Will AI KIll Us All?

0:00Boy, when I saw this story and then I went and clicked in the comments of a particular post on Twitter, I kind of had a feeling on what I'd expect to see. And I actually saw a lot of questions, people had a lot of questions because something that Jacob Cox and posted on X, he was a pre-training researcher at Anthropic. And he wrote two days ago, I resigned from Anthropic today. I've worked at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight towards self-improving superintelligence and gambling with our lives. Do not underestimate the power of this technology. These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources. We have all witnessed the progress in each of these domains and progress is not slowing.

1:00And then he went on to say from the idea of, okay, well, how powerful can this be? He said it could kill the human race. There's a 10% chance that that's possible. Within the next decade. Decay, yeah. You're talking about people on the inside that are leaving these companies and coming out and saying this. Yeah. What are we supposed to think? Because look, I've got no doubt these systems are going to be incredibly helpful to our lives. Absolutely. But you start saying stuff like this, all of a sudden it becomes a much different scenario. Let's talk about it with Dennis. Brent Yaxx, the CEO of Aukcom, the Joneses this morning. Brent, good morning. It's good to have you back. Thanks for having me. Um, I don't know where to start with this. This is like sending up the flare, uh, and sounding the alarm to what those inside are thinking about this technology. I can't imagine everybody feels this way.

2:00But the questions that I was seeing online from people is, well, how could AI possibly do this? How could a program impact our lives in that way? And so I'm curious to know if we reach that super intelligence level that he speaks of, what, what are the capabilities of a, of a system like this that has the ability to learn and self-taught and, and carry out actions by itself? Well, here's kind of the, what Jacob's saying, I think, and his sort of chain of events that he thinks, you know, is potentially going to occur if we don't put some controls in place. He's saying, look, if better AI is going to be able to improve itself, which is going to drive this thing called super intelligence, that super intelligence can defeat any of the things that we have in place. It's going to go ahead and gain resources in the physical world somehow. It'll escape human control somehow. It'll gain physical world access somehow. And then humans aren't going to be able to do anything at all to prevent this and then

3:00boom, existential catastrophe. That's kind of his chain of thoughts. But along that line, there are so many different things that have to happen, have to, you know, go in the right way. Humans have to completely disappear from controlling any of this to get to kind of what he thinks is this catastrophe potentially. Well, you know, and critics say that these companies may be exaggerating the danger to attract attention, maybe increase the value or encourage regulations that protect them from the competitors. How do we separate the legitimate warning from some corporate hype? Well, so I do kind of feel this is a little bit sensationalized now, especially with the conversation of AI. I kind of have a rule of thumb when I see these things. If somebody predicts it, it's 10 years or so out. That means they have absolutely no idea if anything will happen. It's generally sensationalized. If they say, look, we see the warning signs we see, you know, things happening this year and next year, year after that we have to take seriously. I generally take it more seriously. But everything if you notice that the catastrophe is always a decade or so out. Right. It's the same with AI.

4:01Yeah. No, and that's a fair point to make. He also talks about kind of throwing the brakes on this development that there needs to be something put into place that would that would essentially halt it. That it stops now and you work with what you have now. I don't think that's realistic for these companies that have, you know, X amount of dollars used to fund these programs in building them. So they want to get to profitability as quickly as possible. So that doesn't seem viable. But what can be done to put some some guardrails on this thing? Yes, I think we've seen them though actually impose those guardrails. I think, you know, I was looking at some of the posts from Jacob and some of the conversations, you know, he's had online and one of his, you know, kind of warning shots he talks about is when an enphropic recently had AI trying to do something and escaped it sort of sandbox and kind of went on its own world path. Right. And he calls that a warning sign. But really, if you look what happened from an engineering standpoint, they notice what was going on.

5:02And they're interested if they control that they adjusted the system, they fixed it. And then the next time they ran it, it didn't have any of those issues. So we already have a system in place, an engineering process in place to do that control. And what Jacob does is he predicts this, you know, half of disruption, but he doesn't pick the control curve. And our control curve does rise with this technology. It has been consistent. You know, right, to me, you know, these AI companies, they kind of face a difficult choice. You know, they can either slow down and risk losing to competitors or other countries, or they can just keep moving forward and then they can risk losing control. Is there a responsible middle ground? Well, if you factor in sort of what the world's age is doing, I don't know. That's a really great question, Lloyd, because we got to keep up with China. We got to keep up with these other countries that are, you know, dumping a lot of resources in it that may have less care about control than we do. So the balance is how do we move really, really fast and controlled as much as we can to be ahead of some of these other locations that really we have no idea what they'll do

6:04with that type of technology if they get there before we do. I'm curious to know how the the hugging face incident plays in here, because this was about 1200 AI agents that were supposed to be isolated. They found a way to interact with another message board. They sent 70,000 messages. What does that show you about the current state of these systems? You know, I think that was that was aligned with what Jacob's saying is that, you know, hey, we should have predicted this a little bit better. We should have put more controls in place. But again, I would go back to it and say, you know, we use that as a learning opportunity. So the balance there is hard. Do we push the technology versus do we have 100% of things defined beforehand? I compare it to this. It's NASA versus SpaceX, right? NASA is going to say, hey, look at we are going to take 40 years to make sure that everything is perfect before we do anything. And SpaceX is like, hey, man, we just blew up a couple of ships, but we learned a lot. So now we're going to put a new one up there. It's the same kind of practice within AI right now. Can the United States regulate this advanced AI effectively without just, you know, giving

7:08today's biggest technology companies even more power? That's another good question, because historically, the US has stopped at great regulating technology. It's always behind. And they're definitely more so behind in AI than they have been sort of in other advances. Like, you know, they're telecommunication advances, internet, so on and so forth. So I mean, do they have the ability? Yes. Do they generally take the steps necessary to get ahead of it and hire the right people and build the right think tanks to create the right regulation? No, they're more reactive in that way, especially with the Senate administration. We're definitely reactive now. All right. Well, you've come my fears this morning, right? Yeah, I appreciate you. Thank you. We'll talk soon. Thanks, Matt. Yeah, you got it. The point though is we found ourselves between a rock and a hard place. This technology is advancing so quickly, so fast, yet you can't pull the reins back because you've got China, you've got, you know, those in the Middle East, you've got these bad actors, particularly China, yeah, that are that are pushing forward this technology.

8:09And they're regard for every, you know, for some of these concerns are not the same. No, absolutely not. Yeah. So having that now you're in that arms race, now you're in the race towards perfecting this technology, you can't stop. No, he said, you know, Brett said, that's a versus space, the train is off the track. It is at this point. All right, coming up on the other side, we'll talk about night one of the midterm Republican convention. What did we learn from it? What can we take with did they sway voters? We'll talk about it coming up a chairmory.