Transcript
This is an AI transcription and may contain errors
Source: :contentReference[oaicite:0]{index=0}
Justin: Hello. Good morning, good afternoon, good evening, and welcome to The AI Argument with myself, Justin, and Frank, as ever in the People’s Republic of Cork. Frank, how’s your week been?
Frank: It’s been pretty good. Yeah, it’s been a pretty good week.
Did Claude Code just leak its secret sauce?
Frank: I wouldn’t say it’s been a great week for Anthropic. What would you say?
Justin: Well, they’ve done the world a great favour. By accident, they released all the source code for Claude Code, and now all the nerds on the internet have spent the week poring over the code to see all the tricks and hidden features that they have buried in the depths of the code.
Now, it’s not a big deal, everybody. There’s going to be loads now of open source iterations of Claude Code, but they’re moving so fast. It will be a different product in a month, and I think the speed is probably the reason why they released the source code by accident. You are worried about this, Frank. I’m not terribly worried about it at all.
Frank: Well, yeah, I wouldn’t say I’m worried about it on the level of being turned into a paperclip by artificial superintelligence, but I don’t think it’s great for Anthropic. They’ve been in the headlines for two reasons lately.
One is the big row with the Pentagon, and then the other is everyone fell in love with Claude Code and everyone was talking about how amazing it was and how incredible it was, and it was so far ahead of the pack. I don’t even know if that’s true because we’ve talked on the show about how it kind of goes back and forth between Claude Code and Codex. Codex is also a fantastic tool. But regardless, everyone fell in love with Claude Code and that was kind of their secret sauce there for a while, and now the secret sauce is not so secret.
Justin: Well, yeah. I mean, it’s a combination of the model and the harness, right? The two of them together for me. You are right. What I take from it is the human is the weakest link. Somebody made a mistake, and that happens when you ship stuff literally every 12 hours, it seems. They’re shipping a feature. Somebody’s got to make a mistake.
Frank: Okay. Let me ask you a question, though, right? Because they said this was a developer. I didn’t understand the details. I read about them, but a developer basically accidentally left a link to the source code in the code itself, in simple person’s terms. And they very much said, oh yeah, yeah, yeah, yeah, human did this. A human did this. But let’s face it, if Claude Code had done it, would they have admitted it?
Justin: Well, that’s a good point. That’s a good point. I don’t know if they would admit it. I also don’t think Claude Code would have done it, but that’s just me. But yeah, look, there’s far more worrying things that should have upset you that are kind of in the same ballpark during the week. They made a mistake, right? They also dealt with it very well, I thought, by the way. They put out public statements saying, yeah, it was a mistake. We make mistakes. It won’t happen again. So that was cool.
Will AI agents blindly install your next hack?
Justin: There were this week and last week, however, poisoning attacks in some of the public repos.
This is far more serious. So it’s nothing to do with Anthropic, right, just to be clear. But we’re just talking about automated processes and what should be automated and what shouldn’t be automated, right? And lots of people using Claude Code. So there’s a library, a Python library called Axxis on the internet. It’s very basic, but what it does is it allows your computer program to access the internet. One of the most popular Python libraries on the internet, if not the most popular on the internet, and some hackers managed to inject malware into this library, to put it simply. And so then if you downloaded it, it basically gave them complete access to your machine and they could take it over and download all your data and stuff like that, right?
Now, this is a big deal because Anthropic, for their problem, said, oh, that was human error, right? And we can fix the human error. But the problem is with coding tools, they just download packages off the internet as they see fit and as they need. So if a hacker is able to inject malware into it, your coding agent is just going to download the malware and install it without you ever knowing. This is a much bigger issue, right? That’s actually a problem. It’s a big problem. I mean, I wouldn’t like to be a security person at the moment because I think they’ve got a lot of headaches, especially for small indie developers and stuff like that. I think it’s a problem. Big companies will come up with ways to stop this, but small people won’t.
Frank: This is a really good point because I’m not a developer. However, I am curious about the tools, and so I like to sometimes give Codex tasks that I’m giving ChatGPT, just to see, do they perform them differently? Is the end result any better in one or the other?
So I’m talking about simple things, or relatively simple things. Recently I gave it a CSV of all of my LinkedIn contacts, and I was getting ChatGPT and Codex to go through the entire list and enrich data. ChatGPT just goes away and does it in the background. I don’t know what it does. Codex comes back to me and says, is it okay if I do this in order to complete this task? Because I have not given it carte blanche to do whatever it wants, which sounds like I’m being very safe. The problem is it comes back to me and it says, can I do this to complete this task? And I’m like, I don’t know what that is. I don’t know what you’re asking me. Go on then. Go ahead.
