Transcript
This is an AI transcription and may contain errors
Justin: Hello. Good morning, good afternoon and welcome to The AI Argument.
Is Matt Schumer right about “something big”?
Justin: There was an essay written this week by, who wrote the essay?
Frank: Matt Shumer.
Justin: Matt Shumer, and this was an essay where he was positioning this essay as, look at some people live in an AI bubble. You live in an AI bubble. I live in an AI bubble, but 95% of the world do not live in an AI bubble. And he was trying to explain to them what he thought was coming in the near future, saying, look at change is coming. It’s big. It’s like a tsunami that’s gonna wash over us all. And you should start getting prepared for it. Now, do you agree with this?
Frank: Well, that’s a really tough question. I think it was really interesting that this article went so viral because we are immersed in this and so I guess, since whatever it was, since 20, the end of 2022, we’ve been like, something big is happening. So I understand why he’s saying something big is happening now because there are a lot of changes. A lot of people are experiencing that. The coding agents are getting a lot better. I’m not a coder. I’m not having that experience directly, but I’m hearing about it a lot. So if you’re not immersed in AI, I think it’s worth reading that article to hear one person’s experience of, wow, something big is happening, these agents are gonna be able to do my job and we have to start paying attention. It’s worth reading, absolutely, for sure. But I…
Justin: Can I just come in?
Frank: Yes.
Justin: Well, I would say that it’s not just that. So you’ve focused on the coding area, but I think the change that’s coming now, that’s right here, it’s here already. There’s this huge technical overhang. If you look at Claude Cowork, if you look at Claude for Excel, if you look at the legal plugins, if you look at the marking plugins, if you look at the, you know, there’s a graph that I saw yesterday where they had the domains where I was having the biggest impact. And everybody knows it’s having a huge impact in coding, right? But it lists all the other domains, and they’re all going to be impacted the same way over the next 12 months. And I don’t think it’s obvious to a lot of people just how big that impact is gonna be. I saw another, I saw another, there was one of the Collison brothers giving a presentation yesterday in London, and they were talking about the number of pull requests that each developer did, whatever, blah, blah, right? They were just talking about the efficiency of developers, and they had seen from AI that their number of pull requests had increased dramatically, and that if you were to put a value on it, it would be equivalent to $150 million worth of developer time. Now, when it comes down to the impact, right, it’s value created is the only true measure, right? It’s not really pull requests, it’s value create. So they didn’t have to hire a single extra developer, and yet they were able to turn out as much code as if they’d hired an extra, I think it was 800 developers or something like that. That same transformation is coming this year, ’cause the technology already exists to marketing, to find, to legal, to all of the other domains where if you work in front of a screen, there is an agent that already exists, you just haven’t started using it yet, which is gonna fundamentally change the way you work.
Frank: I think that Matt Shumer’s article overstated that perspective a little bit because I think it’s not necessarily that anything he said was incorrect. He talked about being able to prompt an AI agent, go away for a couple of hours, come back and it’s done the work and it’s done it potentially better than he would’ve done it himself. I’m sure he’s had that experience. But I’m also sure he has had the experience where he’s gone away for a couple of hours and he’s come back and it doesn’t work and it’s broken seven other things, and so it’s the reliability that he completely overstated in my opinion right now. And so that’s why I would…
Justin: What matters though? The direction of travel, right? So if you were to ask me last September, was that article overstated? I would’ve agreed with you, right? I did think it was overstated. It was still difficult to do things. Everything changed in December, when the latest suite of models was released and the harnesses to use them, everything changed. And all you’ve gotta do is go forward in one or two iterations. The hallucinations have been going down all year. The models are getting better and better now. The tools are there to make them much, much better. The reliability, that meter of how long it can operate autonomously, is literally on a hockey stick, and you have to aim for where it’s going, not just for where it is today.
Frank: But you made a point there about the thing that the Collison brother had said about, and you said they didn’t have to hire any more developers to get that result. And I think that’s part of this argument, that you still, if you read the Matt Shumer article, you might imagine that he could be replaced, but he can’t be replaced yet. You still need that developer. I am not gonna be able to get those coding results using those tools because I don’t understand coding. I won’t know exactly how to prompt it and I won’t be able to look at the output and say, oh well, almost, but you need to fix X, Y, or Z, and so this is why.
Dario Amodei vs Gary Marcus: who’s right?
