Fei-Fei Li on spatial intelligence and human-centered AI

How can we use AI to amplify human potential and build a better future? And what exactly does “AGI” even mean? To kick off Possible’s fourth season, Reid and Aria sit down with world-renowned computer scientist Fei-Fei Li, whose work in artificial intelligence over the past several decades has earned her the nickname “the godmother of AI.” An entrepreneur and professor, Fei-Fei shares her journey from creating ImageNet, a massive dataset of labeled images that revolutionized computer vision, to her current role as co-founder and CEO of the spatial intelligence startup World Labs. She explains why spatial intelligence—the ability to perceive and interact with the 3D world—is so crucial for AI's development and how it could lead to breakthroughs in fields like medicine, climate, and education. They get into regulatory guardrails, governance, and what it will take to build a positive, human-centered AI future for all. For more info on the podcast and transcripts of all the episodes, visit https://www.possible.fm/podcast/ Topics: 1:55 - Hellos and intros 3:43 - ImageNet and the interplay between data and models 6:06 - World Labs and spatial intelligence 10:03 - Boundaries between 3D physical and digital worlds 11:50 - The difference between LLMs and LWMs 13:02 - What humans are capable of creating with technology 14:04 - Key principles of AI: human agency and respect 17:16 - Stanford Institute for Human-Centered AI 19:13 - What this moment in AI means for humanity 21:06 - Cross-sector collaboration 25:10 - AI4ALL program and the importance of diversity in AI development 27:00 - Midroll ad break 27:09 - Using AI to improve healthcare delivery and treatment 30:20 - Founding history of AI and the meaning of the term “AGI” 33:00 - Future of agentic AI and voice 34:42 - Fei-Fei’s mentor and his advice 37:18 - Rapid-fire questions Possible is an award-winning podcast that sketches out the brightest version of the future—and what it will take to get there. Most of all, it asks: what if, in the future, everything breaks humanity's way? Tune in for grounded and speculative takes on how technology—and, in particular, AI—is inspiring change and transforming the future. Hosted by Reid Hoffman and Aria Finger, each episode features an interview with an ambitious builder or deep thinker on a topic, from art to geopolitics and from healthcare to education. These conversations also showcase another kind of guest: AI. Whether it's Inflection’s Pi, OpenAI’s ChatGPT or other AI tools, each episode will use AI to enhance and advance our discussion about what humanity could possibly get right if we leverage technology—and our collective effort—effectively.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:03Speaker 00

  Humans are capable of creating godlike technology so that we can improve our medieval institutions and raise above our paleolithic emotions or channel our paleolithic emotions into creativity, productivity, and benevolence.

• 0:37Reid Hoffman?

  Hi, I'm Reid Hoffman.

• 0:39Aria Finger?

  And I'm Aria Finger.

• 0:40Reid Hoffman?

  We want to know how together we can use technology like AI to help us shape the best possible future.

• 0:47Aria Finger?

  We asked technologists, ambitious builders, and deep thinkers to help us sketch out the brightest version of the future, and we learn what it'll take to get there.

• 0:58Reid Hoffman?

  This is possible.

• 1:03Aria Finger?

  Reid, it is so good to be back. Happy New Year, even though we're already halfway through the month, we can still say that. And of course, welcome back to our loyal listeners. Thank you so much for listening. If you caught last month's featured mini series called Artful Intelligence, we hope you enjoyed it.

• 1:20Reid Hoffman?

  Yes. Happy New Year, Aria, and great to be with you. The turn of the calendar year is always a good time to reflect and reset. For me, the end of 2024 actually triggered a good deal of reflection on the state of AI. November marked two years since Chat GPT was released to the public, and it's hard to overstate just how much has happened in the world of AI development since then.

• 1:44Aria Finger?

  I could not agree more. And as we look ahead, especially with respect to AI, I think it's valuable to put into context how far we've come in such a short time.

• 1:55Reid Hoffman?

  Well, lucky for us, we're joined today by the godmother of AI. In addition to being a brilliant and world-renowned computer scientist, entrepreneur, professor, and humanist, Dr. Fei-Fei Li is also a friend. I'm excited to have her with us to talk about her pioneering work in AI, including in her current role as co-founder and CEO of the spatial intelligence startup World Labs.

• 2:22Reid Hoffman?

