Tech analyst on AI guardrails and a whistleblower's dire warning

There are proposals from both parties to put more oversight in place, but right now there's no single set of rules every AI company has to follow. Joining us now, AI expert Joe Tavares. Always good to see you, Joe. Thanks for joining us. Yeah, thanks for having me. So you heard Speaker Johnson at the end there say that it's up to the companies to know kind of what's best and what's not. Well, it's the companies that came out and had some pretty stunning words about their own progress. Talk about that. Yeah. I mean, you just said across the board, all of the large AI labs are kind of coming out and saying they want to like slow down a little bit. And to do that, they need to agree, because if one of them slows down and then everybody else speeds up, you know, there's going to be a disparity there. So I think that saying the companies can regulate themselves, they're not going to. They're basically begging for regulation from a government body at this point. Now, I understand, depending on what part of the political spectrum people are on, they can say that regulation is either good or bad. But we all know that, for instance, the auto industry is regulated to a certain extent. All industry is regulated. This industry is regulated by the FCC. And so why is it that it seems like there seems to be a blank check in terms of power for AI companies? With this one in particular, I think the general feeling is that we're in an arms race, that we're locked in with China and Russia and other adversaries. And if we don't get there first, we're going to lose. But I think maybe if anybody gets there first, we all lose if we don't approach it properly. What do you mean by that? Because I've heard you say that before, and I think it's a great point. So the metaphor I like to use is that we're driving through fog right now. We don't really know where it gets bad or if there's anything up on the road ahead. So the best approach is to slow down and drive carefully. Well, with this, we don't necessarily know what might happen. We can't assume anything about what a superintelligence would do because we can only frame things in terms of regular intelligence. So we wouldn't know when or how it got out of the box or what its goals are at that point. What is it that when we hear things like it may destroy humanity in a decade, what exactly would AI have to do for that to happen? I'm not putting it out there. I just don't quite put my head around what that would look like. Yeah, sure. So science fiction, again, if we look at Terminator, it took control of the nuclear weapons and immediately launched them. I think it's going to be something more benign than that, where we ask it to optimize paperclip output and then it consumes all the resources to make more paperclips rather than to do something useful. So it puts a drain on society in many big, medium and small ways. Yeah. And again, it's hard to anticipate that because you don't know exactly what that goal might be. So right now, the big fear is that we're racing towards recursive self-improvement. So the models are making themselves smarter. Well, in order to make themselves smarter, they need to collect more data sets. They need to have additional hardware to do the computations. So what does that mean in the grand scheme of things? They would need to build robots to go gather more data. They would need to consume more silicon and fuel and water to produce more calculations. So again, it's really difficult to anticipate what that might look like. What does it look like in terms of legislation when you have the Trump administration saying it's a hoax, that this thing could ramp up and take over the world? But on the other hand, you have people saying we need some regulation in the middle there somewhere. I think there's like a huge disparity when the people that are working on it are saying, hey, please regulate us. And the regulators are saying you don't really need it. That's again, it's hard to bridge that gap for me. Yeah. It's a gap that we'll keep our eyes on as people continue to talk about this. It's an important one. And so AI expert, Joe Tavares, good to see you. Thanks for joining us here tonight. Yeah. Thanks for having me. Rich, over to you.