So it’s a really good point in terms of people who are not technically proficient using these tools. What trouble are we likely to get ourselves into? Because we’re not aware, oh, you know what, I probably shouldn’t download that library in case it contains malware.
Are we moving too fast with agentic AI?
Justin: Now let’s abstract it one layer more, right, just to make you even more frightened on a Friday. So you obviously now have smart people who are trying to hack into people’s computers, and they’ve identified a vector of attack as, let’s put some malware into some Python code that people are likely to download. But you and I know that you don’t need to put the malware into the Python code. I could just set up some websites with MD files that have malware in them, and I could make sure that they’re going to be found by Google or any of these people. And when your agent is going off, searching for some information, not even downloading a programme, just reading information that’s on the web, that could easily contain instructions which say, oh, take Frank’s bank details and send them to this area over here. And you could be sure that if people are smart enough to poison a Python library, they’re going to be smart enough to put web pages on the internet that are designed to misdirect your agent that’s running on your machine.
Frank: Yeah, absolutely. I keep coming back to it. I just keep coming back to, I think we’re rushing to the agentic age with reckless abandon that we should rein in a little bit.
Justin: Right. No, I don’t. These are just problems. We have to solve them, right? They’ll all be solved. The Python library thing, right? Anthropic did do a thing where they donated compute to all the open source projects, and a bit more of that, right? So you can have the AI scan the code before it gets published, right? So that would help everybody. The malware on the internet, I haven’t got a simple solution for it yet, but it will appear.
I wouldn’t. What you’re trying to do is you’re trying to ban gas cookers before we’ve invented safer types of cookers. It’s like, yes, it is dangerous. Your bacon is cooked when it goes on fire, and that’s okay. We’ll come up with a better way to cook the bacon.
Frank: Yeah, I would say it’s more like I’m trying to ban gas cookers that have a hole in the pipe, when we know they have a hole in the pipe. Let’s ban those ones.
Justin: Put tape around it until you get a better pipe. So look, there we go.
Did OpenAI kill Sora to end side quests?
Justin: OpenAI have had an interesting week. What’s going on with OpenAI?
Frank: Yeah, so we were talking there recently about how, first of all, they had a code red and they said they were going to focus on their core products, and then there was no sign of them actually really focusing on the core products, as in they were still doing this, that, and everything. Then they said, okay, yeah, we’re definitely doing code red now, but this time they said what we’re going to do is we’re going to stop chasing all these side quests, as they called them.
And I think on the last show we had, I was like, okay, yeah, I’ll believe it when I see it, because they’ve cried wolf before with their code red. So let’s see what they actually do. And they’ve actually done something. They’ve pulled the plug on Sora, Sora being the AI video model that they had. And we never even got it. It never even reached Europe.
Justin: Ah, there you go.
Frank: Well, yeah, it’s gone now. Well, it’s not quite gone, but they’re pulling the plug on it. It’s too much of a side quest. It’s too much of a distraction and, I believe, it’s basically using up way too much compute that they could be using elsewhere. What that does mean, by the way, is we talked recently enough about the $1 billion deal that they were doing with Disney and how they were going to have access to Disney’s characters and Disney were going to have Sora content on Disney+. Well, all that’s down the drain now because Sora is going to be removed.
Justin: And just to give you an idea, right, with Disney, from what I read, Disney didn’t know about this. They didn’t realise it was coming. They were told at like 12 o’clock in the day and then at five o’clock in the day the public announcement was made and they were just shut off. So it was a very sudden decision. It wasn’t like a planned out thing in terms of the two companies working together, which is mad.
Frank: It is mad. It is. It’s also kind of mad that the money apparently had not exchanged hands anyway. I just find it weird that they have all this fanfare, it’s all over the media, $1 billion deal, et cetera, and then when they pull the plug on Sora, it’s like, yeah, it’s fine, because we never actually finalised that deal anyway. It’s like, well, what was all that fanfare about?
Will OpenAI’s Spud transform the economy?
Justin: They drive me mental at the moment, right, because you will have seen, I’m sure, all the chatter about Spud, their new model. And this is a fresh pre-training run. It’s another step-jump in capabilities.
Frank: Best name for a model ever.
Justin: Well, I’m thinking of Trainspotting and Spud from Trainspotting. I’m thinking, well, if your model is as clever as Spud from Trainspotting, you know.
