Campbell Brown, Senior Advisor, TollBit
Fei-Fei Li, Sequoia Professor, Computer Science Department; Co-director, Stanford Institute for Human-Centered AI, Stanford University
Moderator: Mallun Yen, Founder and CEO, Operator Collective; Co-chair, Fortune MPW Summit
Fei-Fei Li, Sequoia Professor, Computer Science Department; Co-director, Stanford Institute for Human-Centered AI, Stanford University
Moderator: Mallun Yen, Founder and CEO, Operator Collective; Co-chair, Fortune MPW Summit
Category
🤖
TechTranscript
00:00So welcome, Fei-Fei and Campbell.
00:02It's great to have you here today.
00:03So thank you for being here with us.
00:06So before I start, I wanted to let the audience know
00:08that we are going to try to take,
00:09we probably only have time for one question or so,
00:12so if you do have a question,
00:13please flag down the person with mics
00:15and we will try to get to you.
00:17So great to see you guys.
00:19Good to be here.
00:20Okay, we're gonna start with a question.
00:22How many people in the room think that AI
00:25will be a net positive to humanity?
00:31Okay.
00:33Great.
00:34A net negative?
00:40I see very few.
00:42Now, does that surprise you?
00:46It does.
00:47I didn't realize how many people think it's a net positive.
00:51I actually think we've come a long way
00:56in terms of just over the last three years,
01:00partly because people like Fei-Fei who have taken,
01:02who have given me a framework for how to think about it,
01:05and I think have taken a much more measured approach
01:08to talking about what it means,
01:10both the positives and the negatives,
01:13as opposed to some of the other people
01:14who are very visible around AI
01:17and have been at the forefront of it
01:18who have been somewhat hyperbolic, perhaps,
01:21about the positives and negatives.
01:23And so I think you're seeing people become more comfortable
01:26with more people like Fei-Fei sort of taking center stage.
01:30Thank you, Campbell.
01:32It's true.
01:33I do believe it's true.
01:34I think you're absolutely right.
01:36So you two both raised your hand,
01:38and I also know that you are both excited about AI,
01:41and in fact, this year,
01:43you both transitioned into the AI startup world.
01:46So Campbell, you started out as an NBC news correspondent
01:50for the White House.
01:50You were an anchor.
01:52You joined Meta to lead partnerships there,
01:54and then very quickly,
01:54you found yourself right in the midst of content and AI.
01:58And we actually met through Tollbit, which we mentioned,
02:03which is, for fair disclosure,
02:05Operator Collective is an investor in Tollbit,
02:07and which is content licensing.
02:10And so now you're working on another stealth AI startup.
02:15So what led you to leave Meta and enter into this realm?
02:19Yeah, when I was at Meta,
02:20the two things I was focused on
02:22were content and business models
02:25and helping people, mostly news,
02:26try to make money on our platforms,
02:28and then misinformation on the Meta platforms.
02:31And so I think going back to when ChatGPT
02:33was first released, it was very clear to me,
02:36I think a lot of us,
02:38that this was gonna really disrupt
02:40the sort of information ecosystem
02:43in a way that we had never seen before,
02:45even with social media becoming so disrupted.
02:49And I wanted to try to get ahead of it,
02:50and I see everything right now
02:53through the lens of my two teenagers
02:55and how they consume information.
02:56And so a big driver for me
02:59was trying to figure out how we ensure
03:03that there is accurate, reliable information
03:06that's part of the LLM training process.
03:09And we're not there yet.
03:11I was on ChatGPT last night and asking a question.
03:14It gave me an answer that was just completely false
03:17and so authoritatively false.
03:20And you recognize that there's still a lot of work
03:23that has to be done on that front.
03:24And part of what the startup
03:27that Malin and I are both involved in is doing
03:30is trying to create a marketplace
03:31to ensure that creators of content,
03:35reliable content, timely content,
03:40are being compensated by the tech companies
03:43who are building these models.
03:46Fei-Fei, you are the godmother of AI.
03:49Give me.
03:50Give me.
03:54So this has been your life's work.
03:56You began working on ImageNet in 2006,
03:59and when you did, you were repeatedly told
04:01this was a go-nowhere project,
04:03that it was, yeah, your PhD students
04:05were never gonna get their PhDs.
04:08And at the time, most AI research
04:10was focused on models and algorithms,
04:12and ImageNet was focused on expanding and improving
04:14the data available to train the algorithms,
04:18which we now know has been crucial to the development of AI.
04:22So what prompted the move this year
04:23from years and years in academia to found your own startup?
04:28Yeah, okay, how many hours do we have?
04:31I'll try to be quick.
