AI that mimics humans would ‘destroy the value of human labor,’ expert says

Stanford Digital Economy Lab Director Erik Brynjolfsson joins Yahoo Finance Live to discuss the AI hype, its implications and values, and the outlook for AI on society.

Video Transcript

[AUDIO LOGO]

BRIAN SOZZI: All right with ChatGPT taking the world by storm, the spotlight on artificial intelligence is brighter than ever before. What exactly is the scope of AI and what are the implications that lie ahead? Here to discuss is Erik Brynjolfsson, Stanford Digital Economy Lab Director. Good morning to you. Before we even get too deep--

ERIK BRYNJOLFSSON: Morning.

BRIAN SOZZI: --down this rabbit hole, what is the difference-- you know, what is making this crop of AI different than in years past? And how does it compare to, let's say, even an algorithm? I think there's a lot of confusion out there.

ERIK BRYNJOLFSSON: Sure. Sure. Well, ChatGPT is an example of a new class of AI that is, I think, a really big breakthrough. It's-- these are called foundation models. They include not only large language models that can write stories, poetry, email, ads, many other types of texts.

But many of you have probably played with DALL-E that make images. Others make videos, audio, even write computer code. These tools are having a set of implications that I think are bigger than even their developers expected, initially. Trillions of dollars of value will be created.

JULIE HYMAN: And Erik, you've written about-- you know, there has been a lot made and a lot of, I think, fear out there, as well as excitement, over what these new developments can mean. You've written about what you call the Turing Trap, one of the potential things that we need to watch out for. Can you explain that to people?

ERIK BRYNJOLFSSON: Sure. Well, what-- you know, Alan Turing famously proposed that the test for intelligence was what he called-- what we later called the Turing Test, which was how similar can a AI be to a human? And trying to mimic humans has been kind of a goal of a lot of computer scientists ever since. You know, can we fool humans so you can't tell the difference.

I think it's a very evocative goal but it's also a trap. And the reason it's a trap is that if we make AI that mimics humans, it actually destroys the value of human labor. And it leads to more concentration of wealth and power.

But there's an alternative approach, which is making AI that augments humans, that allows us to do new things we never did before. Don't try to mimic us but try to extend our capabilities. AI that does that is more likely to lead to a flourishing of wealth and more widely shared prosperity.

JULIE HYMAN: So where the industry is going now, which of these things is the goal? In other words, do you think that some of the goals are sort of on the wrong path right now?

ERIK BRYNJOLFSSON: They sometimes are. Although, I'm actually pretty optimistic. I've seen lots of good uses of these tools to even enhance invention and innovation. People are using them to invent new compounds. I was talking to a colleague and his-- one of his grad students had written up a research proposal and he couldn't quite make sense of it. And then he ran it through ChatGPT. And he said, ah, that's what she was trying to say. And he showed it to her. She said, yeah, that's what I was trying to say.

I think there's a lot of people who have some brilliant ideas, they have trouble expressing them. And these tools can be used to help make it more possible for others to get what they're trying to say.

BRIAN SOZZI: Erik, we were talking about today the new Pixel phone, the Super Bowl ad where the Magic Eraser, where you can race somebody or an object from a photo. That-- is that tech for good? And what is the societal impact of something like that?

ERIK BRYNJOLFSSON: I'm really worried, not just about that but more broadly, deep fakes. Just over the weekend, I saw the first deep fake in the wild. There was an ad running on TikTok where some famous people were saying things that they didn't really say. And it was impossible to tell the difference.

So this is going to proliferate. As the cost of making deep fakes goes down, the number of them going to increase exponentially. And we're going to have to come up with some ways of understanding what's real and what's not real.

JULIE HYMAN: This all seems to be moving so fast. I wonder if-- you know that a lot of that stuff is going to fall through the cracks. In other words, like you say that we have to be prepared for deep fakes. It doesn't feel like we're prepared for deep fakes. It doesn't feel like we're prepared for some of the other pitfalls you're talking about.

ERIK BRYNJOLFSSON: I totally share your concern. The technology is racing ahead. And the next 10 years could be some of the best 10 years we've ever had or some of the worst. But we have to adapt our institutions.

One of the reasons I started the digital economy lab here at Stanford was to try to close that gap, to study the economic implications, the societal implications, and we need to hustle to catch up. I'm proud of my colleagues, my very smart colleagues that are inventing these amazing technologies.

But the rest of us in business, and economics, and social sciences need to get on the ball and think about how we can make sure our society is ready for these changes. If we do it right, I do think that we're going to have unprecedented productivity and growth, and we can have shared prosperity but it's far from inevitable.

BRIAN SOZZI: Erik, I think investors have just sent a lot of these AI stocks through the roof. You know, if they bought one of these stocks-- let's say, a C3.ai. If they bought one of these stocks because of announcement of AI, how does AI impact their bottom line? Is it is it a one-time shot? Is it a jolt? Or is it just over time these companies will just see just a faster rate of earnings because of AI?

ERIK BRYNJOLFSSON: Well, we're in very early days. I mean, there's been a one-time shock. But then there's gonna be another one-time shock and more and more shocks coming in the pipeline. I already know that there are new versions of these large language models that are set to be released shortly that are dramatically better than the ones we've already seen. They're being used in lots of other applications.

So we're like in the first inning of this. And the winners right now may or may not be able to extend their lead. It may be that a whole different set are going to emerge. I see a lot of turbulence.

JULIE HYMAN: Now, I know you're not coming at it from an investment perspective. But I do wonder, you know, for laypeople out there who are buying up AI stocks right now, is there any kind of lens that they can apply or analysis they can apply without being super steeped in it, to know who are gonna be the winners and losers in this thing?

ERIK BRYNJOLFSSON: You know, ironically, I think it's almost easier to see the losers than the winners. A lot of legacy companies that were relying on just the old way of doing things, their value is going to be threatened. Some of the movie studios or others that have amazing content creation capabilities, they're now going to become stranded because we can do it much more cheaply and do them in new sorts of ways.

It's a good bet to focus on companies that have some of the best talent and have some of the best resources. Here within a few miles of my house here at Stanford, there are dozens or hundreds of companies that are tapping into this. And I see a number of years of very fast innovation in this field.

BRIAN SOZZI: Exciting times, indeed. Erik Brynjolfsson, Stanford Digital Economy Lab Director, good to see you. Appreciate it.