Frank: Okay, so this is why basically that I would say to people, read the Matt Shumer article, then go and look up Gary Marcus on Substack, your old buddy Gary Marcus, Justin, and read Gary Marcus’s response to it where he is saying no. As you and I know, Gary Marcus thinks that LLMs are a dead end, and he talks about the unreliability of the models and the fact that it’s gonna replace jobs at a kind of a role level is completely overstated. And I think if you read both those articles, then if you’re not immersed in AI like we are, you get a really good indication of where we are right now, which is that there is this huge disagreement about where all this is going. And it’s not, you know, Matt Shumer has come under some fire because he previously had an LLM that he said performed really, really well, and then it didn’t, and people were saying that it was, you know, that he jed the numbers and that he’s not trustworthy, et cetera, et cetera. And I don’t know, I don’t know about that. But what I do know is you’ve got people like Dario Amodei saying very similar things. He’s saying there’s going to be a white collar blood bath in terms of employment. And then you’ve got people like Gary Marcus. Let’s also talk about, you’ve got people like, say, Yann LeCun. You’ve got people like, who else, maybe Ilya Sutskever to a degree; they’re not buying into the LLM vision. So it’s very hard to tell where we are because you’ve got these two completely opposed viewpoints. One is that there’s going to be this blood bath and the other is that they’re just not gonna be capable of taking over a role. I agree that you can achieve a lot more as an expert individual with AI, but I don’t think we’re at the point where it can replace a role as…
Justin: You ever think, do you ever think that Gary Marcus might turn around one day and go, geez, I was wrong?
Frank: Yeah, and this is the…
Justin: He’s never gonna say that, right? He’s never gonna say it. It’s impossible to get a man to understand something whose paycheck depends on, depends on not understanding it, right? That’s just fundamental wisdom.
Frank: If the proof is there, I think he will.
Justin: He said last year, last year that we wouldn’t be able to solve hallucinations. Now, models still hallucinate, but the harnesses around have reduced those hallucinations. So we’ve made enormous…
Frank: He was correct.
Justin: He wasn’t, though. Hallucinations are less now than they were last year. He said it would…
Frank: Yeah. But he said they wouldn’t be solved. They’ve been reduced, not solved. And in a way that makes it more dangerous, the more they’re reduced but not solved, the more dangerous it is because we trust and trust and trust and trust and then we get a hallucination and we just trust.
Justin: That is it, by the way. I do agree with that. And the other thing he said was it was gonna be a total, it was a bubble. It was gonna be a crash and there might be a crash, but there hasn’t been a crash yet. This is the old joke about economists, that eventually they’ll always be right and there may be a crash at some point in the future, but there hasn’t been yet. And he said there would be. So I just, I can’t take him seriously, in that I think he has an opinion and he has a bias and he’s just pushing his opinion and his bias as opposed to looking at things objectively. Now I would…
Frank: I can’t argue with that. And…
Justin: You would hear…
Frank: That’s…
Justin: Go on, sorry.
Frank: Well, no, just that’s the huge problem that we’re in right now. This is what I’m talking about, this uncertain moment in time where you have these experts who have completely diametrically opposed views, and who do we trust? And they all have skin in the game, they all have a, they all have a reason to say what they’re saying. Dario Modi is the head of Anthropic. Of course he’s going to say it’s so good that it can replace jobs. IA Kerr and y Koon are looking at their world model or their completely different approaches to AI. So of course they’re gonna say, yeah, LLMs aren’t the way. Everyone has a reason to be saying what they’re saying. And I don’t know. I don’t know who we can absolutely trust. We just can’t absolutely trust anyone. So I would agree with you there.
Is “closing the loop” the real tipping point?
