Fei Fei Li

speaker

223 appearances 1 recordings 1 series first heard Dec 2024 last heard Dec 2024

Fei Fei Li’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

No recordings in the last 12 months.Older appearances are listed below; set an alert to hear about the next one.

Appearances

newest first · ▶ plays the moment

Armchair Expert with Dax Shepard · Fei Fei Li (on a human-centered approach to AI) · 11 Dec 2024 podcast

But once they start to interface with the world, the impact is not necessarily neutral at all. And there is so much humanness in everything we do in technology. And how do we connect that? I decided to talk about that with the interns.

Yeah, it was around that time, 2018 March, I published the New York Times op-ed. I laid out my vision for human-centered AI.

My overarching thesis is that we must center the value of technologies, development, deployment, and governance around people. Any technology, AI or any other technology, should be human-centered. As I always say, that there's no independent machine values. Machine values are human values. Or there's nothing artificial about artificial intelligence. So it's deeply human.

So human-centered AI should be a framework, and that framework could be applied in fundamental research and education. That's what Stanford does. or creating business and products, as Google and many other companies do, or in the legislation and governance of AI, which is what governments do. So that framework can be translated into multiple ways.

Fundamentally, it is to put humans' dignity, humans' well-being, and the value that a society care about into both how you create AI or how you create AI products and services or how you govern AI. So concrete examples, let me start from the very basic size upstream. At Stanford, we created this Human-Centered AI Institute. We try to encourage cross-pollinating interdisciplinary research

study and research and teaching about different aspects of AI, like AI for drug discovery, AI for developmental studies, or AI for economics and all that. But we also need to keep in mind, we need to do this with the kind of norm that reflect our values. So we have actually a review process of our grants.

We call it ethics and society review process, where even when researchers are proposing a research idea to receive funding from HAI, they have to go through a study or a review about what is the social implication? What is the ethical framework?

That's the very fundamental research example. Now, translate to a company. When we think about an AI product, let's say I would love for AI to detect skin condition for diseases. That's a great idea. But starting from your data, where do you curate data? How do you ensure data fairness?

Exactly. What about the scan of the face? And also the algorithm that detects melanoma, is it trained on... Just white folks? Right, exactly. Narrow type of skin or all skins? What's the impact of that algorithm?

So all those are messy human elements. And then you ask about legislation. Then we come to government. Of course. There is always a tension between how much regulation, how do you regulate? Is good policy only about regulating? For example, I firmly believe we actually should have good policy to rejuvenate our AI ecosystem, to make our AI ecosystem really healthy.

For example, right now, the gravitational pull is that all the resources that The data, the computation and the talents are all concentrated in a small number of large companies. It's all for commerce.

No, not at all.

My own lab at Stanford has zero NVIDIA H100 chips. There you go.

And it's not just competition. It's that the problems we work on are curiosity driven and sometimes they are really public good. For example, my own lab, we're collaborating with hospitals to prevent seniors from falling. That is not necessarily a commercially lucrative technology, but it's humanistically important. And universities do all kinds of work like that.

Now our universities in the age of AI is so under-resourced that we cannot do this kind of work. I have been working really hard in the past five years with HAI, with Washington, D.C., with Congress people, senators, White House agencies to try to encourage the resourcing of AI through National AI Research Cloud and data. And then we have legislation and regulation.

How do you thoughtfully put guardrails so that individual lives and dignity and well-being are protected, but the ecosystem is not harmed?

I was also chatting with Yuval. Did he give the C minus grade to humanity? Did he say that?

He said that humanity has gotten a C minus. And I was like, Yuval, you know, I'm a teacher and a mom. When a kid comes home with C minus, you don't throw the kid out. We help the kid to get better. So first of all, you're right. We're not living in a vacuum. And AI also is not living in a vacuum. AI is just one technology that's among many. So I absolutely do believe that there can be cooperation.

How exactly we cooperate, who we cooperate with, and what are the parameters of cooperation is much, much more complicated. Look at humanity. We have gone through this so many times. I mean, you're always right. We have many messy chapters, even nuclear technology. But we have gotten to a state that there is a fine balance at this point of nuclear powers.

I'm not saying that's necessarily comparable. I think it is.

Showing 181–200 of 223 · page 10 of 12