One Twitter post put it well. It's like, oh, cool. You have millions and millions of downloads right now. Make an appearance on live television. Prove it. Put out a video. Like, just show a human being in a room, and we'll all believe that you're human beings, and this becomes a remarkable story. And you probably become wildly over famous overnight going from being like this band, AI band got a million and a half views, and it turned out they were real. Like, that story would have gone wide.

Speaker 2: I I can't help but think that this story the storyline that you have in front of you that we're gonna talk through is the it's not the act that gets you in trouble. It's the cover up. A 100%.

Speaker 1: A 100%. And just hoping this to lie in a situation like this.

Speaker 2: Have they just come out and not said anything? Yeah. People being like, you guys are AI and just ignore the posts? Yes. They probably would still be accruing listeners today. But the second thing, like

Speaker 1: The the thing that's weird is I'm I think they might still be. But you're right. Like, you could let the mystery kind of live. Except maybe no, you couldn't because someone else would try and own it. So around the same time, like, a different streaming platform, one that we don't really have much here in the English speaking world, it's called Deezer. I was not very familiar with it. Mhmm. Started flagging the content. They added, like, a tag to it saying this is AI generated content. Quote, some tracks on the cell may have been created using artificial intelligence. Other platforms, to date seem to have no such warnings. That French company was able to detect it just using, like, a filtering system, which means they exist. So the question of, like, where did all of these streams come from? In recent years, major streamers have all kind of opened the doors to paid placement on, like, influential playlists. Recommendation algorithms have shifted away from human curators to, like, automated selections based on audio characteristics that are analyzed out of the the the audio file, which therefore can be factored for and put into an audio file by like a music generation

Speaker 2: smart AI.

Speaker 1: You got it. Those changes, Glenn McDonald, a former Spotify data alchemist said, there's fewer reasons why a fake band couldn't be successful on these platforms.

Speaker 2: I'm excited to hear the end of this if they'd know who made this, but I wonder if it wasn't a data analyst for somebody like a Spotify or a Tidal or a Deezer who knew exactly what to make to feed the recommendation playlist to feed, you know, to get mass coverage.

Speaker 1: Yeah. It's it all has the sheen of either a data scientist or a marketer or someone who's less maybe on the, like, songwriting and creative side because they probably would have written some songs, but someone who's more on the releasing and putting stuff out there into the world side of things who came up with a scheme. There was an interesting point here. It was that same McDonald guy was talking about how, like, in the past, the issue was fraudulent listeners, like bots inflating play counts on different platforms. We've talked about this a ton on this show. Hence. They were a bigger problem than the idea of, like, fraudulent content. And his line was like, now maybe it's the opposite. Now the question is shifting towards like, am I is what I'm listening to created computationally is the are the listeners computational, like, the what's real of it all is getting really, really hard to like, separate

Speaker 2: apart. I can't I can't can't help but, like, draw back to the like, the first major one of these was the AI, EQing and changing voices, and it just makes me remember the Drake weekend song that got released and, like, the Internet witch hunt to figure out who did it. I don't know if they ever figured out who did it, but most of the stuff that I read because I went down that rabbit hole was that it was this independent artist named Jake. And this is by no means

Speaker 1: mostly that

Speaker 2: they did it. It's just one of the allegations that the person that had done the research had so much backing for it. Sure. And, yeah.

Speaker 1: I remember that story. What was interesting about that is so if you've ever used Sunio, and we'll talk more about it in a second, but like, they're they are very analogous to, like, a mid journey or an image generation type process. Like, yeah, you can give it a reference image, but honestly, you're probably just giving it text prompts. And I remember that story was, like, a person who had very clearly made songs Mhmm. And was like rapping over it, like just spitting bars and then reskinning their voice to sound like these prominent artists. And I was like, that's such a different thing. And you can have an opinion that it is not okay to do that with an artist's voice and intellectual property. I'm really open to that argument, but that's such a different thing than just, like, yeeting a prompt into Sun Yos and remember what the vocalist sounds like so it's consistent. It's like, that's just so different.

Speaker 2: Totally.

Speaker 1: And it yet it when that happened, it felt like, wow, fake AI music. What what have we become? It's like, the Halcyon days. Like, that sounds great compared

Speaker 2: to this to me. Yeah. Yeah. Well, that like, that Drake weekend song was completely original composition. Yeah. They they wrote the music. Yes. They wrote the lyrics.

Speaker 1: Yes.

Speaker 2: They sang and wrapped the lyrics and then used AI to make it sound like Drake and the weekend.

Speaker 1: Yes. It was a Drake and weekend filter on top of an original composition as opposed to a very, allegedly markedly not original composition

Speaker 2: Totally.

Speaker 1: Or not composed by a human and therefore probably not composed in a traditional sense. It was weird. Like, this thing, the this Velvet Sundown Band is is showing up in all of these places. It's showing up on people's feeds. It was in a because of the 70 sound. I liked this one. There was a playlist about the Vietnam War where it like, a prominent one of, like, classic Vietnam era songs, and it appeared on that. It's, like, no. Definitely not. It was it was it was already a little bit weird at that point, at which point enter a man possibly likely named Andrew Freeland. July 2, this guy, Andrew Freeland, contacts a bunch of media outlets, including rolling stone who runs a pretty prominent story on it saying like, we're the, I'm the, I'm the person that has been posting all of these comments on behalf of velvet sundown. I'm like a spokesperson and kind of like adjunct member of the band gets on a phone interview with Rolling Stone magazine and says like, you know what? We gotta come clean. The band is indeed using AI tools to make their music. They were using Sunyo in their songwriting process. And he described it as kind of like an art hoax thing and sort of like waxing philosophical on, like, have things that are fake sometimes have more impact than things that are real. He's presenting this whole thing as kind of like a marketing ploy, and he's been pretty wishy washy on, like, how much of it's AI and the degree to which it's AI. It's saying it's like, I've got the quotes in front of me here. It's marketing. It's trolling. People didn't care about what we did before, and now suddenly we're talking to Rolling Stone. So it's like, is that wrong? Yes. I added the yes at the end. It's not great.

Speaker 2: Can I can I deviate this conversation?

Speaker 1: And how?

Speaker 2: How is this I struggle with a lot of major music already in the fact that most of it is not written by the artist.

