Elon Musk on AI: humans will no longer be in control in ten years — The Economist

[00:00:00] Speaker 1: I think AI may exceed the sum of human intelligence in around five years. [00:00:07] Speaker 2: In five years? [00:00:07] Speaker 1: Roughly five years is my guess. So there really won't be anything that AI can't do better than humans, apart from being human, perhaps. [00:00:20] Speaker 2: In a more prosaic level, what will life be like? [00:00:25] Speaker 1: The most likely outcome is an age of amazing abundance, where anyone can have anything they can think of. This may sound preposterous, but, well, here we are in 2026. Let's see where we stand in 2036. Now, of course, there are things that could derail this, like a massive global thermonuclear war or something like that. But assuming that there's not World War III, I think we're headed for an age of amazing abundance. So this is, I guess, a message of optimism and excitement about the future. [00:01:07] Speaker 2: You're sure that humans will no longer be in control in 10 years' time? [00:01:11] Speaker 1: I think it is unlikely, in the same way that if you had a vastly more, like, you know, the, like, if the difference in intelligence between AI and humans is vastly greater than the difference in intelligence between AI and chimpanzees, it's hard to imagine that the chimpanzees would be in charge. [00:01:38] Speaker 2: I'm really struck by this. [00:01:40] Speaker 1: And we are essentially an evolved form of chimpanzee. So it wasn't that long ago that we were swinging through the trees, eating bananas in the jungle. And we still eat bananas. [00:01:53] Speaker 2: You said there was a 10 to 20% chance of killer robots wiping out humanity. So you sounded worried then. Now you seem very relaxed about it. Have you changed your view? [00:02:03] Speaker 1: I still think there's risk associated with AI and robots. It's not zero. I guess I've come to my sort of philosophical conclusion is to look on the bright side. Like, I can't see any way to really stop this incredible momentum of AI and robots. And at times I sort of think, well, perhaps, you know, even if there was a stop button, we probably shouldn't press it because the most likely outcome is incredible abundance for all. [00:02:45] Speaker 2: That's what I thought, actually. It struck me that now you seem to have decided it's inevitable. Let's hope for the best and go for the ride. [00:02:54] Speaker 1: Yeah, I mean, yes, pretty much let's enjoy the ride is I think my philosophy at this point. Because, I mean, honestly, there seems to be just inexorable progress here in AI and in robots that even if I wanted to stop it, I couldn't. [00:03:19] Speaker 2: Do you think there's still a 10 to 20% chance that killer robots mean the end of humanity? That's a high P-doom, as they say. [00:03:30] Speaker 1: Well, in the long run, we're all dead, of course. I mean, so, according to the standard model of physics, the end of the universe is a long, boring, sort of thin gray soup of gradual cooling. I mean, it's sort of the heat death of the universe is called. So, if the heat death of the universe is, in fact, the outcome for our reality, then it really is all about the journey. Because the destination is terrible. [00:04:07] Speaker 2: Yeah, but 10 years is not very far. Yeah, I get that, but that could be a long way ahead. Yeah, a very long way. And you're basically saying, if it's 10 to 20% in 10 years, would you go into one of your rockets if you thought there was a 10 to 20% chance of it blowing up and killing you? [00:04:20] Speaker 1: Yes, but let's say you can't do anything about it. I guess the only thing we can try to do is minimize the probability of the VAT outcome. But trying to stop the intelligence in robots, I mean, for the longest time, I declined to participate in AI or was, you know, created OpenAI as a, essentially a counterweight to Google because at the time they were, they had more or less a monopoly on AI. In creating OpenAI and then having Anthropics spin out of that and now currently Anthropics the leader in AI. I mean, these actions have actually resulted in knock-on effects that accelerated AI, which wasn't really my intention. So, it just seems like all roads lead to acceleration of AI. So, then I'm like, okay, well, you can just sort of be sad about it or join the club, I suppose. [00:05:29] Speaker 2: Well, you've certainly joined the club. [00:05:30] Speaker 1: Yeah. [00:05:31] Speaker 2: You've definitely joined the club. [00:05:33] Speaker 1: If you can't beat them, join them, you know. [00:05:34] Speaker 2: A number of the other leaders of model makers have been pushing