04:32First of all, I have lived in the AI technology ecosystem
04:37for more than two decades,
04:39and especially privileged to be working at Stanford.
04:42So I always felt that the world I live in
04:45is a very porous world between academia, technology,
04:50industry, entrepreneurship.
04:52So it didn't feel like I was abandoning
04:55a particular part of my world.
04:58So you're right that my own life's journey
05:02has been seeing the transition and the coming of age of AI,
05:07starting from the ImageDev project,
05:10which is seeing that the critical element,
05:13now that we know we have known
05:15as the three fundamental elements of AI,
05:19the computing chips, the neural network algorithm,
05:22and big data, and that was my ImageDev project.
05:26Now, fast forward to now, to 2024,
05:30one thing that has been always fascinating me
05:33and driving me is really unlocking the full capability
05:39of the science of intelligence, of machine intelligence.
05:44And if you look at the evolution of intelligence,
05:47Campbell already mentioned LLM.
05:50Most of you probably know it as large language models.
05:53We are starting to realize that we can create machines
05:57that unlock the linguistic capability of intelligence,
06:03whether it's conversations, summarizations,
06:06knowledge, search, and reasoning.
06:10But there is a huge part of intelligence
06:13that advanced animals like humans
06:16have developed over the course of evolution,
06:20and that is everything to do with seeing
06:23and interaction with the 3D space,
06:26which I call spatial intelligence.
06:29In fact, evolution took 540 million years
06:33to create the capability of perceptual reasoning
06:36and understanding and interaction,
06:39whereas language, compared to the half a billion years,
06:42took about a million years-ish.
06:45So there's a profound capability
06:48in the science of intelligence
06:50that is centered around the ability to understand 3D world,
06:55reason with the 3D world, generate 3D world,
06:58and interact within 3D world.
07:01And that's what I'm so excited about,
07:04that together with my former students and co-founders,
07:09we feel this is the moment to bring this capability
07:13to the world and unlock so many use cases
07:17and business opportunities.
07:19It's so fascinating.
07:20So Fei-Fei, when I called you the godmother of AI,
07:24I saw you sort of cringe a little bit.
07:26And in fact, when we were talking last night,
07:30you said that when you thought about starting this company,
07:33you weren't sure that you should be the CEO.
07:36And then I'm gonna say one more thing
07:37that I frankly find just really annoying,
07:39not that you did, that speaking of media,
07:42was that, do you guys remember when the New York Times
07:45came out with that list in December
07:46that was who's who behind the dawn of the modern AI movement?
07:51Yeah, yeah, exactly.
07:53And so there were the CEOs of OpenAI,
07:56Anthropic, Microsoft, Google.
07:58There were some researchers
07:59who made tech advances in AI,
08:01and there were a couple of investors
08:02who mostly just invested in the AI companies
08:05and didn't really do anything on the tech.
08:07So there were 12 women, I mean, sorry, 12 men,
08:11and no women, and most glaringly,
08:14Fei-Fei, you were not on that list.
08:16And even though ImageNet, your work was the foundation
08:19that enabled all of those companies, including the research.
08:22Wow. Yeah, yeah.
08:23Yeah.
08:24Woo!
08:31So I just had to get that off my chest.
08:35You weren't the only one.
08:36Twitter exploded after that with all the women
08:39who were obviously very offended,
08:41and I think even some of us spoke to reporters
08:44at the New York Times about it
08:45because it was such an outrage,
08:47and I think they feel like they,
08:49I mean, I think they recognized
08:50that was a total media failure,
08:51and they really screwed up.
08:53Yeah, so when you see the headlines today,
08:56they do tend to be dominated by one gender,
08:59and so would love to hear your reaction to this.
09:04Well, thanks, Mellon.
09:06Yeah, so first of all about this Godmother of AI title,
09:11it came around, I don't know, a year ago.
09:14I start seeing that.
09:16I had to take a pause because it really is not my personality
09:20to call myself godmother of anything.
09:24It really, I did cringe, but I had to,
09:29you know, my natural reaction is let's just reject this.
09:32I don't want this title,
09:34but I really had to take a pause and realize
09:38in the entire history of science and technology,
09:41so many men are called whatever,
09:44founding fathers, godfathers,
09:47and if women are so readily rejecting that title,
09:52where is our voice?
09:53And I'm standing here.
10:02So I kind of with a cringe, I did not fight back
10:05that title, and yes, the New York Times list.
10:10I actually am a pretty shy person,
10:13but I think I went on social media and called.
10:17It was just wrong.