Speaker 1: There

Speaker 2: are songwriting camps and groups and people that make entire livings just coming up with writing the lyrics for and pitching songs to popular artists. So if you look at the credits on any CD, most of it, most of the songs on any top 50, top 100 Billboard artists, they haven't written themselves. There will be major singer songwriters credited on every single song. They might do a twist tweak to the lyrics here to make it more personal to them or Sure. Some small thing.

Speaker 1: They're still singing it. They they can decide whether it's a word, and they perform it, perform it.

Speaker 2: They become the performer.

Speaker 1: I see where you're going

Speaker 2: with this. But but largely, I kinda wanna call it a fraud because it's it's not a fraud, but it's like it's not like Fleetwood Mac, you know? Like, they wrote music, sat in the studios

Speaker 1: Mhmm.

Speaker 2: Did drugs, came up with the songs, and recorded them and like made a generation of music that like people still adore to this day with real instruments. Nowadays, everything is done in a DAW. Any kind of technical expertise to creating music is largely inside of a computer now, less on a drum kit or with a guitar. Like, yes, there are still the ability to integrate that stuff, but you can still use MIDI, compose it, put it through the right plugins and VSTs. Like, we're having a whole discussion about anyway, the the soul of the creation of the music is more about artistic expression through music, less of technical expertise that leads to artistic expression, which is where I think it was more in the 70s.

Speaker 1: Yeah, I follow what you're saying that it has changed. I don't feel as though it has necessarily become, in the ways that matter to me, less humanly authored.

Speaker 2: It's

Speaker 1: just that the structure surrounding it has changed. Like, Fleetwood Mac is a way of presenting a group of musicians making music together. Yes. We we call it a band, and there's the

Speaker 2: there's a Yeah. Yeah.

Speaker 1: They all play an instrument, and that's what how it works. There's not really under the hood, though, that much of a difference between a band with a singer and a few instrumentalists and a singer, who's the front person, and a community of instrumentalists and producers and creators that have come and written the stuff around them. It's still human beings coming up with musical ideas and figuring out how to weave them together to create a song. Typically, in pop music around a vocalist. It's different and it doesn't feel the same because it's like typically different musicians on each song. It's It's definitely a lot more factory floor. You're right. There's, like, a that personal touch to to all of it is is less so. But it's still authored by humans in a room, certainly with drugs sometimes, together in those big songwriting houses. Like, so there there is still it's still human authored, even if it's authored in this very weirdly optimized way that it has become over the last, like, twenty years, it feels like.

Speaker 2: Yeah. Yeah. I hear your point. But to like, the the thing for me is like we've already made the jump from singer songwriters. You know, Taylor Swift went from being Taylor Swift, the country singer songwriter to Taylor Swift, the pop star who has other people write her music. And she did so very successfully. She still wins awards for writing country music that she doesn't perform because she was a country songwriter. And the jump for me is less so, where it's like if I if I wanna find inspiration in music by having an AI come up with concepts for me, is that any worse than me asking it a question about, like, hey. We wanna do something in marketing, and I have a question about campaign optimizations, or I'm going to sit down and develop this software system, and I want to talk about the architecture of it so that I don't miss anything. Like, I'm I'm engaging with an artificial intelligence to help prog like, progress what my goal. Yeah. If we're already at the point where I'm a major billboard artist and I just get sit in a room with a producer and listen to 700 song pitches in an afternoon and choose the ones that I like, Is that any different? I guess that your argument is that that humans created those pitches where I'm saying the AI created the pitch, and that's the major difference.

Speaker 1: I mean, so in that situation, you still have a human vocalist performing on top of an instrument.

Speaker 2: Of course. Yeah.

Speaker 1: And if the instrumental was created through a generative mechanism, that's still different than a human being going like, shit. I just don't know what chord to play next here on my guitar, which is plugged into my DAW. Like, to me, there's

Speaker 2: just But but but but very I would say, like, just given how how fear theoretical music is, especially how formulaic certain genres are, like, they know what chord they're playing next. They know what transitions hit the audience marks for like, I I like, we're I'm talking about a very specific set of formulaic music, like, in true artistic expression, for sure. But, like, you don't hear a lot of that. You have to go find that music. That music's not pushed out to you on a on a now trending list on on a music platform.

Speaker 1: I'm very I'm aware of how people seem to react when they find out that a thing wasn't played by a human being, which is this feeling of, like, I I came to this to see you sweat. Like, I came to this. You're you put your shit down. Yeah.

Speaker 2: Without dying But it's not even their shit most of the time. It's somebody else's.

Speaker 1: But still someone's. So there's there's a spread here. Right? Some people would be totally some people probably are listening to Velvet Sudden and going, I just love the songs and don't really give a shit. A lot of people didn't. But the question of, like, is this a real band or a totally synthesized band or something in the in the in the middle starts to really veer towards, like, no. This is this is totally synthetic. And the story in his telling, Andrews, is that Velvet Sundown is like a prank to prove a point. He'd like fit they'd feigned outrage at the AI accusations, which earned this extra wave of, like, publicity backlash, and the strategy seemed like it worked. The band's streaming listenership, like, triples. It goes from, like, 300,000 monthly listeners to over 900,000 monthly listeners in, like, a matter of days as part of this giant press cycle. And during that interview, Freeland talks more about how the band was made, which I wanna bring up here because of what we're talking about here. Mhmm. At first, he hedges, saying AI was only used for brainstorming musical ideas to your point. As the discussion goes on, this is on the same call. He concedes that, like, no entire songs were generated using Sunyo. Though he wouldn't say which ones. And specifically, so if you look at how Sunyo works, he confirmed that they were using this thing called the Persona Features, which is like you can synthesize a consistent lead vocal across a bunch of different tracks so that this thing that's presenting itself as a real band, sweating in a bungalow in California, sounded consistent across all of these tracks. This is like a kind of an AI voice cloning feature inside of the project. Like, Timberland used it on his art artistic process Tata, which is like an AI generated thing that is, like, publicly invisibly and kind of celebrating itself as being AI, which I find much more interesting. So it seemed like velvet Sundown's creators had the AI write and sing pretty much all of the material. There was maybe a little bit of human direction. Freeland then gets asked what I think is maybe the one of the most important questions about whether they're juicing their streams via, like, playlist manipulation and bots. And he just said, like, he kinda hedges and goes, like, I know we got on some playlist that have a ton of followers, and it seemed to have spiraled from there. Still unclear. So now everyone goes, okay, the big reveal, they would send down is a AI band after all. It's confirmed by an insider. We are importantly not at hoax within a hoax yet. Freeland's admission comes out, makes a bunch of headlines, and then confusion sets in over who Freeland actually is and whether he's telling the truth. Because then on July 3, days after his official Rolling Stone interview, the band's official Spotify linked social accounts put out an urgent claimer saying that the individual calling himself Andrew Freeland is not associated with the band. Rolling Stone quickly publishes publishes an update saying that Freeland, the professed representative of the allegedly AI hoax band, seems to be a hoax himself. The band says we have no affiliation. We have a hoax inside of a hoax. Everyone's scratching their heads. Half of people are saying this is very clearly all AI. Other people now are starting to swing in the other direction saying, no. It's not. That guy was a liar. BBC summed it up pretty well. Rolling Stone reported the band was AI accredited, quote, only for the magazine to report shortly afterwards that the spokesman him himself was a hoax. Freeland kind of admits like, claims the stunt was a deliberate plucks to plot to hoax the media about the hoax.