Frank: I thought it was just a nod to how smart the Irish are. And as we go for artificial superintelligence, we need to be using more Irish names like Spud.
Justin: Maybe it’s so energy efficient you can get two probes and stick it into a spud and that’ll power the entire model. That would be cool as well. But I’m just sick of the hype. I’m like, yeah, whatever, okay. Release the model already. And then I much prefer the way Anthropic do it, which is like on a Thursday evening, it’s just like, oh yeah, we just developed superintelligence. You can get it on the API and it’s gone live. And that’s it. There’s no hype. You don’t know about it. Feels better to me.
Frank: Spud, what was it? Spud is going to apparently rapidly accelerate the economy or something. Was it?
Justin: It had the capability to transform the economy, was what they said. But sure, they said that about the last model too.
Frank: It’s interesting that you mentioned the contrast with Anthropic. Anthropic had another leak recently where somebody realised that their content management system on their website was not secured. And so there was a load of material on the content management system which had not gone public, had not been approved to go public, and was available if you knew how to get to it. And so there was information about an upcoming model from Anthropic there as well. And in that case, it wasn’t like, oh, it’s going to rapidly accelerate the economy. It was, this model could be really dangerous because it’s really good at cybersecurity stuff.
Justin: Yeah, I mean, again, I fall on the other side, which is if it’s really good at cybersecurity stuff, use it for cybersecurity, right? Your attackers are going to use it. But at least you control the model, right? You can see when the attacker is attacking you. So that should give you the advantage. You’ve got, like, you can see around corners where they can’t.
Is Sam Altman the wrong CEO for OpenAI now?
Justin: But look, I know you’re sort of throwing a bit of shade on Anthropic for all of this, but Anthropic is a company that’s gone through enormous growth in the last, you know, even in the last six months. I mean, there was one week where their usage base was growing 50% week on week.
From a user’s point of view, I can tell you the problem with Anthropic is that their infrastructure’s creaking, so it seems, at the moment. They’ve had a number of outages over the last two or three weeks in particular, just because of the explosive growth in the numbers of people that are using their product. So there are growing pains, right? You’re hiring people. They’re new on the job. They don’t know exactly how all the systems work. It’s tough. So I would. They’re not acting. I’m contrasting it where OpenAI are doing stuff and they’re acting out of self-interest and hype and whatever, right? They’re making the froth frothier, right, to push up their valuation. Whereas Anthropic are trying to do the right thing as quickly as possible. So I would give them a bit of slack.
Frank: Yeah, I did hear that bit where you were talking about the infrastructure creaking. I came across a lot of people a bit unhappy with the rate limits being greatly reduced on Anthropic lately. On the OpenAI side, didn’t you say on this show, was it in the last episode or the episode before, I think you said that you thought that Sam Altman would be gone? That was your big prediction for this year.
Justin: By the end of the year, yeah.
Frank: Because for a company that basically became synonymous with AI, and ChatGPT was AI and probably still is AI for most of the general public, not nerds like us, and they did that by forging ahead and doing things that nobody else was willing to do. Now, I might argue that it was a little bit foolhardy and all the rest of it, but they did it. They stole a march on everybody else, and now they seem to be completely reactive. Everything is like, oh no, Google have done this. We better do this. Oh no, Anthropic have done this. We better do this. And I was thinking about what you said about Sam Altman and I was like, is it the case that he is just a brilliant leader for a scrappy start-up, but is he the leader that you need to be a global enterprise?
Justin: And to be fair, there are different skills. One is a skill to go from zero to one, right? So, you know, scrap together the resources, get people aligned, and build the first product that works. But then to go from there to have a global company is a totally different ask. And yeah, I think he’ll be gone for two reasons. One, I think the side quests are going to come back to haunt them, right? Huge distraction, a lot of money that maybe could have been spent on other things. And there’s a good chance that Anthropic are going to overtake them in revenue by the middle of the year, I think another month or two. And that will look poor in terms of investor confidence and so on. So yeah, look, we’ll see. We’ll see what happens. I see also they’ve canned adult mode, OpenAI.
Frank: Yeah, I heard that. I couldn’t find a definitive statement on it, but all the rumblings seem to be that they’ve decided they’re not going to do that either.
Justin: I think, given the AI psychosis that we’ve seen out in the general public, actually canning adult modes, for once I’ll agree with you, I think is actually probably a good idea. I’m somewhat glad that that didn’t happen.
