Short Wave - Trailblazing Computer Scientist Fei-Fei Li on Human-Centered AI

Episode Date: November 10, 2023

AI is popping up everywhere nowadays. From medicine to science to the Hollywood strikes. Today, with computer scientist and AI pioneer Fei-Fei Li, we dig deeper into the history of the field, how machines really learn and how computer scientists take inspiration from the human brain in their work. Li's new memoir The Worlds I See traces the history of her move to the U.S. from China as a high school student and her coming-of-age with AI. Host Regina G. Barber talks to Li about her memoir, where the field may be going and the importance of centering humans in the development of new technology. Got science to share? Email us at [e-mail] pcm.adswizz.com for information about our collection and use of personal data for sponsorship and to manage your podcast sponsorship preferences.NPR Privacy Policy

You're listening to Shortwave from NPR. Faye-Fei Lee has been called the godmother of artificial intelligence, or AI. And although she's a computer scientist at Stanford now, her journey in science began with physics. I just loved Albert Einstein. I loved it so much. I went to Princeton University and majored in physics. What captivated Fifei was the bold way physicists questioned everything unknown. What's the smallest particles that make up acid? What is the boundary of the universe? What is the beginning of time?

These questions stayed with her, from her childhood in Chengdu, China, to arriving in the U.S. with her parents as a 15-year-old and into college. It was there where she started reading works from famous physicists and asking some new questions. And I start to realize these physicists who spend their life wondering about atomic world, physical world, also wonders about a world that is equally mysterious and equally daunting. The questions are equally audacious, but seems even harder to answer, and that is about life, about intelligence. One summer in college, Fei-Fei did an internship at UC Berkeley.

The team she joined would be studying molecular neuroscience and biology, and Faye Faye worried she'd feel out of place as a physics student. But when she got off the plane, she was a very much. was greeted by an electrical engineer. He told her they weren't just going to be studying the brain, but trying to reverse engineer it. In their work, they'd show video clips to a cat and try to reconstruct the footage using only the signals detected in the cat's brain. So I kind of pivoted from being so enamored by just the question of the physical world, the atomic world, to the question about the world of intelligence. So all this, you know, the curiosity for brain, curiosity for computation,

for modeling all kind of converged. When Fei-Fei returned to Princeton that fall, she felt like a different person. She knew she wanted to continue this exploration using computation and modeling to understand what she calls the mystery of intelligence, both in humans and machines. To me, both animal human intelligence and machine intelligence are sort of the two sides of the same coin.

So today on the show, how people design designing AI take inspiration from the brain, where the field may be going, and how centering humans in AI development may be the best way to make sure the tech is used for good. I'm Regina Barber. You're listening to Shortwave, the science podcast from NPR. Faye-Fei Lee's book, The World's I See, Curiosity, Exploration, and Discovery at the Dawn of AI, came out this week. It weaves together stories of her coming of age alongside the history and development of AI. She begins the book in 2018. where she's getting ready to testify in front of Congress about the responsible use of AI technologies.

And your goal was to show people that AI could be used for good. Why do you think that needed to be said? Right. So we have to recognize its power. We have to recognize the opportunity, but also recognize the consequences. And I don't know, serendipitously or coincidentally, we're yet again recognizing how important it is to talk about AI communicate that in the policy world and why it's important to use AI for good because it's powerful. So you talked about using AI for good, and I know you've been a proponent of human-centered artificial intelligence. Like in your book, you write, AI must become as committed to humanity as it always has been to science. What does that mean to you? To me, it means that we should put humans in the

center of the development of AI, as well as the deployment application. and governance. And how do you think that will translate into how we use AI? Creating a tool like AI is a feat of human intelligence. In the meantime, we use tool to, you know, make the world better. So that's one side of AI, how it should center humans to hopefully be used for good. Another thing in your book I was fascinated by was some of the mystery in the field. Like, there's a lot we don't know about AI.