Speaker 2: How many hoaxes can you do in this?

Speaker 1: Just layers of hoaxes. In a very long Medium blog post on July 3 published hours after the official band socials denounced him, Freeland, came clean kinda describing himself as a professional in online safety and self professed expert in red teaming tech platforms and counter disinformation tactics. When he saw the story kind of blowing up, he thought, oh, they don't seem to have a huge social presence yet in spite of all these views, so he rebranded some old unused Twitter handles that were the right age to Velvet Sundown squatting on the band's identity, used generative AI to, like, reskin all of it to

Speaker 2: look real,

Speaker 1: and then goes on a publicity tour. Reporters start reaching out to the accounts. He's ready, sets up dummy emails as, like, the band's PR person and goes and does all these interviews. Notably, when speaking to Rolling Stone, Freelancers, the journalist never questioned the authenticity of his account or the existence of two separate Velvet sundown profiles on x. People were just really anxious and eager to find out what this was and who was behind it, and he sort of just presented himself as a perfect avatar for this. Still, I have to say allegedly, AI generated band.

Speaker 2: I'm looking at Andrew Freeland's medium where he there's only three posts, but it goes, I'm the guy running the fake Velvet Sundown Twitter. So, yeah, I did make Velvet Sundown. And so, yeah, I'm also Margo Blanchard, the fake journalist that's been reporting on it as well.

Speaker 1: I'm not convinced Andrew Freeland exists.

Speaker 2: No. No. Definitely not. I can tell by looking at his little medium photo that he's also this is an AI photo as well.

Speaker 1: That's where the next hoax with it, I can stop being dramatic about it. It's like, I think then the fourth layer maybe is that I don't think Andrew Freeland exists.

Speaker 2: He is probably sitting in a studio with his buddies making actual music, drinking, and doing drugs, and laughing.

Speaker 1: Oh, I

Speaker 2: would love that. Yeah. I would love to read all

Speaker 1: of this. If it's a real musician and not, like, just an Internet person, I think that would be great. So July 5, Velvet Sundown's official Spotify bio is updated with, like, a a flat out acknowledgment, and this is at least a step in a good direction. The Velvet Sundown is a synthetic music project guided by human creative direction and composed voice and for visualized with the support of artificial intelligence. I think support is still, if I had to guess gut instinct, a disingenuous word, but it is at least acknowledging it is synthetically created. And then the quote, this isn't a trick. It's a mirror. An ongoing artistic provocation designed to challenge the boundaries of authorship, identity, and the future of music itself in the age of AI. Man, I added the man again. Sorry not to add stuff to this. As of right now, the creators are still kind of, I I think to the best of my knowledge, anonymous, a third Velvet Sundown album, Paper Sun Rebellion, dropped on July 14. To my mind, very it's like you just you don't write and record songs that quickly. And the band's popularity has grown. They've got over a million months of listeners on Spotify by the point that that album dropped. The day after he did this kind of dramatic I did this post, Freeland or someone using that name that may or may not be his posted another update saying he'd been hacked. He is now at this point somehow a less reliable narrator than the made up AI band at the heart of a hoax within a hoax

Speaker 2: within a hoax

Speaker 1: within an Andrew Freeland within a velvet sundown.

Speaker 2: Listen, you've gotta respect the troll. It's quite the troll.

Speaker 1: It's quite the troll.

Speaker 2: The thing that I don't understand is Timbaland has launched Stage Zero, this whole AI entertainment company, which is, again, as you said, fascinating and totally something we should talk about. For sure. Why has he not signed Velvet Sundial? Like, they got they can they come with a massive amount of press. They have publicity, a massive user base. Like, if if you were in A and R, an AI A and R, like, is there a bigger AI band right now? I think I actually have

Speaker 1: a really good answer to this, which is in a money grubbing context. Mhmm. Going to an AI band created with some by someone would mean having to split the profits with them. Mhmm. When that person has inadvertently approved that it is so trivially easy to create this content and game these systems to get to that level of success, that why would you ever share the profit? Why not just spam, like, throw AI slop spaghetti at the wall until you get a hit of your own? Maybe you can get Rolling Stone to cover it. Like, that's how I would be approaching it.

Speaker 2: I haven't listened to any of Tata's music. Maybe. But she just from, like, straight impression, I I get the feeling, like do you know who Rose is? She's in Korean New Zealand Taylor Swift, would be the the, like, the analogy that I would draw. Like, I listen, like, I've heard I think it's Rose, Rose. I can't remember her exact name. But when I hear her music, I just assume I'm listening to Taylor Swift. Because all of the mannerisms and the little, like, you know, accents and stuff that Taylor Swift does to, like, move the music along and, like, her her song structures, Rose has

Speaker 1: nailed. Interesting.

Speaker 2: It But Rose is a real person. And if you if you had played the music for me and told me that this was a clone of Taylor Swift's music, I would be like, yes. It is. This is an AI clone of Taylor Swift's music.