Frank: Yeah, yeah, yeah. There’ll be an awful lot of romantic AI writers very disappointed with that decision, though.
Do Anthropic’s limits make Claude too costly?
Justin: Actually, could I just go back? Because we should talk about it a little bit, right? The rate limits on Anthropic, right? So just to explain, in case people don’t understand, right, so what happens with Anthropic is you pay your 20 quid, or your 100 quid, or your 200 quid, and then you get usage, right? And they do it by sessions or hours. So you’ve got a weekly limit and then you’ve got a five-hour limit, right? That’s how they set it up.
And they were growing so fast that during peak hours they didn’t have the infrastructure to handle all the requests, right? So what they did was they reduced it, almost like what they call flash pricing with Uber. So during peak hours they basically reduced your limit so that you would get to that limit much quicker. And for a developer that’s really frustrating. Like you’re in the middle of doing something and then it’s like the lights turn off. I was in the flow. I was in the flow, and you stopped me, and it’s really, really annoying. They’ll fix it, though. I’m confident that they’ll fix it.
Frank: I only realised, when looking into this Anthropic rate limiting, that apparently if you have a 20-a-month subscription to Anthropic, say, it’s still token-based. If you have a $20-a-month account with ChatGPT, it’s actually message-based. And so you are much more restricted on Anthropic because if you’re chatting with ChatGPT, you could have an essay in your prompt and it still just counts as one message. Whereas in Anthropic, if you have that essay in your prompt, that’s using up way more tokens. So you’re using up much more of your allowance on Anthropic. I thought that was interesting, that that’s why the ChatGPT limits seem so much more, that you get more for your buck.
Justin: Yeah, yeah, yeah. More with that. But I mean, also, it’s also the reason why Anthropic actually make some money one day, whereas OpenAI won’t.
Frank: True, true. Yeah. So we’ve talked about what’s going on with Anthropic.
Did Google just make local AI much more viable?
Frank: We’ve talked about what’s going on with OpenAI. Google, what’s going on with Google? Because in my feeds, they’ve been a little bit quiet lately.
Justin: Well, not so much, right? So Google did two things that are, one is directed at killing Anthropic and OpenAI, and the other one is directed at reducing the price of memory for themselves. So memory, as you may or may not know, Frank, and I know that you keep a close eye on the memory markets to see what the price of memory… The price of memory has gone through the roof, through the roof. Sam Altman, I think, had signed up to buy 50% of all the memory for this year 12 months ago or so, which is a deal he’s welched on as well. But the price of memory still has been going up.
So Google this week released some research called Open Quant. So what this is, I won’t get into the details, but it’s a method where when you’re serving a model, you can reduce the amount of memory it uses by six times. So a 6x reduction in the amount of memory that it uses, and it speeds up the amount of tokens you can get. So an 8x increase in the amount of tokens.
Now, this is kind of a weird piece of research to publish because you would imagine that it’s commercially sensitive. That’s a big deal, right, to be able to do those things. So there’s only two possible reasons that I can think of. One is everybody’s doing this anyway, and the researchers want to be the first to publish the paper, right, so they get the prestige and the kudos for doing that. Or they just want to reduce the price of memory. So if they have a need to buy a whole load of memory, well, the best thing you can do is release some papers that reduce the price of memory. That pays for the research and a whole lot else as well.
So that was the first thing they did. Second thing they did, and it made me really think, is they open sourced a model called Gemma 4. So Gemma 4, in its biggest incarnation, apparently is on a par with Sonnet 4.6. Even if it isn’t, this is a big deal because similar models will be on a par with Sonnet 4.6 and others in another six months or so, and you can run these models on a Mac Mini. They’re small and they’re very, very powerful.
And I think the commercial reason they’ve done this is maybe to have a pop at Anthropic and OpenAI, right? So yeah, you can pay them for your models or, look, here’s one for free, and you’ll get a large number of people doing that. But it made me think, right, that for the last two or three years we’ve been climbing up this intelligence curve, right? And so every time a new bigger model came out, you had to go for that one. So everybody’s using Opus at the moment, or they’re using the latest one from OpenAI because they need the raw intelligence.
But there comes a point in that curve where you don’t need the raw intelligence any more, right? I don’t need superintelligence to do a search of the web and do some simple task. And I do think that for large organisations, sensitive organisations, healthcare, hospitals, let’s say, for the sake of argument, hospitals don’t want their data going outside of the hospital if at all they can stop it, right? There’s all sorts of risks with the data leaving your ecosystem. So if you can get a couple of Mac Minis and you can run a model in your hospital that’s as good as Opus, I think there’s going to be a big market for that. There’s going to be companies that want to run the models locally so that the data never leaves their networks.
