AI safety requires action, not promises

SPEAKER_1 (0:00)

Hi, I'm Molly Graham, host of Worklife, a podcast from TED. The most important lessons about work usually aren't the ones anyone teaches you. They come from experience, from uncertainty, from figuring things out as you go.

On this show, my expert guest and I explore how careers really unfold. Through change, through doubt, through the decisions that shape who we become over time. Because those moments aren't the exception, they are the work. Listen to Work Life wherever you get your podcasts.

Megan McCarty-Corino (0:34)

An AI futurist is getting deja vu this week. From American Public Media, this is Marketplace Tech. I'm Megan McCarty-Corino.

AI leaders suddenly seem to agree. It's time to slow down the AI race. Anthropics' Dario Amodei published an essay last weekend calling to pace the frontier, slow the development of advanced AI to allow safety measures to catch up. He said Anthropic will give access to independent safety evaluators. OpenAI's Sam Altman and XAI's Elon Musk seem to concur.

It's all giving deja vu to Amy Webb, founder of the Future Today Strategy Group, who has been pushing government leaders around the globe to adopt a similar framework for AI for almost a decade. She laid out her vision in her 2019 book, The Big Nine, How the Tech Titans and Their Thinking Machines Could Warp Humanity.

Amy Webb (1:35)

The world does not need another manifesto. What we need are for folks to get together and commit to action with accountability. So it's wonderful to say that we will make some of this explainable. We will have models being tested. We will have people observing those models being tested.

But what the largest companies are calling for right now will eventually harm every other startup in the space that doesn't have the same ability or certainly capital flow to do what the big boys are talking about.

Megan McCarty-Corino (2:04)

Yeah. Explain that for me. Break that down.

Yeah.

Amy Webb (2:07)

Look, at this stage, everybody knows what OpenAI and Anthropic and DeepMind are. These companies have a lot of cash at their disposal, so they can potentially put people in places to do reviews and sandbox testing and all of these other things that we seem to want. But if you're a startup with 50 people and you are similarly inventing new tools, you just don't have the same kind of resources. So calling for safety measures without actually doing anything to create the circumstances where everybody would be able to participate is not actually addressing the problem. What it is potentially doing is making it much harder for the smaller companies to compete. It's a way of creating an artificial moat.

Also, these companies don't need regulation to prevent them from creating dangerous tools. There is an easy way for them to address that problem, and it is to do things differently inside of their companies. And by the way, technology doesn't necessarily respect borders. So, we have to create something that others around the world can also participate in.

And I think the best way forward is to use an incentive structure. So, let's come up with some carrots rather than relying on sticks.

Megan McCarty-Corino (3:27)

I want to go back to what you said about the idea that these companies can be prioritizing safety without some sort of external control or accountability.

That's something that we have seen over the last several days, including from former AIs, our David Sacks. We argued, fine, go take the lead. You don't need regulation to do this. You don't need to be worried about governments to do this. Go ahead, you companies that want to do this. It makes sense. I mean, what do you make of this school of thought?

Amy Webb (4:02)

Look, I don't think Sam Altman or Dario or any, most other tech leaders are out there to intentionally cause harm for anybody. I don't think that's what they're doing. I do think they've gotten themselves into a situation where there's so much money at stake that they are pressured to push on and to move forward.

Regulation is not the way out of this, but neither is blasting from the rooftops new manifestos and calls for action when there's very little that will be done on the other side of it. We've been here before. Do you remember Facebook had its Supreme Court of Justice? The Oversight Board. That's what it was. And there were like nine people that were meant to evaluate cases that would bubble up from a billion or more users to provide a fail safe against development of tools or content that was out there that was not in the best interests of everybody. You know, they certainly couldn't keep up. And when it became no longer needed, Facebook shut that service down.