Speaker 1: Which is funny because it feels like a thing that could only happen in the AI in the age of AI cloning except for the fact that it is the oldest story in music. Someone does something original, and then to your point, an industry, like engine steam engine boots up to just be like, how do we make another one? Totally. And it's like, by getting them to make more because it's them. And it's like, no. It's not what I'm talking about. Like, how do I get one? And suddenly, a bunch of people have to go do an impersonation of someone. And it becomes a style and a trope. And then someone else rejects that and does something new. And the whole thing starts over and over again. Yeah, exactly. Yeah. I I wanted to talk about that the second I saw the story back in July, but it there just wasn't quite that extra twist of us having anything to, you know, to say about who's behind it. And then within, like, two weeks, there was way too much to say about who is behind it. And

Speaker 2: yet somehow And none

Speaker 1: of it was real. Exactly. It's a nightmare. It's just like, the great group of vacation has begun.

Speaker 2: Here here here is stuff. I wanna close this segment by reading a quote from Timbaland about his new AI entertainment company, Sage Zero, from Billboard.

Speaker 1: We should end every episode with a quote

Speaker 2: from Timbaland. Well, we're still gonna chat, I think, after the break.

Speaker 1: Oh, yeah. No. We're not done yet. Yeah.

Speaker 2: But, I'm not just producing tracks anymore. I'm producing systems, stories, and stars from scratch. Tata is not an avatar. She's not a character. She's a living, learning, autonomous music artist built with AI. Tata is the start of something bigger. She's the first artist of a new generation. A pop, not k pop, a pop, AI pop is the next cultural evolution, and Tata is its first icon. Timberland. Timberland. Grammy award winning mega producer.

Speaker 1: I'm looking at other quotes

Speaker 2: from Any fear you have about where AI is going, Timberland has just stepped up, driven them down your throat.

Speaker 1: Timbaland riding in the the tip of the spear of the four horsemen of the AI apocalypse.

Speaker 2: He's like, I can I can write good music, but the annoying thing has always been working with artists? So I'm just gonna replace them with AIs that do what I tell them.

Speaker 1: If I have to see Justin Timberlake in studio one more time. Exactly. Another quote from Timberland, you don't question where music comes from because it all comes from God.

Speaker 2: Strong power quote.

Speaker 1: It's a power quote, and I wonder how to reconcile that with the quote you've just said, and maybe we go to ads. Starting something new isn't just hard. It can be downright terrifying. You put a lot of work into a thing. You're not entirely sure it's gonna work out. You're taking a huge leap of faith. I've started a few things. Now I know I was right for believing in, you know, the idea, the product, despite all of those fears and hesitations. But boy, does it sure help when you have a partner like Shopify on your side. Shopify is the commerce platform behind millions of businesses around the world and 10% of all e commerce in The US. From household names like, well, hacked podcasts merch to brands just getting started, you can get started with your own design studio with hundreds of ready to use templates. Shopify helps you build a beautiful online store that matches your brand style. Did I mention that that iconic purple shop pay button is used by millions of businesses around the world? I don't know why I wouldn't. I should. It's why Shopify has the best converting checkout on the planet. It also helps boost conversions, meaning less carts, sort of getting abandoned in the parking lot and more sales for you. It's time to turn those what ifs into sign up for your one dollar per month trial at shopify.com/hacked. Go to shopify.com/hacked. One more time, that's shopify.com/hacked.

Speaker 3: This Father's Day, do more with dad and spend less with low prices guaranteed at the Home Depot. Get him fired up with a new grill and accessories, like the next grill five burner for just $299 so you can spend more time together while he becomes the grill master he was always meant to be. Or build memories with savings on top brand power tools so you can tackle projects side by side. Give more and do more together this Father's Day with help from The Home Depot. Exclusions apply to homedepot.com/pricematch for details.

Speaker 1: Study

Speaker 2: and play. Come together on a Windows 11 PC. And for a limited time, college students get the best of both worlds. Get the Unreal College Seal, everything you need to study and play with select Windows 11 PCs. Eligible students get a year of Microsoft three sixty five premium and a year of Xbox Game Pass Ultimate with a custom color Xbox wireless controller. Learn more at windows.com/studentoffer. While supplies last, ends June 30, terms at aka.ms/collegepc.

Speaker 1: And we're back.

Speaker 2: And we're back.

Speaker 1: And we're back.

Speaker 2: And we ended before the break talking about Timberland.

Speaker 1: Long may he reign. Truly one of our, like, generation's greatest pot makers. Like, all the goofing aside, if you've ever seen video of the man in a studio, he is quite remarkable.

Speaker 2: He

Speaker 1: is. Anyway, what were we what were we saying?

Speaker 2: We we have to talk about other random things now.

Speaker 1: Yes. We do. So We do. How's your day, how's your crypto investment going, Scott?

Speaker 2: Oh, yeah. Take the Nathan. Let me check. I think last I looked, I was down 41%. But you know you know how these things get Highs

Speaker 1: and lows. Highs and lows.

Speaker 2: Highs and lows. It's red, so I should probably buy more. Isn't that the the rule?

Speaker 1: Yeah. Totally. Throw, throw good money after bad. That's what they say. Anyway, we don't gotta dwell on that. What should we talk about here?

Speaker 2: Well, I've got something that's available in Canada. I don't know if you've seen this, but the AI friend necklace is now launched in Canada. And you can buy it, Jordan. I can buy you an AI chip that you wear on a necklace around your neck that listens to everything going on and makes snarky comments about people and things that are happening around you.

Speaker 1: I would normally so we don't normally talk about multiple stories in the same kind of wheelhouse, which would mean that I would normally have a lot more, like, exploratory questions and interest and no, this can go straight to hell. What? I am I'm wearing it. What am I doing? I don't like this.

Speaker 2: It's it's essentially the, the new generation of AI wearables, and this is essentially a friend. So it's literally a voice chatbot that you wear on your neck that listens to what's going on and converses with you as you wander through the world every day.

Speaker 1: So it's a microphone on a necklace?

Speaker 2: Yeah. With a chatbot integration.

Speaker 1: Right. But my phone has a microphone and chatbot integration.

Speaker 2: Yeah. But it's not Yeah. But shut up. There's no way this thing has any actual Internet connectivity, so you probably still need your phone to utilize it.

Speaker 1: Yeah. That's kinda where my I was going with it a little bit. I was like, like, the microphone.