Justin: I think the phrase I would pay attention to that a lot of these people talk about is closing the loop. So they talk about closing the loop. The AI companies want to close the loop internally. What they want to do is have AI that’s able to develop the next AI, learn from its mistakes, then develop the next AI, and it’s sort of a virtuous circle, right? And they believe that they’re getting there. They already have harnesses which do a lot of the work for them already. So I think the same applies to all these other domains, like when you can close the loop in finance, when you can close the loop in marketing, when you can close the loop, you know, so for example, you’re a small software development company and you get a request from somebody that says, I’d love if you had this feature or if it did something. Basically autonomously, it picks it up, evaluates it, does a bit of market research, does all the things you would do as a human, implements the software. Maybe then it gives it to the engineer to say, right, I’ve implemented this feature, it was asked for this, here’s my research and here’s the finished update. Give me a go or no go. And you go yes or no. When you say go, it goes and promotes. That’s closing the loop in my mind, happening, right? It’s not there yet, but the direction of travel is so obvious that we went from, and I think this was, you know, the something big is happening, this was ’cause it’s my lived experience too, a year or two ago, you would type in, write me a function to do this, and you would get a piece of code, you would copy the piece of code, you’d put it into your code editor, you’d run it, it wouldn’t work, you’d take the error, you’d paste it back and you’d go over it, back and over it back. And I remember a good design principle at the very start of this, in fact up until last November, was if you’re writing code with AI, use what was the old Unix principle, which is every function should have a very defined purpose and be very small. And so the AI would just focus on little small chunks of code and that would work, right? That all changed in December. Now you can just give it a whole code base and say, go make this change, go make that change. And it isn’t always perfect. You’re right that it does hallucinate, but it is so different than what we did last year that, if you just follow the curve, do you know what the futurists do? Follow the curve. It’s the, you know, Gary Marcus is totally off base on this one, and he’s just, he’s keen at, he’s chasing the god of the shadows, right? Oh, but you don’t do this little thing. Oh, but you don’t do this. Oh, it does this wrong. And he’s not wrong. It does make mistakes and do all these different things, but it makes far less mistakes than it did a year ago and a year from now it’s making far less mistakes again, to the point where it will make less mistakes than a human. And that’s all that matters.
Frank: But if we look, well, okay, so let’s not dive too deep into this, but I would say it does matter because there are different types of mistakes than a human would make. But also I would say that the curve that you’re talking about, or the path forward, I don’t think it’s a given. And I think we’ve seen in the past, we’ve made huge leaps forward with AI, but we’ve seen in the past that AI has had that pattern of making huge leaps and then we think we’re nearly there at AGI and then it stalls out and there’s what they call an AI winter and nothing happens and investment dries up and whatever. And I just don’t think that that’s impossible. I don’t think that what you are talking about is impossible either. And that’s exactly why I just think we’re in this uncertain moment. What I do think, and this is the same thing that I’ve been saying from the very beginning, so I’ve probably said it on the show a few times, but I think what we know for absolute certain is that the labs are working very, very hard on creating a technology that will absolutely be able to replace human roles. And so to me what that means is, you know, Matt Schumer is saying, learn the tools. Great. Learning the tools is a great thing to do for right now in this moment in time. But in terms of the future, I don’t know whether the labs will get there or not, but I definitely think that at the highest levels our governments should be considering what on earth are we going to do if they do get there? So I’m not convinced that this is the tipping point that Matt Schumer perceives it to be, or that I think maybe you perceive it to be. I’m not so sure that it is the tipping point, but we need to be thinking about, well, ooh, what if it is? I’ll acknowledge that much.
Justin: So, I mean, serious data vu, like we literally had this conversation two years ago and I don’t know which one of us said it, or what we were, the exact frame of the conversation was. But I think at the end, the conclusion was that until a lot of people lose their jobs or there’s some sort of financial turmoil or something big happens, governments tend not to do anything. They just sit around and go, well, it’s not a problem today, so I’m not gonna do a whole lot. And until it actually becomes a problem, then they’ll start to think about it. And I do think that the thing that we do all agree is like, that’s probably not a good idea in this instance. We should probably start thinking about those things now and talking about them.
Frank: Well, yeah, absolutely. We should have been thinking about them in 2023.
Should Anthropic defy the Pentagon?
Frank: And speaking of huge problems, Anthropic have a couple of problems that they’re dealing with at the…
Justin: Hmm. They do. So they have one of their lead researchers has decided to leave and become a poet. And in totally unrelated news, and not at all linked whatsoever, the Pentagon has said that there’s gonna be serious consequences for Anthropic when they refused to allow their software to be used for valence of American citizens. Do you think, Frank, that at some point in the future, who knows how far in the future, they may become a limited liability company in, let’s say, Dublin or Paris or Frankfurt, somewhere European, where perhaps the…
Frank: They do certainly seem that little bit more aligned with our approach than, say, OpenAI. So yeah, being threatened by the Pentagon, I mean, this seems huge. So Anthropic, they signed, I think it was, I don’t, I can’t remember when, but they signed like a $200 million deal with the Department of Defense as they were called at the time, now rebranded as the Department of War.
Justin: Of posture there, but there we go.
