SCOTT DETROW, HOST:
For years, the conventional wisdom has been that if you want to work in tech, you should learn to code, and if you want a job anywhere, you should avoid liberal arts majors like philosophy. But that is all changing as AI has endangered coding and a growing number of AI labs have been looking for new hires from a surprising pool of candidates. That is right - philosophers. That is, people who have thought deeply about ethics, reasoning and human decision-making. Benjamin Sutherland recently wrote about this phenomenon for The Economist and joins us now to talk more about it. Welcome.
BENJAMIN SUTHERLAND: Thank you, Scott. Glad to be here.
DETROW: Let's just start with the why. Why are AI companies looking for philosophers right now?
SUTHERLAND: Yeah, well, there are a couple things. I mean, some of it is just to develop models that have better reasoning processes. Philosophy students are taught inferential reasoning, logic, and that can be helpful for just developing longer chains of thought, as the AI people call it. But the bigger area is really to try and figure out how to make AI models that are safer. And that's obviously become front and center as these models become so much more powerful.
DETROW: How many philosophers are we talking about, and how high up on the food chain are they at some of the big companies? Is this just kind of like, we'll have a few and kind of let them think about this? Or are they really central to the decision-making?
SUTHERLAND: Yeah. No, I wasn't actually able to get good numbers on how many are being hired, but I think they actually are in pretty senior positions. They're the - philosophers are the ones who are writing out these constitutions. They're writing what Google calls its AI Principles. So they are really shaping how these models are being developed.
DETROW: In your article, you talk about a couple different philosophical schools of thought that some of these AI companies are really leaning into. They want to make sure they're part of the models, part of the thinking. I'm wondering for, like, an average end user, is there any way that we can detect the presence of this or, like, which philosophical school the company has leaned into?
SUTHERLAND: Yeah, absolutely. I mean, Anthropic's Claude models are just obsessed with honesty. You keep hearing that term in its - in the output, and that's a sign of kind of the deontological structure. There's certain things that a model cannot do. It can't say anything that would be untrue. And - now, you can run into problems with that because there's times that an untruth could, you know, perhaps save a life or it could be for the greater good. Then the other kind of opposing school is called consequentialism. And that's kind of a form of utilitarianism. And the idea is you look at what is going to create the greater good. If you take the notion of self-driving cars, if there's going to be an accident, the car has to be able to quickly decide what's the least tragic way to crash. And if it chooses to hit one person rather than, you know, a family of five, that's a consequentialist decision.
DETROW: Is there a sense that one of these schools of philosophical thought will win out over the other in terms of how broadly they're adopted within this field?
SUTHERLAND: Well, one thing I can say is that we're seeing increasing pressure on lawmakers in a lot of different countries to regulate AI. And I think the way that's going to happen is going to be primarily with the deontological school of thought because laws are, in fact, deontological. They're saying this behavior is illegal. You can't do it, no matter, you know, what you would like the outcome to be, no matter what your intent is. And so in order to abide by these laws, I think we'll see more and more deontological systems kind of working their way into these models.
DETROW: Well, Benjamin Sutherland, thank you so much for talking to us about your reporting in The Economist. This latest article was titled "Why Big AI Labs Are Hiring So Many Philosophers." Thanks for joining us.
SUTHERLAND: Thank you, Scott. It's good to be here. Transcript provided by NPR, Copyright NPR.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.