Speaker 2: The biggest thing that I'm interested in is k. Is the rumors about what OpenAI is building

Speaker 1: Oh, sure.

Speaker 2: Are very similar to this. So I'm intrigued to see that the negative response for this, if that will adjust OpenAI's direction.

Speaker 1: So for anyone that doesn't know, OpenAI spent a small nation's GDP acquiring, Johnny Ive, the industrial designer behind one of most of Apple's great products. His company essentially Yeah. Brought them in to be designing hardware for OpenAI. They've there's been rumors that they're like, well, it's not a wearable. It's something else. But it's not a phone. It's like a third device that you would carry with your laptop and your phone. It's this thing, but it's not a wearable because we know you hate those because the rabbit r one sucked. Personal opinion. The AI pin was not successful at all. A lot of these companies are falling apart. There's this weird thing of, like, it seems like you would have to be talking about a wearable, but people don't seem to like these AI wearables. So what is it you spent $6,000,000,000 on? That's to catch people up.

Speaker 2: Yes. All the speculative rumors is that I've seen are that they are essentially making, an iPod Nano or whatever the tiny little iPod. Yeah. The

Speaker 1: cool. I bought. That you wear

Speaker 2: on your like around a necklace, which is very similar to what this this is. Yes.

Speaker 1: That is again, like, for all practical purposes, we've talked about this on the show before, but, like, that's a microphone accessory for a phone. Like, it might be able to do some of the processing locally on it, but if it doesn't have a screen, you're gonna go over to your phone. My theory being that what this effectively is is gonna be a I'm guessing, and I don't wanna sound overly pessimistic, but, like, a trivially useful hardware thing that is used as a gatekeeping thing for a software feature. They're gonna roll out some crazy agentic shit, and the little piece of hardware is gonna be the thing you have to buy along with order to get access to it. That's my gut of where this is going.

Speaker 2: It's more

Speaker 1: of a software thing with, like, a little hardware appendage that, you know, has some sensors in it.

Speaker 2: A little creepy always on microphone that's listening and digesting every piece of intellectual property that comes out of your mouth or anybody else's mouth around them. Yeah. I I suspect the same. It's your friend, Scott.

Speaker 1: The hardware component of friend connects to an iPhone via Bluetooth continually listening. I'm just reading marketing copyright. Yeah. Listening to all interactions around the where pressing a button on the net necklace allows the user to speak to friend with a message sent to the iPhone in response. Friend is also able to send messages without a conversational prompt responding to the what the wearer is doing, which is if I'm understanding correctly, just means it sort of, by definition, has to be listening all the time, which is a choice.

Speaker 2: It's I think it's becoming this is

Speaker 1: Yes.

Speaker 2: To me, this is becoming like a demarcation point as to whether society rejects this or accepts it. A 100%. Because there's a few of these products now that are, like, complete privacy violators that are just, like, I'm always listening to everything that's going on around you, and I am consuming the intellectual property and sending it back to the server farm in China. And are you cool with it? And I feel like this is yeah, I feel like society is now in this in this point of being like, are we cool with this? I don't know if we're cool with this. And OpenAI just spent $6,000,000,000 to try and, like, push the coolness of this. Yeah. And I don't think society's gonna love it. There's rumors that the new AirPods aside from being able to, like like, obviously, the new Beats Pros have the heart rate detector in the earpiece.

Speaker 1: Mhmm.

Speaker 2: There's rumors that they will have always on microphones that will have similar functionality. Rumors, speculation.

Speaker 1: And Apple has historically been pretty good at compartmentalizing what they run locally on a device. We know these models that are leaner and can run locally or like, I could see them trying to differentiate themselves from the pixels of the world by saying we do all of that locally. So, yes, it's always listening. It has the convenience of AI with none of the privacy violations of AI or something their appley version of that marketing copy.

Speaker 2: Copy. Totally. Apple Apple has always been very good about privacy and confidentiality. You know, they've gone to task with the governments and stuff like that over keeping private information private.

Speaker 1: Yes.

Speaker 2: So I can I they would be my trusted play in this? If I had to make a bet on somebody making something that people would actually adopt and use, it would probably be them because they also have more money than a lot of the country's GDPs. So Yeah. They they don't need to particularly try and get rich. They are already rich.

Speaker 1: Yeah. And there's they have a lot to lose as well.

Speaker 2: And That's true.

Speaker 1: Their privacy like, there's billboards on the side of highways taught using the word privacy. And it's like you you can really you could really do some damage if you roll one of these things out in a bad way. And I think that they probably feel a little bit like gun shy after the Apple intelligence rollout. So I would imagine that they're gonna be very careful. There was people talking about the meta ray bans. It's like, does this device pass or fail the meta ray ban test, which is after you take your finger off the talk to it button? Can you keep talking to it and it will answer you? By and so the thing people do is, like, can you still hear me? And does the device after it by how it's presented as working should no longer be listening going? And it was someone going, no. I'm not listening anymore. It's like, cool. Love that. Love that. Like, it knows it's not supposed to. So that's I think that's gonna be, like, we're we're all gonna it's like, I'm I love my AirPods. They're they're a great product. And the idea of heart rate monitors is cool. And the idea of Siri not being useless is great. But how you roll that stuff out is, like, people are gonna stress test the living hell out of that.

Speaker 2: Yeah. The Google the Google XR glasses, like, the next generation of Google, like, way past Meta Ray Ban functionality.

Speaker 1: Totally.

Speaker 2: I'm I'm so intrigued by them. It's cool tech. Yeah. It's cool tech.

Speaker 1: All of the, like, wavelength displays and, like, pass through stuff. I'm, like, as a nerd, what do

Speaker 2: you mean

Speaker 1: cool stuff?

Speaker 2: I mean, I have a hard time not preordering a pair of those. Sure. And then I'm gonna have a long philosophical conversation over a glass of cognac with myself about whether I'm okay with it. And also in what regard, like I this, yeah, it's going to be very fast. And Google is another one of those companies. They've, you know, when it comes to these things, like Friend is a smaller startup.

Speaker 1: Yes.

Speaker 2: It's like the Googles, the Apples, the Microsofts, the majors, I think you're gonna have totally different bumper rails on their stuff. And I think that's gonna be fascinating to watch the the moral development of society around, are we cool with this and how it works?

