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

It was pretty bad.

Yeah, I'll start with this room and write everything. Well, okay. So first of all, I've had years of training as a scientist. So after you formulate a hypothesis, you do have to come up with a plan. My PhD thesis had a mini version of ImageDesk. So I got a little bit of practice. But yeah, our idea was to create a data set and a benchmark to encompass all the visual concepts in the world.

So we had to start with WordNet. We had to figure out what is visual. We have to figure out what are the concepts we need and then where to get the source images and how to curate it. Every step of the way, like Dax, you were saying, we were just way too optimistic at the beginning.

Yeah, I was just fearlessly stupid.

And then we start to hit all these walls of Jia and I and other students. But Jia was the main student. We had to just deal with every obstacle that came. Now, science is a funny thing, right? Sometimes serendipity makes a world of difference. What was really critical was the Amazon Mechanical Turk, the crowdsourcing platform. Amazon, nothing to do with us.

We're like, oh, we have all these servers sitting in our data centers and we have nothing better to do. Let's make an online worker platform so people can just trade little tasks.

Exactly. Which I didn't know it exists. I was in New Jersey, Princeton and trying to pull my hair out. And then some student who did his master at Stanford came to Princeton and just mentioned it casually and said, do you know this thing? That was really, really quite a moment for me.

Yeah, 10x. That was one of the technical breakthrough that really carried this whole project.

Not only years and years, the budget. Hiring undergrads or whatever just doesn't cut it. The budget was not going to cut it. My tenure was on the line. It was a dicey few moments.

Yes, I did go rogue a little bit.

So it's very important to call out the year. My sabbatical at Google was 2017 and 2018. That was my first sabbatical. I finally had a sabbatical. And it was a conscious decision for me to go to Google because this is right after AlphaGo. So AI was having its first hype wave, at least public moment. And Silicon Valley, of course, was ahead of the curve and new AI was coming.

So I had multiple choices, but I really wanted to go to a place for two reasons. One is to learn the most advanced industry AI, and Google was by far the best. But also to go to a place where I can see how AI will impact the world. And Google Cloud was a perfect place because cloud business is serving all businesses.

So at Cloud, being the chief scientist, I was able to see the technology translating to product and product impacting healthcare, hospitals, financial services, insurance companies, oil and gas companies, entertainment, agriculture, governments, and all that. But in the meantime, it was confirming my hypothesis that this technology has come of age and will impact everyone.

It was the first tech clash. 2017 was right after Cambridge Analytica.

Yeah, social media's algorithmic impact can drive societal changes. It was also around the time face recognition bias was being publicized for the good reasons of calling out bias. So before that, tech was a darling. The media doesn't report tech as a force of badness.

That's true. Hollywood is always ahead of the curve on that.

So it was a tech clash that came at us very fast. Google has had its own share. I was actually also witnessing the struggle that Google was coming to terms with defense issues.

Yes. In hindsight, it was a mixture of many things. It wasn't a single event. I remember it was summer of 2018 and we were just coming off this turmoil. In hindsight, they're small, but at that point, And I was just like, I'm about to speak to maybe my memory is wrong, but I thought it was 700 interns from worldwide who worked at Google that summer. And they're the brightest from the whole world.

And they were hand selected by Google. You know, Google is really a machine of talent. And what do they want to hear from me? Of course, I can talk about come work at Google. That's my job as someone who was working at Google. But I felt there was more I should share.

Really coming from the bottom of my heart at that point, something that you will appreciate is that the math behind technology is clean, but the human impact is messy. Technology is... so much more multidimensional than equations.

Showing 161–180 of 223 · page 9 of 12