for different ways to try and minimize the risks. You know, just last week, Demis Asabes came out with a proposal for a private public regulator. Do you think you guys should be working more actively to get something in place to control the risks? Or do you think that's impossible, too? [00:05:57] Speaker 1: Yeah. My recommendation to Demis, because I talked to him for a few hours before he published his piece, was that the most immediate thing that we could do is to have the leading AI companies at least just meet. We'll have some sort of call once every few weeks and just discuss any safety and security issues. [00:06:23] Speaker 2: But why haven't you done that yet? I mean, you don't all like each other very much. Yeah. But you're not all, I mean, you're also all competing, right? You all want to be first. [00:06:32] Speaker 1: Yes. But that's where I think the competitors can keep each other honest. So meaning like if, I think it's quite difficult for someone in the government who doesn't have a deep technical understanding and isn't driving the frontier of AI to know whether something should be released or not. However, the competitors, I think, can, you know, if given sort of a week or two to review a new model, can highlight issues and have an incentive to keep, all the competitors have an incentive to keep the others honest. [00:07:08] Speaker 2: So you'd have a system where the model makers would check each other's work. If there was something that worried them, you'd tell the government. Yeah. The government would then think. [00:07:16] Speaker 1: If there's something that was worrisome and that company was not doing anything to address that risk, then that would be the moment for government to step in and take action, which is literally what just occurred with us. [00:07:27] Speaker 2: How soon does all this have to happen? [00:07:31] Speaker 1: Well, I think it would be good to do that, I think, immediately, really. [00:07:36] Speaker 2: Yeah. That's what people keep telling me. We have to get this some kind of system in the next six months because of the pace at which the models are improving. [00:07:45] Speaker 1: Yeah. Six months is a long time. [00:07:49] Speaker 2: So you think shorter than that? [00:07:51] Speaker 1: Well, I mean, there's so many AI breakthrough announcements, sometimes multiple per day. I mean, how many were there last week? [00:08:01] Speaker 2: A lot. [00:08:01] Speaker 1: I lost track, you know. [00:08:02] Speaker 2: Yeah, a lot. So you basically, you know, whether it's Demis' proposal or others, you need to get together and get this set up kind of now, it sounds like you're telling me. [00:08:14] Speaker 1: Yeah, I would recommend that we at least have some sort of informal, you know, weekly or biweekly call. And like I said, just that there's maybe a week or two weeks of early access by competitors. And this is why I think the incentives work is because, you know, competing companies, I think, are likely to understand if something is a security risk. And they're not going to be shy about highlighting that their competitors' models should be delayed in release. [00:08:52] Speaker 2: You guys don't much like each other and you don't much trust each other. You're insulting each other on social media constantly. So do you think you can do that with people that you dislike, where there seems to be such kind of personal vitriol between some of these leaders? [00:09:09] Speaker 1: Yes, well, I'm not a fan of Sam Alderman because if you started a non-profit that was meant to be an open source AI company and, you know, owned by the world and it somehow got turned into an $800 billion for-profit company with closed source, I think you'd be like, well, wait a second. That's the exact opposite of what I donated the money for. That's my issue. I think it's a legitimate one. And then Dario, I think, also is not a fan of Sam Alderman because he felt, in fact, the reason the Anthropic team left OpenAI is because they didn't trust Sam Alderman. Otherwise, Anthropic wouldn't exist. They would still be at OpenAI. I think Dario is a very principled person. He cares about things a lot. He cares about the future of the world. And I think everyone I've met at Anthropic so far has been, like, no one set off my evil detector. You know, they always seem like they're well-intentioned. And the road to hell is, I think, mostly paved with bad intentions. There are a few well-intentioned paving stones in there. So we don't want to be complacent. But at the end of the day, if we have to talk, we'll talk. I mean, set aside our personal differences for the good of the world type of thing.