Speaker 1: It's tough because this is kind of an aside from, like, the ethics of all of this. But, like, just as a person, I love the idea of small hardware startups. This is an in the weeds thing.

Speaker 2: Totally.

Speaker 1: But, like, when Rabbit started, I was like, that's awesome that you're doing this rats. Like and and all of the AI hardware, you know, era has just been that over and over again where I'm like, I love the idea of the non incumbent coming in and being, like, supply chains have gotten cheaper. The ability to create something from scratch has gotten cheaper. Let's make some crazy crap. Let's throw something at the wall and see if it sticks with hardware in the way that people have been doing it with software for years. I just want someone to have an idea that's different than what if we took the sensors that are already in your phone and put them in a different little box and have those sensors talk to the chatbot. Like, I just want a different idea than that. That would make

Speaker 2: me happy. For you. You ready?

Speaker 1: Lay it on me.

Speaker 2: We're gonna tie this whole episode together with this idea. K. Let's make a jukebox that takes prompts rather than, like, song requests.

Speaker 1: Thanks. I hate it.

Speaker 2: It's direct directly jacked into Sunio.

Speaker 1: Yeah. Yeah. Yeah.

Speaker 2: You set up, like, a full old school looking jukebox, and maybe there's an audio interface or maybe there's a keyboard on it. And you just type a prompt into it, and it'll generate and play the song in the bar. That's

Speaker 1: you would you'd make a lot of money for, like, a long weekend. I think a lot of people would come in to try that once and then never come back. But it's for a minute, you would you would you would have a thing there. And then people would go like, oh, no. I wanna listen to good music. And then they would go to a different bar where someone's playing. Like

Speaker 2: The the funny thing is that it would take music. Like, you could do this in a hackathon. It's like a Raspberry Pi

Speaker 1: Yeah.

Speaker 2: With, like, a network connection and then a set of speakers. It would be so easy to make, and it would be the thing that I would get a wired article for. Like what you're saying? Like, I can already see it.

Speaker 1: Yeah. We're talking about this friend thing, and it's like a little bit jibbit jibbit thing with a microphone in it that, like yes. You would a 100% get a bunch of

Speaker 2: press release. Media coverage.

Speaker 1: Yes.

Speaker 2: Yeah.

Speaker 1: It's we would talk about it, probably.

Speaker 2: Probably. Should we do it? Should we do this? I'm good. A hackathon.

Speaker 1: I like the idea of a hackathon. I mean, we've talked about this before. Like, I wanna do our our great great hackathon at some point. I wanna do a great game jam. I wanna do a great hackathon. Like, we we we're scheming. But, yeah, it's interesting. It's just funny to think that every it's like I was at a a bar the other day that it had kind of like a hi fi bent to it. It was like a bar and restaurant, but they had, like, someone had really put the time into, like, the acoustic treatment and speaker arrangements so that it hit every table right. It was like that was kind of the thing

Speaker 2: I dig it.

Speaker 1: About it. And the idea of going into a bar and it's just like a bunch of, like, little Apple HomePod streaming like that hissy Sunio generated music is just like the antithesis of that. It's very, very funny to me.

Speaker 2: The going to, like, jazz bars in Japan, so much fun. Like, beautifully done.

Speaker 1: It's a cool concept for bar because, like,

Speaker 2: the acoustics

Speaker 1: of an environment, and a restaurant bar. Go to a restaurant that doesn't have acoustic treatment, and you're like, why is it sound and feel bad in here? And to go and be like, that means there's a spectrum. Let's go to the far other sound of it and just make, like, the air feel right for music is like, that's cool. It's nerdy, and it's in the weeds, and it's not for everyone, but I I appreciate

Speaker 2: it. I am such a I won't go back to really good restaurants that have bad acoustic treatments. Like, I'm so sensitive to it. It drives me up the wall. There's a really nice restaurant in Edmonton owned by a friend of mine. Makes great food. I've been twice. And every time I go, I have a terrible experience because it's, like, 96 decibels of average sound the whole time you're in there. And it just gives you, like, this anxiety feeling. Sure. I am just like, I just don't like it.

Speaker 1: I have an idea for your idea that I don't think I like. That I like this this is my denim to it. You could do the jukebox. K. But you could also gamify it. So on each table, it's almost like, you know, you can have, like, a Jenga thing.

Speaker 2: Mhmm.

Speaker 1: You could have artists, like, little blocks or something that have artist names and genres

Speaker 2: and Amiibos? Yeah. Little Amiibos.

Speaker 1: And everyone can, like, stack and be like, I wanna do Ragtime, sung by, Charli x x, like and start creating the that that weird mix and match thing, and then you play that over it. So it's almost like the music is intentionally chaotic and kind of cringe. Do you remember That could be something.

Speaker 2: Do you remember going to iPod battles?

Speaker 1: Oh my god.

Speaker 2: Do you remember these? I I feel like you were

Speaker 1: talking about.

Speaker 2: Yeah.

Speaker 1: This this would have been on the edge Yes.

Speaker 2: So, like

Speaker 1: for me. But

Speaker 2: Like like, hip little bars and stuff used to always have like iPod battles and it would be like two trendy businesses would go in and they essentially got the aux and they would just play sets and then they would compete and then people would choose a winner. You could gamify the same thing. Like, you go to a bar, and every table gets the ability to create one song, and then that gets played. And then That's pretty funny. Somebody at the end of the night gets voted. A leaderboard goes up.

Speaker 1: Yeah. Sure. It's very funny to think of, like, the bar that's main line item is not food or drink or staff or ambiance or furniture or API APIs. Just like API tokens. Just like we

Speaker 2: we boil a notion

Speaker 1: for a night of the weirdest music you've ever heard. It's, someone will do it. There there's a pop up working on it right now.

Speaker 2: Okay. I'm gonna transition us. Take us away. Because I feel like we've talked about this enough. But if you do make an AI powered jukebox, I expect full credits.

Speaker 1: You will not get it. Andrew Freeland will offer you nothing, Scott.

Speaker 2: In the exact same vein, there is a malware group that has stolen a bunch of art, and as part of the ransom, they are threatening to feed it to all of the major AI's learning data if they don't pay the ransom.

Speaker 1: Oh, wow. When you say stolen art, what do you mean?

