Can We Avoid AI Catastrophe? (with Tech Ethicist Tristan Harris)

Tristan Harris (0:06)

We are releasing the most powerful, inscrutable, uncontrollable technology that we've ever invented. We're releasing this technology faster than we deployed any other technology in human history.

It's already demonstrating exact behaviors we thought only existed in sci-fi movies, but they're happening in real life, and we're deploying it under the maximum incentive to cut corners on safety.

Sam Fragoso (0:33)

This is Talk Easy, I'm Sam Fragoso.

Welcome to the show.

For the last three years, AI researcher Jacob Coxon has been on the inside of both OpenAI and Anthropic, where he's focused on training and pre-training new frontier AI systems, including GPT-40. But as you may have seen, he resigned from Anthropic this past week over concerns that neither company is acting responsibly. Quote, they are racing straight to self-improving superintelligence and gambling with our lives. As you'll hear from our guest today, Tristan Harris, Coxon is far from the first person in the tech world to sound the alarm, but his public disavowal has spread like wildfire, in part because his resignation has been reaffirmed by other AI safety researchers, including one currently doing alignment research at Anthropic, who said, quote, we really do earnestly believe AI can kill all humans. I personally think it is greater than 10 percent within the next decade. In the wake of these declarations, better late than never, Washington has decided to take an interest. Next week on the 16th, Bernie Sanders will host a private briefing for senators with top artificial intelligent researchers. An hour after that briefing ends, and just a few doors down, our guest today, Tristan Harris, will be hosting the first screening of the AI Doc on Capitol Hill. Harris was previously a design ethicist and product philosopher at Google before going on to co-create the Center for Humane Technology, where he's helped inspire a wave of activism and legislation around the globe to protect kids and teenagers from the addictive lures of social media. The question is whether Harris and others in this movement will be able to reach members of Congress in the same way they did with the tech issue for AI. Can they persuade a historically geriatric Congress to create guardrails for a technology that's probably evolved since the beginning of this sentence? Democrat James Clyburn of South Carolina, great James Clyburn, recently admitted that up until a few weeks ago, he didn't even know what Chat GPT was.

SPEAKER_3 (2:59)

Believe it or not, I didn't know what that was until about a week ago.

Which way? Chat GPT or whatever it is.

Sam Fragoso (3:05)

That's just the one politician who admitted it. When I reached out to Harris for comment this morning, he described this moment as an inflection point, citing a recent report done by two independent research organizations, that the technology right now is, quote, more than 50 percent of the way to full-blown AI takeover. I realize this all sounds in reads as terribly apocalyptic, dystopian, Black Mirror-like, whatever term, analogy, phrase you want to use, it's probably apt and appropriate, and I feel it too.

For what it's worth, I know Tristan does as well. Despite all of that, despite all those fears and frustrations and concerns, many of which you'll hear in today's episode, when I reached out to him for comment, he did send me this text and I'll just share it with you. He said, quote, that we are in a narrow window, a very narrow window, to pause and pivot to a pro-human future. That this isn't a foregone conclusion, despite what the Internet tends to tell us every day. This is Tristan Harris.

Welcome to the show.

Tristan Harris (4:23)

Good to be with you.

Sam Fragoso (4:24)

You know, you have spent the last few years advocating for AI safety and restrictions. Every day it seems like you're on this fight.

Are you all right?

Tristan Harris (4:39)

Yeah. I mean, it is interesting to point your attention at something every day that is advancing incredibly quickly. There's actually something, I think, for people who in general work on AI as a topic, that it has a kind of psychological impact for just about everybody who works on it.

Sam Fragoso (4:56)

Is it kind of like how owners start to look like their dogs after a while?

Tristan Harris (5:00)

Yeah, something like that. Maybe AI people start to look with the same bags under their eyes or something like that as they all stare into the void. I mean, AI is a technology that is just so unlike all their technologies, it just moves faster than every other thing. It's quite profound and confronting what it represents as things we have to respond to. I think that that just impacts people.