And once the models become smart enough, and what I’m saying is I think in another six months or so they will be smart enough, well then that has an interesting commercial impact on OpenAI and Anthropic because I don’t need to buy my tokens from OpenAI and Anthropic any more. I can just install the server with a couple of GPUs and I can just serve the models myself. So that’s kind of an interesting thing. And then I was also thinking that it seems like a terrible idea, like it’s a fast depreciating asset. Well, the hardware might be fast depreciating, but the models are getting better every year. So I can probably upgrade the model on my hardware and never change my hardware at all. So maybe it’s a very wise investment. So that’s all the stuff that’s come out of Google. It didn’t make a lot of fanfare, but it certainly made me think.
Frank: Yeah, it’s funny that the Gemma didn’t actually reach me at all. I’m surprised. And I’ve lost track, right? So to the best of my knowledge, have Anthropic ever released an open source model? I don’t think they have, have they?
Justin: Not to my knowledge, no.
Frank: OpenAI made a big fanfare, as they do, that they were going to release open source models. And I think they did, but I don’t think they were brilliant, and I can’t remember the last time they open sourced a model either.
Justin: I mean, it’s an age ago. Probably three months.
Frank: Yeah, yeah. In AI terms, yeah. So yeah, it is interesting. And I wonder as well what… I don’t know what the primary driver for Google releasing it is. Maybe, as you say, it’s just to have a swipe at their competitors. And if you’re going to have an open source model that is good up to a certain degree, and then you want a more powerful model to do something else, are you going to use the same brand of model? And is that an advantage? I don’t know. Interesting, interesting.
Justin: A pattern, by the way, just if people are interested in it, a pattern I’ve heard a number of people using is that you go to the really powerful model, like an Opus model, to come up with the plan for the thing that you want to do, and then when you have the plan, you give it to your local free, essentially free, model to go ahead and implement the plan. And I’ve heard a number of people get good success with that type of pattern.
Frank: Sure.
Is Bernie Sanders right about stopping new AI data centres?
Frank: I’m going to bring this up now because I didn’t bring it up earlier. We only have a few minutes left. I didn’t bring it up earlier because I knew we’d spend the whole episode talking about this, so I’ll bring it up now and we’ll just do, like, a quick 60 seconds on it instead. But Bernie Sanders and AOC have put forward this bill to put a moratorium on the building of data centres in the US. What do you think?
Justin: Well, I mean, I think it’s a funny situation in America that you have one party appears to be working for the Russians and the other party appear to be working for the Chinese. I mean, who’s it going to benefit by not building data centres? It’s ridiculous.
Frank: Well, I mean, Bernie brings up all of the things that we talk about on this show. He brings up the simple facts that they’re so energy hungry and so water hungry, and a lot of…
Justin: Water? They’re not water hungry. That’s ridiculous. I mean, they use a teaspoon of water for every whatever hour that they run.
Frank: Yes, but the entire data… I mean, I believe there have been problems in neighbourhoods where data centres have been built, because that might be the case for one query or whatever, but when you have an entire data centre, it kind of hogs the water for the area, I believe. I’m not… And so there’s that. Then he brings up the potential job disruption that’s coming. He brings up the existential risk, which we talk about a lot here. And so he brings up all of the issues that we talk about all the time here. And he’s just saying, look, none of these conversations have happened. We haven’t had a proper conversation about what we’re going to do about any of this stuff. Those conversations need to happen, regulation needs to happen at a federal level, and until we have that, we should stop building the data centres.
Justin: He is literally the one who makes the rules, right? So why are you stopping? Make the legislation, have the conversation, do whatever it is, but don’t be stopping the companies building the data centres. Go do your job. Why would you stop somebody else from building data centres, which is their job, because you’re not doing your job, which is to make the rules that will manage the output from the data centres?
Frank: Yeah, I think his point is that if we stop to have the conversation, let’s just use existential risk because that’s the fun one, so if you stop at a government level to have the conversation about existential risk, but you allow the thing to continue being built over here, you may come to the conclusion that actually the data centres should be stopped, and meanwhile artificial superintelligence has actually been created over here and is turning us all into paperclips before we can do anything about it.
