Could AI kill all humans within the next decade?

0:00It's been nearly 30 years since humanity was forced to rise from the ashes and rebuild following judgment day. August 29, 1997. We all remember. Right, sorry, what? Did that happen, dark? I don't think it happened. I could have been wrong. Maybe this is all an illusion. So there was no such thing as a T2000. No. No. Not really. That's a relief. I know. Yeah. But could they soon be a reality? Should 79-year-old Arnold Schwarzenegger get back into game shape? Well, that's what some people are wondering, following a viral tweet from someone claiming to be a former anthropic AI researcher who says that there's a 10% chance that AI will kill us all in the next decade to help us walk through this cheery conversation. Oh, man, it's just nothing but barrel of laughs on the program today.

1:03Technology and lifestyle expert, Carmie Levy with us this afternoon. Hey, Carmie, thanks for taking the time. I've never heard it describe as a barrel of laughs, but yeah, I'll take that. I kind of had to shake my head when I saw this headline pop up in my feed because you sort of have to wonder, okay, if now insiders in the industry are warning us, you know, AI could kill us all in 10 years. Yeah, I think we should probably maybe perk up and take notice because this is something that I think we've probably needed to discuss for a while. And maybe we're finally approaching the point where the technology is getting just good enough that it is worrisome. Yeah, well, and here's the thing. I mean, it'd be nice to be able to write this off as, you know, maybe an angry ex-employee, but you had the Alliance Science Lead at Anthropic replying to this tweet saying, quote, we really earnestly believe AI could kill all humans. Yeah. Not the encouraging statement that you think it is. So what's the thought process here? What are the concerns that those who are in the weeds are seeing as the potential path

2:06towards the obliteration of the human race? So in order to understand the what, it helps to know the who and the who is this guy, his name is Jacob Cox. And he was, he first he worked with OpenAI, then he left OpenAI because he felt that Anthropic, which he was going to join, both in their both AI companies, he felt Anthropic was more cautious that there were more ethical and in fact, Anthropic was founded by ex-open AI employees to be more ethical and moral because they were concerned about potential abuses involving the technology. And so his job was he was responsible for pre-training the AI models. So basically, he was literally had a front row seat to the way this company operated. That is the process by which huge amounts of data are loaded into models. And that's what gives AI tools their intelligence. So he saw this happening and his concern was was that both of these companies were putting profitability over safety that the priorities were in absolutely the wrong place, that they're racing toward what we call super intelligence, which is where AI at some point in future

3:11will become smarter than humans. And it will be able to, you know, when humans say we're going to turn you off, the technology will be able to say, no, sorry, that's not going to work. That really is SkyNet in a nutshell. And his concern is is that we're approaching that point. And these companies are behaving in a way that suggests they don't really care about safety, all they care about is making sure that they win the race. And so he not only quit Anthropic, but he's quitting the industry so that he can now go public and share those concerns. And the fact that one of his former bosses at Anthropic, you know, essentially backed him up is incredibly problematic. They are suggesting that there's a greater chance, about a 10% chance now that within 10 years there could be a mass extinction event if we don't change course. So I think that's hyperbolic sure, and that's probably why we are paying attention, but it's really no different from what we were talking about a couple years ago when Dr. Jeffrey Hinton also quit Google to travel the world and warn us of exactly the same thing.

4:14So and of course Dr. Hinton is a Nobel laureate. He's probably the Canada smartest computer scientist. So I think we should listen to him as well. And quite frankly, maybe start discussing ways that we as a society and the governments that that you know that's the control and make the rules that they start prioritizing this and really get cracking on ensuring that there are better protections in place and that we don't allow these technology companies to make up the rules as we have been doing all along. And as we've seen pretty clearly, they're not really not very good at it and all they care about is profitability and they certainly don't care about us. Yeah, I'm half expecting my computer to start talking to me here any moment and it's probably listening to you right now. Yeah, yeah. Yeah. All high, how 9,000 of me like, I'm sorry Courtney, I'm afraid I can't let you continue with this conversation. Where's Dave? Right. Here's the thing too, right? There's another headline today that I think kind of fit into this and maybe offers another avenue towards a potential mass casualty event because you had anthropic basically halt

5:18potential plots by scientists who are using AI models to do research that could have developed biological weapons. That feels worrisome too. Sure it does because it shows that you know AI, we're shifting gears in AI development. We've reached a point where AI can work autonomously. AI agents can do complex tasks, complete them without any human intervention. So if you're a researcher, for example, and you want to maybe create a new drug, you just tell the AI agent, go ahead and do that. Here are the parameters. Here's everything I know about it. Here are all the details. Go off and do it and report back to me once you have design. Which sounds amazing because this could cure disease, it could do all sorts of great things. But at the same time, if you use it to create incredible next generation weapons, it can be used to kill people as well. I can say them and it can kill them. And so anthropic is on the one hand, they say, we're no longer going to allow research using our tools in this area.