Speaker 2: Like, they've stolen digital art from, like, they targeted a group called Artists and Clients, the big website that connects artists and clients. They hit it with a ransomware attack. Four zero four Media, Friends of the Pod reported this, and they have stolen all of the digital artwork from this site. And as part of the ransom demand, they're threatening to expose it and feed it into the learning datasets for large language models. Oh, wow. So that is now, like, a new threat to the artistic community to be like, pay your ransom, or else we're gonna essentially give your art style to the to the AIs. Yeah.

Speaker 1: I mean, it's kind of a limp threat at a certain point. It's like, oh, they're gonna scrape it anyway. You found it. Like, they're gonna get it eventually. Mhmm. Interesting. So

Speaker 2: Okay. Full circle this episode we talked about is, like, media creation and AI.

Speaker 1: Yeah. I mean, maybe it's just a theme episode. The other one was the, okay. Well, actually, the one other story I wanna talk about does have an okay. This is retroactively a theme episode. K. Sheen. Are you familiar with the fast fashion brand Sheen?

Speaker 2: Oh, yes. Yes.

Speaker 1: Yeah. Yeah. I was like, you you've s h e I n. Yeah. Isn't it Shen?

Speaker 2: Or is it Sheen? I think it's Shen.

Speaker 1: Shen? Shen.

Speaker 2: I don't know. I'd

Speaker 1: That brand.

Speaker 2: That brand.

Speaker 1: Did did you hear about the AI generated model? No. They they use the AI generated model to, like, sell a shirt, and it it seemingly inadvertently used, UnitedHealthcare CEO murder suspect Luigi Mangione. Scott stepped back to the mic when I said that.

Speaker 2: Oh my god.

Speaker 1: Yeah. So, you know, kinda just leave that one there. Just just, you know, a comedy affair with AI in this one. They took down the listing. You know, I mean, it's technically just his likeness, but that's a photo of Luigi Mangione in a shirt with some a little floral print on it.

Speaker 2: I'm looking at it right now, and that is shocking.

Speaker 1: Yeah. I mean, don't don't do that. And I'm so curious what the prompt was. But, quote, we have stringent standards for all listings on our platform. We are conducting a thorough investigation, strengthening our monitoring process, and we'll take appropriate action against the vendor in line with our policies. So they're saying that an external party, created the image.

Speaker 2: The thing the

Speaker 1: cross chatter of, like, handsome men and then a lot of comments of people saying, like, I find him handsome, and then it just, like, got in the soup or someone was just just did a just did a funny. Either that or not.

Speaker 2: Through. The diffusion had learned from public photos and because they were in the press so much, a licensed photo set of like, we work in marketing, so it's like I could see how you could get to this photo. You're like, I need a racially ambiguous person with curly hair and, like, you know what I'm saying? Like, it's kinda Yeah. Yeah.

Speaker 1: Yeah. Like, I

Speaker 2: can see how they got a prompt to this photo, and then it probably went digging through the it's learning set and found a photo of him. It was like, oh, this looks right.

Speaker 1: Yeah. A lot of people sure seem to be sharing a photo of this young man. He must be a good Totally. Could be. Could be. Or someone just asked it to put Luigi Mangione in that shirt. And it got through.

Speaker 2: Hey, maybe it's There's no bad PR. Right?

Speaker 1: This might be bad.

Speaker 2: Well, not now that we're stuck in a thematic episode as you've interactively made this thematic

Speaker 1: Yes.

Speaker 2: I don't really the next thing I was gonna talk about isn't on this theme, so I'm just gonna leave it alone. But, hackers have been sneaking malware into SVG images bypassing antiviruses because SV SVG images, which there are boatloads on on the Internet these days, are essentially just XML files.

Speaker 1: So Anyone that doesn't know what what is an SVG file? Like, what is it normally?

Speaker 2: So a scalable vector graphic is an SVG file. It's essentially a non raster vector image that can scale and resize dynamically because it's coded into XML, the colors and shapes and dimensions that you look at. So instead of seeing like a JPEG as a rasterized image, it's bit pixels. You know, every every pixel is like a value for the color, etcetera.

Speaker 1: Mhmm.

Speaker 2: SVG is more of like a dynamically generated one based on an instruction set that comes into this XML.

Speaker 1: Yeah. It's an image executed via math instructions as opposed to pixel values.

Speaker 2: Thank you. Yeah. Very, very helpful, very functional on today's Internet, especially when we're scaling between tiny cell phones and massive ultra wide monitors.

Speaker 1: It's vertical and wide. Yeah. Yeah. Useful.

Speaker 2: So the, but because they are just JavaScript or, sorry, just XML, they can embed things in them including JavaScript. So so, anyway, there's a group out there that have been stringing stuff up into SVGs that is bad for you.

Speaker 1: Okay. So I wonder what the takeaway there is. Don't open SVGs?

Speaker 2: No. No. No. I think they're so prevalent that I don't think you can't not. I think it's more specific around. It's just that it's happening. I don't think there's probably really any hard takeaway.

Speaker 1: Deal with it.

Speaker 2: Yeah. I'm sure I'm sure the the Microsofts and the Apple OSes of the world will begin filtering them for malicious tags and stuff like that. So if it hasn't already done been done. So

Speaker 1: Totally. I could see browsers being on that extremely quickly. It's like a it seems like the ideal delivery vehicle.

Speaker 2: Totally.

Speaker 1: The last one that I had here, and it is sort of still on, on theme was the story of prompt lock. So, prompt lock, this team of six professors and research researchers at NYU's Tandon School of Engineering put out this thing called, like, called a prompt lock. It's a proof of concept experiment built to, like, test ideas in a very controlled lab setting. They uploaded the program to a thing called VirusTotal, which is this Google owned, like, malware database

Speaker 2: Mhmm.

Speaker 1: Without explaining its academic origins, which caused a security company called ESET to flag it as, like, an active threat and issue a warning. So we got this big cycle about what is this word it come from, and then everyone figured out it was a research project. The thing that was interesting about it is that, they they called an orchestrator program. So traditional malware, it's like the attacker writes it it very care carefully crafts it to execute a very specific piece of malicious code. This being, like, designed as an orchestrator, the program doesn't contain all of the attack logic inside of it when it gets executed on the system. It delegates the planning and coding to a large language model inside of it. So basically, the malicious file runs on the like, runs the orchestrator. Instead of executing the precoded attack, the orchestrator communicates with the LLM inside of it using natural language prompts. The LLM generates the malicious code live. Everything needed for each step of the compromise based on what it's learned, which is then executed automatically. The orchestrators launched, the attacker steps back, the AI runs the entire process. So just like a different style of ransomware developed by this internal research team put up as, like, a demonstration of how it could work and a warning that gets this big response. Really interesting.