Can you talk about that? Yeah, so AI for me. me, especially the private part of AI for me, is a fascinating science, right? Of course. How do you model thoughts? How do you get a computer to see? Because what you get on a photo are just lights and colors and shades, yet you read out a cat. So it is mysterious.

It is whimsical. Now everybody use powerful AI products like chat GPT. But even there, how come it can talk to you in a human-like language, but it does stupid errors amass? You know, how could Einstein, you know, think about E equals MC square and he doesn't have nearly the kind of big data that chat GPT has from the entirety of the internet? These are mysterious questions. Okay, so I want to switch gears for a second. There's also like tons of different areas within the field, but when you've been a pioneer in is computer vision,

like how computers process images, how they quote unquote see. Can you talk more about what computer vision is? Okay, so the way computer sees, first of all, just like humans, right, you have to have eyes and retina to capture the world. You use cameras or sensors. you capture the sensory information, but you have to make sense of this. Well, to make sense, you need to define

what do you want to make sense of? Do you want to see the objects? Or do you want to see surfaces so you can navigate? But one, for example, important task is to see objects. You know, how do you see? There is a car, there is a pedestrian, there's trees, there's pavements, and you have to teach computer through mostly examples.

of these items or photos of cars, photos of pedestrians, or videos, and then you make sure the algorithm learns very well what these things are. And then you want to interrogate it or test it. And mathematically, what we call this process is generalization, is you give it a new car on a different road and see if the algorithm is capable of actually recognizing there is a car on the road. And mathematically, that's what AI is really trying to create these models that can generalize. Okay, so speaking of self-driving cars, your book touches on some ethical questions about AI.

Like, what are the biggest ethical issues in the field right now? I think probably the biggest issue of today's AI is that the technology is developing really fast, but the governance model is still incomplete. And in a way, it's inevitable. I don't think we ever create governance model before a technology is ready to be governed. That's just not how our society works. These are ethical issues. For example, fairness is a very deep issue.

You don't want to be deploying these models to do, say, a micro-loan application assessment when you have historically biased data. drive these financial decision making, right? Some are not ethical per se, but they're profound like jobs, like workforce. You know, this technology right now, the latest wave is really disrupting knowledge workers. Look at the Hollywood artist's strike. It's partially a reaction to what's going to happen to our jobs and to our pay and all that. But policy and business do have power, for example, reskilling and upskilling workers, recognizing and forecasting areas of disruption and understanding how resources can be well used to help in areas or in jobs that

will be most disrupted. All this is part of the solution making. Yeah, I mean, it reminds me of the internet. Like, people didn't expect it to be a place where things like, hate groups could connect what it was. Could it be that we haven't learned from our mistakes from the internet? And for AI, maybe there are uses that we can't even predict. Yeah, exactly, right? So I'm deeply concerned that we might be making similar mistakes again. But I guess the way I cope with my concern is actually to take actions. This is why I wrote this book. This is why I founded the Human Center AI Institute because I think by collaborating with people of all disciplines, we at least have a chance to raise the voice and to look at these issues.

Let's end on a high note. With the potential of AI for the future, what are you most excited about? Yeah, I am very excited by this technology because it's such an innovative tool to superpower humanity's capability to do good. I guess I'm really excited about using AI to augment people's productivity, whether it's our caretakers to elderly, to clinicians, or it's teachers to superpower their teaching and students learning. I think drug discovery, new materials, we've seen protein folding. I think that is boundless possibilities, and that does excite me.

Faye-Fei Lee, thank you so much. Thank you, Gina. Faye Lee's book, The Worlds I See, is out now. You can find a link to it in our show notes. This episode was produced by Rachel Carlson and edited by Burley McCoy. Britt Hansen checked the facts and Patrick Murray was the audio engineer. Beth Donovan is our senior director and Anya Grenman is our senior vice president of programming.

I'm Regina Barber. Thanks for listening to Shorewave from NPR.