Frank: So, and apparently Claude was the model. They signed, I think the Pentagon signed deals with most of the companies, but I believe that Claude was the one that they trusted with their most secure information, their most highly classified information. But there’s been some tension between them around what they can actually use Claude for. You mentioned mass surveillance, so apparently that’s one big sticking point that Anthropic are saying; Anthropic have pretty tight rules and what I read was that they are willing to kind of open them up a little bit for the Pentagon, but their two kind of hard lines are mass surveillance on American citizens and fully autonomous weapons that could fire without a human in the loop.
Justin: Yep.
Frank: They seem to me to be lines that are pretty reasonable, but the Pentagon are making the point that legally speaking, there’s nothing stopping them from doing that. And so why should Anthropic be able to stop them from doing that? I believe they’re now not making it black and white either. They’re saying, look, there are gray areas here that we need to consider. They’re not just saying, we absolutely want to mass surveil the American people.
Justin: They will. Make no mistake, if the technology exists, they will. That’s their job essentially. And this is just a better technology to do it. I think this is good for OpenAI and it’s good for Anthropic. Why, you ask me, why is that the case? Because you’re gonna end up with different companies with different principles and companies should be allowed to do what they want to do. And so if I’m a private citizen and I’m okay with the government surveilling me, well then I can graduate and go to the OpenAI ecosystem. And I know that I’ll be surveilled more than likely. However, if I don’t want that, and in particular, I can imagine that there will be certain companies, large corporations, and then individuals as well, who say, I don’t wanna be just, I wanna use, I wanna work with a company that have principles around how data is used. And if you’re working with one company, let’s call it, I don’t know, company one, I’m not even gonna say who the company is, and they’re working very closely with the American government, you would have to think that if they’re embedded with the American government, my data could be used by the American government. Maybe, maybe not. They’ll say that it’s not and all that sort of stuff. But if they’re happy to have their model used in certain ways, that would have to be a consideration for me. And then I have another company who has said, no, in fact, we got kicked out by the American government because we have principles about how your data’s being used. That’s a huge marketing coup for that other company. If you’re looking to do deals with people, I think at some point there will be people going, hold on a second, I don’t want my data being used in particular ways, even in countries like America where the rules are much looser. I still think people think that way, but definitely for companies. I think this is, you know, it’s a badge of honour for Anthropic to say, no, we are responsible, we have these rules, and we stand by these rules.
Frank: Yeah, and if I remember correctly, we’ve seen this before a little bit in terms of, I’m drawing on my memory here, but I think if I remember correctly that it came out that Facebook, as they were called at the time before they rebranded to Meta, had built back doors into their systems for the American government, and it came out that Apple had absolutely refused to do that same thing. So we’ve seen this kind of thing. It has happened before and I agree with you generally, but what I would worry about is apparently the Pentagon have threatened Anthropic with classifying them as a supply chain risk. And apparently that is unheard of, to threaten a domestic company with that. And what it would mean is that anybody who wanted to work with the Pentagon would have to prove that they had cut all ties with Anthropic and did not use Anthropic in any way, shape or form. And the Venn diagram of companies that work with the Pentagon and companies that Anthropic might want to sell to is a pretty huge intersection. So my worry would be, would they survive something like that?
Justin: I mean, only Anthropic can answer that question, and I don’t know, and I haven’t seen that Venn diagram. And again, it makes me think, I would hope that Anthropic would stand on their principles, right? It’s a dangerous situation where a government can bankrupt a company because they don’t do what they want them to do. So that’s not a situation that you want to be in. I think that there’s enough business in the world that a company like Anthropic could be very successful not doing business with a specific government if that government wants to use their models in a way that they didn’t agree with. And there is a country on the western periphery of Europe which has a very low and generous corporation tax rate, has almost all American multinationals with head bases, and it’s a huge concentration of talent. And listen to the accents of the researchers when you listen to a podcast or a webinar or something like that from any of the AI companies; they don’t speak with Californian accents, they speak with accents from the entire world. And so California is good because there’s a concentration of AI companies there, but I think a company like Anthropic would do just fine in Dublin or Paris or Frankfurt. And I think, again, if it comes down to, and this isn’t, we don’t talk about politics here, but it comes down to a very, if there was an enlightened European set of governments that sat in and said, well, look, we will make up for the shortfall in the revenue that you’re going to lose if you come and set up a base in the European Union, we will then use your models and you’ll make the same revenue as you would’ve done in America, and then they have a place where they can operate according to their principles. I don’t know that that’s happen, but it could happen.