6:20But very quietly in the background, they're also saying, yeah, but we're not going to fully hit the pause button because if we don't do it, in other words, if we're not engaged, then someone else is going to pick up the baton and they're going to do even worse than we do. So they're not fully ready to stop this research. And that is problematic. They're essentially deciding that they're going to be the judges of what shouldn't should not be done. And frankly, that's not a decision we should be leaving up to an AI company. It's a decision that's up to society through better legislation, better laws. Sorry, I was just googling cabins in the remote woods. Yeah, let me know when you find a really great off-grid place because I might very well join you. But we're kind of at the point now where these companies are essentially behaving as if the rules all apply to them. And here in Canada, we don't have legislation. The US doesn't really have anything. Europe has the AI Act, which they are now starting to use in the EU, that law, which is

7:20really world leading, to start going after these AI companies and they are starting to see them in court. And it would be really nice if Canada followed suit. And if we at least were in the same ballpark as the rest of the world in terms of leveling the playing field and making sure the tech companies don't become too powerful. But we've always lagged when it comes to legislation. And unfortunately in the AI era, that's not going to do us any favors. Yeah, because here's the thing, right? Like we talk about it seems to me once every year, maybe twice every year, we talk about, oh no, here is this asteroid hurtling towards us. Then it's going to miss us by 50,000 miles or something like that, right? And we can really do about this or that rather. But when it comes to this, you talk about the alarm being sounded by some of the individuals that you mentioned. I was reading a story on Axios today, where they were talking to experts in this field and asking them, okay, in the next decade, what do you think about AI possessing dangerous abilities, cyber attacks, power centralization, competitive dynamics. And the polling on that was low, but not low enough to make me feel good about things.

8:24A lot of these were coming in with the 20% of experts saying that, yeah, there's a good chance that all of this could happen. Yeah. So one in five today in 2026, but the technology is a moving target. It is going to get better. And that curve, that steepness with which it is improving, is developing faster than any previous technology in human history. So today, it's one in five experts say, yeah, we're going to have a problem in 10 years. A year from now, based on the state of the art of the technology then, it'll be more than one in five. And that number will continue to go up as the technology becomes more powerful, more autonomous, has a greater scale and scope and all of these things. And of course, as infrastructure becomes more capable, we have more data centers, we have better networks that can allow even higher levels of compute. So basically, it's almost like SkyNet getting a brain transplant, it will be far more powerful than it is today. If we don't figure out how to control it now, the further down the line we get towards super intelligence, the harder it's going to be to rein it back in, which is probably

9:27why we're seeing so much urgency in these calls today, as the clock is ticking. And the longer we wait, the harder it's going to be to reverse course. At least that new Apple iPhone seems cool. I found that interesting that both of these stories broke on the same day. Because on the one hand, it's like, oh, you know, doom and dystopia. And yeah, should we call Arnold Schwarzenegger out of retirement? I'm always a fan of that. And then on the other hand, it's like, oh, look, new shiny new phone from Apple. And the two of them are somewhat connective because of course, these are AI powered devices. And it's these devices that we're going to use in the AI era. But at the same time, it can reinforce us that we don't realize or we spend $3,000 on this folding new iPhone duo. And we don't realize that we are part of that AI generation. And we are going to use AI services on it that are leading us down this, you know, frankly, concerning path. And everyone loves the good Apple product. So why not just open up your bank account and buy it? Indeed. And after this conversation, I like a good ice cream product. I probably have four liters of it in my future tonight. I've got some black, I've got some black cherry in my freezer.

10:29And I think I'm heading up for some right now. Well, I guess I want to thank you for this conversation because it was enlightening, even if it was a little bit soul crushing. And hopefully, you know, again, you can't be doing this with our eyes closed. So it's good to have a frank and honest conversation. And hopefully we can get at least another 20 years out of the human race before it's all said and done. I will do my best because I'm all about positivity. Okay, Scott, I mean, appreciate that.