AI safety fears grow as experts warn technology could spiral out of control

We've been talking about this here. You know, the growth of artificial intelligence has been a major cause for concern recently, especially now that one researcher from Anthropic, he says that he quit his job over fears of what the technology could eventually become. Governor Pritzker in Illinois responding to that resignation, saying it is time to sound the alarm when it comes to reigning in AI. So we wanted to dig a little bit deeper about these concerns with an expert, Professor Christian Hammond from Northwestern University joining us to talk about this this morning. Professor, thanks for joining us. Oh, thanks for having me. So this story about this researcher at Anthropic who said that, you know, open AI and Anthropic, none of these companies are, you know, acting ethically here. He quit his job. He's worried that AI can become out of control. Can you give us some context here? When we say out of control, what does that mean? What does that look like? Well, all these all these systems are really based upon probability. And so at any given point, in order to seem more human, they when they're choosing words, for example, they're sort of guessing and about the best word. And so you're you're it's very hard to predict what they exactly what they will say, because, in fact, they're guessing at each at each moment. That's fine if it's words. But when you start putting control in the hands of these things, then they're you have to remember, they're guessing. And so we're not quite sure exactly what they're going to do. So I guess that's my next question is, you know, they're guessing. We're not sure. I mean, even even people at Anthropic, you know, some of these researchers have said, you know, we're building these systems. We know how to build them, but we don't really know what's going on behind the scenes. That is a scary thing to hear, especially if you're just a normal person in the universe that, you know, uses chat GPT. What do you what do you make of that? Why? Why are we making things that we and we don't know how they work? Well, it's less that we don't know how they work and more of their complex enough. So they're hard to predict. And there's a big difference. We know how they work. They do a lot of they do a lot of math. But but predicting exactly what they're going to do next is is more difficult because the options are are are enormous. So go ahead. Oh, no, no. Keep going. Keep going. I mean, the the thing of it is talking about a thing you're having a conversation with the fact that it might be wrong is is not devastating. Nobody dies. But you start putting you start putting control of elements of your life, your email, your media, your communication with the outside world in general. You start putting the control of that in their hands and you don't know what's going to happen. That's an odd thing to do. We you don't want to put control in the hands of a thing that might be random. But in that but in that sense, you know, I mean, we we do have so much tied to A.I. now. And now you've got this, you know, ex-anthropic researcher who's coming out and saying that A.I. could cause human extinction by 2030. That is a scary headline. So can you put some context to this? Do we need to be worrying? It's a scary headline. We need to be concerned, but we need but worrying is probably overstating it. A.I. has a possibility of killing us by 2030. Where does that come from? He's just making it up. He's like the systems themselves. He's just making things up to fit a scenario he thinks is likely. And this was a while back. There was a Google employee who quit Google because he felt like he felt like the system had become sentient. You are these are human beings who are making guesses about what is going on in the world. And simply because he's he's inside of anthropic doesn't mean he has a he has a monopoly on on on on futurism, on guessing the future. And it's an odd thing that we we get excited about these these big events. But the reality is that is that it's the it's the it's the allowing system to have control over the elements of our lives that are important to us. That's dangerous. It's it's not worrying about extinction of the race. It's worrying about, you know, email being sent in your name that doesn't really fit your pattern or it's buying things that you don't want because the system is confused. These are the things that that will have a tremendous impact on us personally, much more so than worrying about the extension extinction of humanity as a species. OK, so I guess that can that can ease our fears a little bit here, because that was that was kind of a shocking story that came out yesterday. So it's good to have this context. When we talk a little bit about AI, you know, I I've been one of these people that's like, why don't we just unplug it? You know, if if we don't want these systems anymore, if we don't want them to send the emails in our name, if we don't want them to buy the thing, why don't what what what do you say to people who who just think, well, if we don't want these systems or machines, why don't we just unplug them? What do you say to that? Then what you've got to do is take a look at the history of viruses in computer viruses and in modern times, your viruses that spread like crazy. And even though people know they're there, they're hard to stop and they're jumping, they jump from machine to machine. If you build a super intelligent system, one of the first things it's going to figure out is how to escape. I mean, that's that's what it's going to do. And you can't shut it down in one place because we had designed the Internet to be stable with regard to attacks like shutting things down. And it will it's it will be ubiquitous. And the question is not so much the AI. It's who we're going to put in control of what humans were going to put in control of these systems and what are their goals, because these systems are simply going to follow the goals of that they're given. They're not making them up themselves. And it's the people that we have to be concerned with. So that that is actually that kind of follows my next question. You've got this kind of like it's so it feels like a space race, but with open AI and anthropic and then you've got countries outside of the U.S. trying to do this as well. And the argument is we can't shut these systems down because if we if we aren't making the technology, then some other nefarious actor is going to make the technology. So so where how do you trust? I mean, I know Anthropic has sort of been positioned as this ethical AI company, but then, you know, Jacob here, on the other hand, who just quit at Anthropic, says neither of these companies are acting in an ethical way and they're putting human lives on the line. So how do you how do you trust the people like you mentioned? We have to be concerned about the humans in charge. How do we trust that what they're doing is the right thing? I think right now we don't. I mean, I mean, you're right. Open AI, Anthropic, they are they're racing against each other. They're trying to become the best. But in doing so, what they're doing is they're giving up on testing and safety and harm. And no one is tracking. No one is really tracking the harms that are being created by AI systems. We all are living on the promise of what they'll give us. But where there's not enough of understanding of where the harms are and they don't have to huge harms. They can be little harms. And understanding that that killing us and killing you know, killing us all is is not as worrisome as buying a slightly less slightly more expensive item than we needed to spend because the system has different goals than we have. And it's these thousand little things that will that will impact us, not not not extinction. OK, so you're talking about more like a thousand paper cuts instead of, you know, this mass extinction here. When you talk about these harms that no one is tracking the harms, what safeguards need to be in place? Who needs to be tracking those little harms? It's not clear. But the federal government doesn't seem to be all that interested in doing it. The companies are simply are not interested in doing it. They want to make sure that the obvious the obvious harms aren't aren't there. But aside from that, they don't go into they don't go into the details, which is to say sort of erosion of erosion of our own thinking by using these systems. So but this is a place where if if we as as users demand more testing, demand verification that these systems are are not harmful, if we simply demanded its users, then then the companies will have no choice but to respond. But we don't demand it. And that's the thing that's the problematic is we're living with us, you know, with systems where at the bottom of each of the Web pages, it says, just a reminder, the system is probabilistic and can make mistakes. And you've got to check all the answers. We're living in a world where the technology itself and the way it's used is kind of faulty, but we live with it. It almost seems like maybe these headlines that we see that are, you know, maybe they're waking us up to the fact that we do need to demand more as users. So this is a really interesting conversation. Professor Hammond, thanks for joining us. I wish we had more time to chat, but we have to leave it there right now. But thank you for being with us about this. Oh, and thanks for having me. It was fun. Yeah, absolutely.