10:19It was wrong that in 2023,
10:22any media will look at a technology
10:24that so many people from all walks of life,
10:27not just men and women,
10:28but people of all backgrounds have contributed,
10:32and they didn't even try to make a little bit of effort,
10:36and we do still see that today,
10:39and this is why, as Campbell was alluding earlier,
10:42that it's so important for me,
10:44no matter how advanced a technology,
10:47how fancy, advanced, and cutting-edge,
10:50and sci-fi a technology looks like,
10:53we have to put humans in the center of it.
10:56You know, we have to put human representatives in the center.
11:00We have to put human well-being in the center,
11:04and we have to put human values in the center
11:06of this development of this technology.
11:08It's just so important.
11:15Campbell, when Fei-Fei's book first came out,
11:17you were one of the first ones to raise your hand
11:20and say, we gotta get this out here,
11:21so tell us a little bit about that,
11:23and by the way, if you haven't gotten this book,
11:26it's a phenomenal book.
11:28It was one of the first books I read
11:30that was kind of an introduction,
11:32and I highly, highly recommend for just,
11:36if you're curious about AI and everything,
11:37but also just as an inspiration to young women,
11:43your life story, and maybe you can give people
11:47just the thumbnail even of being a young girl in New Jersey
11:53and growing up with your parents.
11:54I think it's pretty incredible,
11:56and there are a lot of young women at META on our AI,
11:59or when I was formerly there,
12:01but on the AI research team,
12:03who I heard speak about you as that inspiration.
12:06We had a book party when Fei-Fei's book came out,
12:08and several of those women were there,
12:09so I do think it is not only for getting a primer,
12:14if you will, on what's happening in the history,
12:17but also one of the most inspiring stories
12:21of anyone sort of overcoming odds in this country.
12:26Campbell, you're one of the first people
12:27who threw a book party for me.
12:29I was so grateful,
12:32overseeing New Jersey from your New York home,
12:36but it's true.
12:37I wrote the book, I wanted to tell the AI story.
12:41My personal coming of age and AI's coming of age
12:44is so intertwined in so many ways,
12:48including how my work was involved in this,
12:50but it was also, I was deeply encouraged
12:53by my friends and colleagues at Stanford
12:56to tell the story so that the young women,
12:59the immigrants, the people of all walks of life
13:03will find a voice in AI,
13:07because today's AI, like you say,
13:10it's so hyperbolic, it's so,
13:14we hear words like global scale,
13:17infinite productivity, billions of whatever,
13:21but at the end of the day,
13:24whoever makes AI changes the world,
13:27and it's so important people from all backgrounds
13:30feel they have a role,
13:33either it's in development of AI
13:36or in creating products and business of AI
13:40or doing interdisciplinary research of AI
13:43or using AI to create something they care about
13:49or in the governance of AI.
13:50Everyone who cares has a role,
13:53and that was the reason I wrote this book.
14:00So both of you actually worked at,
14:03have worked with social media sites.
14:04You met up for seven years,
14:06and you were on the board of Twitter for two years
14:08until it was dismantled.
14:10So it's widely understood that LLMs are trained.
14:12I'm sorry.
14:15You used a very interesting word.
14:17I did.
14:22Anyway, it's widely understood that LLMs are trained
14:24on all the content that's out there,
14:26and then some reports suggest that 90%
14:28of the information on the internet
14:29will be AI generated within a few years.
14:31So we've talked about the lack of women in AI,
14:35the social media, news sources.
14:38Campbell, any words of thought?
14:41I mean, the one thing that does worry me right now especially
14:46is the misinformation.
14:48I mean, I don't know how closely people are following it,
14:51but these tools are incredible,
14:55and it's almost impossible to tell.
14:57I mean, it is impossible to tell what's real and what's not,
15:00and we haven't developed tools yet to help us do that,
15:03whether it's watermarking for things
15:05that were created by AI or things that weren't
15:08or identifying what is and isn't.
15:10And you've seen examples, the hurricane recently.
15:14There was so much misinformation flying around
15:17to a point where it was literally endangering people
15:20who FEMA was trying to help,
15:22and you've seen numerous examples.
15:25I don't think anything has sort of shocked us to the point
15:28that we're ready to address it in a regulatory way
15:31or anything yet, but I am very worried
15:35about how we think about this and keep track of it.
15:39The one thing I would say that I think is meaningful
15:42for people in this room is that brand, I think,
15:44becomes increasingly important to be able to know
15:47that this is a trusted brand I'm familiar with.
15:50When you see something, whether it's a news organization
15:52or a company, in a world where we're just being flooded
15:56with information and you cannot identify the source,
15:59looking to trusted brands seems like it's going
16:02to matter again, and it's worth investing in.
16:06Thank you so much, Campbell and Fei-Fei,
16:08for joining us today.