Speaker 2: This is smart research.

Speaker 1: Really cool research. Like, really interesting.

Speaker 2: Yeah. But, like, this is this is how you get things like COVID. You know? It's like I yes. It's like we're developing them in a lab to be prepared for when they happen. And it's like, hey. Now we have it. And it's like prompt lockers like, hey. We built this crazy LLM powered hacking toolkit so that we can start to prepare for the world that's gonna have it. Also, here it is.

Speaker 1: Yeah. A 100. And it's like they're they've built it in such a way as to it they don't have to break the guardrails of even, like, relatively conservative systems for this to work. Like, it's Mhmm. Each one of the requests is really, really compartmentalized to be kind of, like, harmless so that pretty much any LLM will reply. Like, say it scans the computer for valuable files, and it decides whether to, like, steal or encrypt or destroy them. And then, oh, write a little extortion note and then run a little bit of code to do that thing. None of those individual things of, like, delete some files are outside of the bounds of a system. Write a hypothetical extortion note is, like, you probably prompt something to get it to do that pretty easily.

Speaker 2: Yeah.

Speaker 1: It's cheap and pretty hard to detect. It I think it chat g p t five API rates, which seems above what this would need to run. It's 23,000 tokens to do, like, a full run of it, which is, like, 70¢. So the ability to create, like, bespoke ransomware on the fly live is, like, 70¢.

Speaker 2: And that's only if you're paying GBT 5. Yeah.

Speaker 1: Use an open source model that's small, and it's, like, you bring that cost down to basically zero and you get, like, unique attack code every single time bespoke live on the fly. Crazy. Yeah. This one was interesting. I almost I I wanted to talk to you about Velvet Sundown, but this one almost there's less

Speaker 2: to this one. Made the deadline. Yeah.

Speaker 1: There's less to it because it's just, like, you know, they made it. They put it out. People found it. They thought for a second it was in the world and then it became clear that it's a research project. But like you said, this does seem like how in six months, we're just dealing with, like and then, anyway, someone ran one of these inside of the Pentagon and or inside of Google or inside of a bank or something and, like, a calamity unfolds.

Speaker 2: Well, the so one of the other stories that I was gonna talk about is kind of a knock on to what you just said is the supply chain attack structures going on inside of, like, development libraries is out of control. Not just development, but development in kind of auto package manager libraries. So like MPM, they found supply chain malware attacks and 2,600,000,000 weekly downloads in MPM packages.

Speaker 1: That ain't great.

Speaker 2: So so now you've got all these vibe coded people, like, vibe coding things

Speaker 1: Right.

Speaker 2: Using these development packages that are coming with prebaked malware. That malware is about to be like a independent AI system that, like, maximizes the damage locally. Anyway, it is gonna be a really interesting decade.

Speaker 1: Yeah. I mean, we're already talking about stuff that I remember three years ago being, like, someday this is gonna happen. Yeah. I think we at some point talked about, like, there's gonna be an AI band that blows up, like, all these things. It's like, okay. Here we go. Yeah. So, you know, write down any predictions you make because we're probably gonna be covering it in a year or two.

Speaker 2: We don't need to write them down there archived in our podcast.

Speaker 1: This is true. This is true. Our podcast about fake AI bands and AI friends and ransomware powered by it's good, good stuff.

Speaker 2: I'm gonna add futurist to my, email tagline.

Speaker 1: Futurist crypto investor.

Speaker 2: Have you ever met somebody that calls themselves a futurist? It is such a fascinating the person who has the audacity to put futurist on their, like, email signature is a very specific person.

Speaker 1: It's in a world of made up job titles, somehow the most made up job. Like I, I predict the future be like that is you just call yourself a snake oil salesman. Like, just put that in LinkedIn, if you're gonna do that, like, if you're gonna just make an unverifiable claim, that strongly implies fraud, just commit.

Speaker 2: Wow. But what if you've got a proven track record?

Speaker 1: Totally. I bet technology is important in the future. Make sure it's Like, it's like when you go to see a psychic at a thing and they're like, yeah, they're like, oh, someone in the audience has lost a loved one. It's like, it's statistically inevitable. You're just making blanket claims. Their name starts with an r. It's like, yeah, there's only 26 letters. Come on. Like, it's just it's it's head games, man.

Speaker 2: Math. Math.

Speaker 1: It's math. Math. Anyway,

Speaker 2: I think we end there.

Speaker 1: I think we end there. Another one on the bucket. Appreciate

Speaker 2: it. Have enjoyed riding with these random after the break sessions. There's a lot of positive feedback. Also, from Jordan, I would like to thank everybody that liked and subscribed and preregistered and

Speaker 1: That was

Speaker 2: for the game.

Speaker 1: Final thing on my to do list for this episode was a big heartfelt thank you to everyone who wish listed and went over to the page after last episode. It meant a lot, that you guys A lot

Speaker 2: of positive feedback in the comments.

Speaker 1: Yeah. It was really kind. It means a lot. We got a lot of work ahead on that episode, but it was really nice to see it. Everyone was very, very kind. So it means a lot. We'll catch you in the next episode. I'm gonna be traveling over the next week or so. So hopefully, we've got a little bit more live reporting, for the next one, and we'll, we'll catch you catch you then.

Speaker 2: Yeah. I'm also gone after you get back.

Speaker 1: We'll we'll figure out when to record.

Speaker 2: We'll figure out what to do.

Speaker 1: I think you're back.

Speaker 2: And I'm going to Japan at Christmas.

Speaker 1: Hell, yeah.

Speaker 2: Hell, yeah.

Speaker 1: Okay. Maybe you do a little live recording.

Speaker 2: Yeah. Absolutely.

Speaker 1: A little reporting boots on the ground.

Speaker 2: Boots on the ground. On the that. That's fine.

Speaker 1: Yeah. Sick. Okay. K. Well, we'll catch you there.

Speaker 2: Take care, everybody.