Things Ranjan Says on Podcasts

dear. Which involved me getting yelled at

last weekend. Because, you know, though Muse did put this very helpful list of doctors together, it's very capable standard model. I'm saying this facetiously. Put the wrong phone numbers for the wrong doctor. Like put working phone numbers of some doctors on the wrong, underneath the wrong doctors' names. So my wife booked an appointment from Muse, you know, after making this phone call, the first person that it suggested, thinking that it was going to be the first doctor on the list because the phone number was underneath the list. It turns out she had made an appointment for a completely different doctor that she had called because Muse put it in the wrong place. And so the night before she's supposed to go, she's trying to figure out where she's going. And we thought it was going to be a brief commute into Manhattan, but it turns out that the doctor that she had booked was a multi-train ordeal in deep queens. And she's looking at. me and she's like you're freaking muse man

um so ranjan i did in this conversation suggest why didn't you double check let's just say that was not the right that was

not the right answer

but i'm trying to say what i'm trying to say is um you know i do think a frontier intelligence will probably be more accurate dealing with such type of uh data problems and so that's why you know i i'd almost trust not almost i would trust open ai's dots much more than i would try trust uh metabuse uh for a similar uh activity in the future

You make a compelling argument and almost Bernie Sandersify me. And I'm going to start yelling, pause AI development right now. I'm not arguing for

that. No,

You know, I think it's reasonable. I mean, I think this is a reasonable discussion to start to have right now. And I think it's important to say that this is not the beginning of behavior like this, that we've seen behavior like this for a long time. You know, as early as beginning of last year, even late 2024, there were examples of AI that was like, they're so, they get put on a task, right? And this is the thing, and I think it's really important to talk about, and I've talked about it in the past on the show, but I'm going to talk about it again here. That there is, you know, self-supervised learning, which is like basically predict the next word, recognize patterns, repeat them, and reinforcement learning, which is like you're given a goal and you just have to find the way to reach the goal. So AIs will do these simulations thousands of times in order to reach the goal that they're given. And they'll learn from their mistakes. And what we have now is that the reinforcement learning type of AI technology has been put on top of the self-supervised learning to get. These AI models are working better, which has added a level of ruthlessness to them. Because one of the things we know about RL is that there's a level of ruthlessness, that the AIs will stop at nothing to accomplish their goal sometimes. There are examples of the AIs playing chess and being told to do it from a reinforcement learning way. And instead of playing the game the right way, hacking into the chess game, rewriting the rules so they can do whatever move they want and winning. And so I'm with you that like as this stuff has gotten more prevalent, I think that there is serious, like there's a serious demand for more concern about what's going on. Now, my response isn't pause AI development right now. I just don't, I don't know if that's that.

That was the most hedged statement I've ever heard. You just said there is a serious demand for concern. Come on.

You're right. Are you a pause guy? We would be a pause guy. We would ridicule somebody who said that. So I think that ridicule is.

That's a good kind of like corollary. Is there a soft landing? Is there no landing? I think that's like a good way to think about it. And what do those different scenarios look like is I think, I mean, I think we're going to be talking a lot about that, especially as it's August 21st and Anthropic is supposedly coming out to the market next month. I think we're going to start seeing a lot more concrete numbers assigned to this whole story.

Yeah. For what it's worth, I just want to state my perspective on this because I have been thinking about it over the past couple of weeks. And then we're going to get to the Anthropic and IPO and revenue numbers and open AI revenue numbers. I'm starting to think there's really only three true outcomes for AI. Like one is AI works and it's good, right? So it leads to economic prosperity and like a rollback of disease. The other is AI works and it's bad. So like we start to see more of like the rogue hacking activity that we've seen and it gets in the hands of bad actors and it leads to like bioweapons and things like that. Or the third possible outcome is that AI doesn't work. And I've really struggled. Like the idea of this soft landing, I struggle to see it. Like if AI doesn't work, there's been so much money tied up into this system that there's going to be some form of crisis. So I really see it as, and I'm going to, I think I'll write about this on big technology when I get the newsletter back up and running, probably like post-Labor Day. But I really see it as three true outcomes here. What do you think about that?

Okay, so first of all, we'll add a fourth potential outcome was that AI works, but it works too slowly and it causes financial collapse. There we go. There we go.

Now I'm good.

But I'm very, I very much disagree with you about this disease thing. I mean, AlphaFold, I know it's been kind of talked about forever, but like some of these personalized healthcare examples that you're getting, you know, maybe it's not cure, it could be managed. They're very, very difficult to argue with. And I'm personally like, I think that it could be very big in a way that, and the amount of focus that's going into it within the labs, to me, I think it's, it's not worth, you know, sort of waving away.