Separating AI Hope from AI Hype

**Arvind Narayanan** (0:01)
The people who believe that superintelligence is coming very quickly, tend to think of most tasks that we want to do in the real world as being analogous to chess, where it was the case that initially chess bots were not very good. At some point, they reached human parity, and then very quickly after that, simply by improving the hardware and then later on by improving the algorithms, they're vastly, vastly superhuman. We don't think most tasks are like that. This is true when you talk about tasks that are integrated in the real world, you know, require common sense, require a kind of understanding of a fuzzy task description. It's not even clear when you've done well and when you've not done well. We think that human performance is not limited by our biology. It's limited by our state of knowledge of the world, for instance. So the reason we're not better doctors is not because we're not computing fast enough. It's just that medical research has only given us so much knowledge about how the human body works and how drugs work and so forth. That's one. The other is you've just hit the ceiling of performance. The reason people are not necessarily better writers is that it's not even clear what it means to be a better writer. It's not as if there's going to be a magic piece of text that's going to persuade you of something that you never wanted to believe, for instance. We don't think that sort of thing is even possible. So those are two reasons why in the vast majority of tasks, we think AI is not going to become better, or at least much better than human professionals.

**Cindy Cohn** (1:34)
That's Arvind Narayanan explaining why A.I.s cannot simply replace humans for most of what we do. I'm Cindy Cohn, the Executive Director of the Electronic Frontier Foundation.

**Jason Kelley** (1:46)
And I'm Jason Kelley, EFF's Activism Director. This is our podcast series, How to Fix the Internet.

**Cindy Cohn** (1:54)
On this show, we try to get away from the dystopian tech doomsayers and offer space to envision a more hopeful and positive digital future that we can all work towards.

**Jason Kelley** (2:04)
And our guest today is one of the most levelheaded and reassuring voices in tech.

**Cindy Cohn** (2:08)
Arvind Narayanan is a professor of computer science at Princeton and the director of the Center for Information Technology Policy. He's also the co-author of a terrific newsletter called AI. Snake Oil, which has also become a book, where he and his colleague Sayash Kapoor debunked the hype around AI and offer a clear-eyed view of both its risks and its benefits. He's also a self-described techno-optimist, but he means that in a very particular way. So we started off with what that term means to him.

**Arvind Narayanan** (2:40)
I think there are multiple kinds of techno-optimism. There's the Mark Andreessen kind where, you know, let the tech companies do what they want to do, and everything will work out. I'm not that kind of techno-optimist. My kind of techno-optimism is all about the belief that we actually need folks to think about what could go wrong and get ahead of that so that we can then realize what our positive future is. So for me, you know, AI can be a profoundly empowering and liberating technology. In fact, going back to my own childhood, this is a story that I tell sometimes. I was growing up in India and frankly, the education system kind of sucked. My geography teacher thought India was in the Southern Hemisphere. That's a true story.

**Cindy Cohn** (3:27)
Oh my God.

**Arvind Narayanan** (3:29)
Oops. And, you know, there weren't any great libraries nearby. And so a lot of what I knew, I not only had to teach myself, but it was hard to access reliable, good sources of information. We had had a lot of books, of course, but I remember when my parents saved up for a whole year, and bought me a computer that had a CD-ROM encyclopedia on it, that was a completely life-changing moment for me, right? So that was the first time I could get close to this idea of having all information at our fingertips. That was even before I kind of had Internet access even. So that was a very powerful moment. And I saw that as a lesson in information technology, having the ability to level the playing field across different countries. And that was part of why I decided to get into computer science. Of course, I later realized that my worldview was a little bit oversimplified. Tech is not automatically a force for good. It takes a lot of effort and agency to ensure that it will be that way. And so that led to my research interest in the societal aspects of technology as opposed to more of the tech itself. Anyway, all of that is a long-winded way of saying, I see a lot of that same potential in AI, that existed in the way that Internet access, if done right, has the potential and has been bringing a kind of liberatory potential to so many in the world who might not have the same kinds of access that we do here in the Western world with our institutions and so forth.