Frank: It will be very interesting to see what happens, ’cause I mean, obviously one way or the other, it’s ferocious pressure. On the other hand, you’ve got Dario Moai just published that article in which he said our own governments should not be allowed to use AI any which way they choose. But you mentioned that they lost their lead safety guy. He resigned from Anthropic, and one of the things he cited was the pressure that they come under to abandon their values. And this is a clear case of that, massive, massive commercial pressure to ditch the values that they have about how AI is used in military and government settings. So yeah, really, really…
Should AI ever pull the trigger?
Justin: Have you seen, have you seen the videos of the Chinese robots this week?
Frank: I didn’t, I haven’t actually watched them. The kung fu, the coordinated kung fu videos.
Justin: That, well, maybe, and who knows how real they are these days, ’cause you worry about this stuff. But they literally had humanoid robots with guns running through fields and they were shooting the guns.
Frank: Oh, those I did not see. No. And yeah, so that hard line of Anthropic’s, like no autonomous AI that fires a weapon without a human in the loop. If you take that hard line and then you think about a story that we talked about just a week or two ago, where the cops arrived, guns drawn, to apprehend a teenager who was brandishing a gun and it was in fact a bag of Doritos. Now, if that was an autonomous AI agent with the ability to shoot, if you put those two stories together, that hard line doesn’t seem like it’s one that we should be crossing.
Justin: I think that’s a great point, by the way, ’cause I read somebody else and they were talking about, I think that that’s where the position Anthropic specifically don’t want to get into. And when there was this, there was a lot of heat on Anthropic this week because Open Law, Anthropic stops people from using their plans with Open Law. So if you had a $200 plan, you still weren’t allowed to use it with Open Claw, a lot of heat from the developer community. But there’s also speculation that Open Claw is gonna do something that’s gonna harm somebody. So it’s gonna do something silly. They’re gonna put it in control of something and somebody’s gonna get hurt. And Anthropic just do not want to be associated with that. And so you can talk about, you know, countries don’t have morals, they have interests, or they don’t have principles, they have interests. And the same is, to a certain extent, true of companies. They may say that, oh, we don’t want it for this reason or that reason, but underlying all that, there might be just, look, bad things will happen with this technology and we don’t want our name associated with that. And that’s a reasonable position to take too.
Frank: Yeah, and it’s a good point because another story that we had on our radar was the fact that Peter Steinberger has been hired by OpenAI, so he’s the creator of Open Claw. We’ve been talking about him a lot over the last couple of episodes. We’ve been talking about the Open Cloud agents a lot. And there was a story that I would love to have done a deep dive on, but we just didn’t have time this week. But there was a story of an Open Claw agent writing a hit piece on a human who rejected their Git submission. And that particular guy who was looking after an open source Git repository, he said, yeah, you can see the funny side of this, but also what if it had uncovered something about me online that it could actually have blackmailed me with? And what if a future AI system, like he made the example of, what if I…
Justin: What if you had been a former member of the Royal family?
Frank: He was saying, like, what if I go for a job in the future and someone asks ChatGPT to do research on me, and it sees this essay from another AI agent saying that I’m a hypocrite, that I’m anti AI, blah, blah, blah, blah. How might that impact me? I still think, yeah, as you’ve said, Open Claw is, there are so many risks involved and I’ve seen a lot of people saying, oh, because Peter Steinberger is a big fan of Claude. A lot of people were saying, oh, you know, Anthropic, Mr. Trick here, not hiring him. That was shortsighted of them. I totally get it. I was surprised when I heard that he’d been hired by OpenAI and that they were gonna keep Open Claw going. Now they’re gonna support Open Claw, but in a foundation. So they are making a kind of a separation there, but they’re still supporting it. And I’m like, really? I would not want to be associated with that if I was a frontier lab, because, as we know, there’s nothing stopping the frontier labs putting something like that out. The reason they haven’t is because it’s risky and dangerous. So hiring him and supporting it I actually thought was kind of odd.