Justin: You see, I think existential risk, I just don’t think that there is an existential risk, right? Or it’s so small that there’s probably more chance of a meteor hitting the planet than existential risk. Economic disruption, however, I think is a real thing, and I think we’re going to start seeing it very shortly. And so it’s good to have that conversation. But you can’t stop progress, right? What are you going to do if you stop building the data centres? You’re just handing the economic disruption…
So let’s lay it out, right? Let’s say the economic disruption comes because I can now replace all office work and all robots with AI, right? And you say, well, do you know what, now until we sort this out, no more data centres. None. Okay? We need to figure it out. You can all go off and spend two years talking about it and thinking about it. In the meantime, there are other countries in the world that will have gone through all the job disruption of replacing all the office jobs and the robots and stuff like that, and they’ll be selling you some stuff for peanuts for next to nothing, in which case you’re still going to suffer from the disruption. It’s just not happening within your borders, it’s happening outside of your borders. You can’t insulate yourself from the disruption, so you might as well own it. You might as well let the data centres be built, and Bernie, go make the rules or talk about it or come up with a plan. It’s great that he’s having the conversation, by the way. That’s good. I just think that the initial reaction to, oh, stop doing that thing, is the wrong reaction.
Frank: I would be curious. I think he’s earnest in his desire for the moratorium, but I also think he probably knows full well that there’s no way that this bill gets passed. You’d have to be completely unaware of the current administration’s attitude to AI and technology and business to think that this bill would pass, which I doubt he is. So yeah, I agree with you that I think the huge benefit here is that he’s bringing the conversation into the spotlight. He’s making the conversation happen at the highest levels, which we’ve been saying really needs to happen since we started this show. Hopefully he is also maybe providing cover for people who might be proposing more palatable legislation and regulation, and maybe also for state-level preventing data centres being built at a state level if it is genuinely not in the best interest of the local community.
Justin: Yeah, it does. I’ll tell you what it does. I agree with you in terms of it’s good to move the conversation forward. And it does make me wonder if part of the legislation, if I was to look at the legislation, could you do cool things, right? It’s a whole new world, right? And if you’ve got a lab that creates ASI, right, do you pass a law which says, well, 49% of your company has to be owned by the public. Now, define the public, right? Maybe it’s everybody on the planet. Maybe it’s everybody in the country. You can argue over that, right?
But this is such a key piece of technology. I guess where the thought is coming from is I don’t trust the government any more than I trust AI, right? So I don’t want the government to be controlling it or the government to be making rules. But I do think that if you want to pass benefit and some ownership back to… Well, we just give everybody some shares, right? And then I can also imagine in this new future, this AI may be making a lot of decisions, right? What if everybody has a share in the AI? It’s very easy then to place votes on certain things or to give feedback to the AI. It’s really good at talking to people. And maybe if there was some sort of question over, should we build another data centre, let the AI send out a question to everybody. Should we build another data centre? Yes, no. You get a much more reactive feedback loop than waiting for four years for Bernie Sanders or…
Frank: Yeah. Didn’t Sam Altman at one stage promise us a future where we’d all own a slice of the compute, which is kind of-ish similar to what you’re talking about? Yeah, so to…
Can an AI metal band really go on tour?
Frank: Final story to finish up on, to get away from existential threat and economic disruption. Here’s a really fun little story.
This guy somewhere in Europe used Suno, which we’ve talked about before, which is the AI music generator. I love it. I think I was telling you before, I actually have an EP on Spotify thanks to Suno, and I’m officially a songwriter registered with IMRO in Ireland. And that was all thanks to AI music generation.
So this guy created a fake Japanese metal band and, unlike my EP, it went viral and some of the songs had millions of views. So the AI-generated Japanese metal band is now going on tour.
Justin: How?
Frank: He actually went and he hired Japanese musicians to perform the music and they’re doing gigs.
Justin: Unbelievable. And do you know the name of this band? It should be called Spinal Tap or…
Frank: I have absolutely 100% forgotten the name of the band.
Justin: They should call it Suno Tap. Isn’t that sort of…
Frank: Neon Only, Neon Only. Not Suno Tap. Suno Tap sounds like some kind of thing for your nose.
Justin: There you go. Is he making money?
Frank: I don’t know. I was not able to find out if the gigs had been a success or if people had attended or anything, any follow-up. But I’m going to keep an eye on it. Hopefully there will be more to report on that.
Justin: Look, good. Frank, you’re onto your difficult second album, but who knows, maybe this will be the breakthrough.
Frank: Justin, pleasure as always. I’ll chat to you again next week.
Justin: Alrighty, take it easy. Talk to you next week.
Frank: Bye.