Justin: You’ve given me a brilliant business idea, by the way, Frank, as often happens, so, yeah. Yeah. So you’re saying they did a hit piece, a hit job on this guy. Right. I remember hearing about a business years ago, and what they did was the opposite. So let’s say you had some terrible skeleton in your cupboard and it got uncovered and you got some news press and stuff like that. There’s a company that exists in America, and what they will do is they will flood the internet with positive stories. Frank Prendergast donates to this charitable foundation. Frank Prendergast helps this child, crossroad, all that sort of stuff, so that when you go and do exactly as you described, the search on Google, all these positive stories come up and those negative stories are all buried on page 10 and nobody ever sees them. That’s a perfect use case for Open Talk. Frank. Frank has done, Frank’s done this terrible thing. Can you just go and write loads of good stories about him and say what a nice guy he is and bury that bad news for me, please, and it’ll do it really well. Unless it’s Carrie Marcus, in which case it should refuse to do it. Shouldn’t do it for Carrie Marcus. Joking, of course, I love.
Was GPT-4o more than just a model?
Frank: I know we usually try and end on something a little bit funnier, a little bit uplifting, but I’ve actually got very sad news to finish up on today. Justin, I have lost a very good friend. GPT-4o is gone, is no more, is not available in ChatGPT in the consumer interface. This is one of the things that has driven me to go back to Claude a little bit. So this is interesting just for a couple of reasons. I’m sad to lose it. I am genuinely sad to lose it because it was a really good model for writing in my opinion. Sam Altman has even come out recently and said they dropped the ball on the 5 series of models in terms of its writing abilities. Even though when they first launched it, they were all, oh yeah, it’s a brilliant writer, it’s a brilliant writer. It’s not; it writes florid prose that says nothing, and he’s admitting now they dropped the ball on it. So I am a little bit put out that he removed 4o before he has addressed the writing issues in the 5 series, but most other people apparently are upset to lose 4o for quite different reasons and it’s quite interesting.
Justin: Most is doing a lot of work in that sentence, Frank. That 50% of the…
Frank: When you say…
Justin: Most of…
Frank: Sorry, of the people, of the people who are sad.
Justin: Yes.
Frank: Yes. Most of the people who are sad to have lost 4o, it’s not because it was just a better writer. It’s because it was a companion, a friend, a support. A lot of people who, we’ve talked on the show before about people who believe that they have awoken their AI and that AI is this entity who has come to bring mankind to a higher level of consciousness or something. A lot of people who believe that they’ve awoken their AI tended to have done it with 4o, and so they have lost this spiritual guide that they had. Some people found it just to be a warmer companion, so maybe not necessarily a conscious entity, but just simply a warmer form of AI that provided a better form of comfort and therapy and support, and those people are genuinely feeling distressed. Some people have said that they literally are feeling the symptoms of grief, so, yeah.
Justin: How many of them are once, know how many there are, Frank.
Frank: Well, I know there’s 22,000 have signed the petition because I signed it myself, so I saw the number when I signed it.
Justin: Out of all the requests that go to OpenAI servers, how many of them went to 4o recently?
Frank: Oh, I don’t know. And…
Justin: 0.3%, Frank, 0.3%. When you say most people, we’re talking about most people within 0.3% of the entire population of people who use AI with OpenAI. So I think it’s maybe, you say 22,000, but I’d say there’s a lot of bots in there. I’m thinking it’s maybe, look, I’m not putting down, I’m sure they feel sad, but it’s another model will come. Every old model finds its old shoe.
Frank: These particular people feel like 4o just had something intangibly special about it. And I don’t necessarily feel that, I just felt it was a better writer, but if you do feel that, yeah, to have that yanked away, I think it’s an interesting problem for AI labs to have to deal with.
Justin: Wait till the AI is embodied in human form and people start to form emotional relationships with them. And then there’s a product recall and your 4o buddy is having to be taken home again. You’re like, no, no. I love her imperfections. They’re the way she is.
Frank: Black Mirror is like, is there any point in making another series of Black Mirror when we are literally just living in Black Mirror now?
Justin: Just listen to The AI Argument. That’s all you need. Justin, have an amazing time in your secret location that cannot be disclosed where you’re, are you, are you actually secretly a spy? Is this, is that what’s going on here? Well, if spies drink pada, I’d spend a lot of time by the pool. Yes. I’m secretly a spy.
Frank: Well, yeah, James Bond, right? Except it wasn’t a pina colada. It was what? What did he drink?
Justin: Oh, martini not stirred. Yes, shaken, not stirred.
Frank: You’re definitely a spy. Fascinating. Okay.
Justin: Yes. Look it, I will enjoy, we will catch up with you at some point next week, but I am travelling to another secret location on Friday. So it may be Thursday or Saturday. Who knows?
Frank: Excellent. I’ll chat to you then. Good luck.
Justin: Brilliant. See ya.