  We'll get into the mechanics and value of training machines to develop human-like spatial intelligence, along with her views on public-priorit partnerships, regulation, and the importance of empathy in AI innovation. These are themes Fei-Fei touches on in her moving memoir, Worlds I See, which details both her rise as one of the world's leading AI experts, but also her childhood in China's middle class and the difficult transition she and her family faced when they immigrated to the United States.

• 2:55Aria Finger?

  Fei-Fei is also the creator of ImageNet, the database that revolutionized computer vision nearly two decades ago. As Stanford's inaugural Sequoia Professor of Computer Science and co-founder of the Human-Centered AI Institute, she has a long track record as a leader in ethical, inclusive, and human-centered AI innovation. She has been named one of Time Magazine's 100 Most Influential People in AI. And honestly, that's just the tip of her biographical iceberg.

• 3:27Reid Hoffman?

  Here's our conversation with Fei-Fei Li.

• 3:37Reid Hoffman?

  Fei-Fei, it's excellent to see you. Welcome to Possible.

• 3:40Speaker 00

  Likewise, great to see you in Aria.

• 3:43Reid Hoffman?

  What gave you the idea for ImageNet? Like, what was the "ah, we need to do this"?

• 3:50Speaker 00

  It's very hard to pinpoint a very moment, but it's centered around the time of 2006, where I was so deeply doing research using machine learning algorithms in trying to understand objects and images. And no matter where I look, I couldn't escape the fact that there is a mathematical concept called overfitting with machine learning models. And this is when the model complexity and the data the models are using don't quite match, especially if the data...

• 4:35Speaker 00

  It's not just the data volume, it's the data complexity and data volume don't really...

• 4:42Speaker 00

  Drive models in in an effective way. And of course, not all models are created equal. We now know neural network models have so much higher capacity and and representational power. But but those jargons aside, there is definitely an interplay between data and model. And everywhere I look, people are not paying attention to data. We're only paying attention to model. And that really was kind of the moment I had the insight. I think we need to not only look at models or in a way or look in the wrong way. We need to look at data and use data to drive models.

• 5:27Speaker 00

  And of course, at that point, moved my early career job to Princeton as a faculty. I came upon the work called WordNet. And WordNet was nothing to do with computer vision. But it was a wonderful way to organize concepts in the world. And I like the name WordNet. And one thing led to another image that came about because of the need. I I I was so passionately believing the need for big data and diverse representation of the visual world.

• 6:06Reid Hoffman?

  I started kind of the at a midpoint in your AI career with kind of the amazing ImageNet. And now we have World Labs. And so I want to kind of draw the line from ImageNet to World Labs. So what's what's the idea with World Labs? What's the thing that you are building towards that is a key portion of kind of where we're going and how to understand this, both World Labs itself and it's as a trend in AI?

• 6:32Speaker 00

  Yeah, Reid, we talk about this, right? It's our favorite topic is where technology is going. One thing I have been obsessively thinking all my career, especially post ImageNet, to be honest, is really what is intelligence and how do we make intelligence happen in machines? And it really, really, to me, boils down to two simple things. If you look at human intelligence. One is that we say things. We use linguistic.

• 7:08Speaker 00

  Communication as a tool to talk, to organize our knowledge and to communicate. But there is another, uh, the, the half of intelligence that's so profound to us. And it boils down to we do things. You know, we do it by cooking an omelet. We do it by, you know, taking a hike. We do it by friends having fun and really enjoying each other's presence that goes way beyond any word we say. It's just the way we can comfortably sit in front of each other, hold a beer can. All this, right? It's part of intelligence.

• 7:50Speaker 00

  And that part of intelligence is really grounded in the ability to make sense of the three D world we live in, to perceive it and to translate it into a, a, a set of, you know, understanding and reasoning and, and predictions so that we can do things in it. That ability, in my opinion, is called spatial intelligence. It's really the fundamental native ability that embody intelligent animals like humans have, which is the, the ability to deal with three D space.

• 8:32Speaker 00

  So ImageNet happened because I was on a quest of putting labels on the pixels in two D images. And two D images for humans are projections of the three D world. So you can see it was a baby step towards understanding the fuller world, visual world we live in. And that baby step was critical because whether it's for humans or animals or machines, comprehending those objects and images, labeling it is a key first step. But now that, um, gosh, more, more than fifteen years, fourteen years have passed.

• 9:17Speaker 00

  I think we're ready for a much bigger quest. It's almost a home run quest to unlock the most important other half of intelligence, which is.

• 9:30Speaker 00

  The the question of spatial intelligence. Now, what makes spatial intelligence really interesting is that it has actually two aspects. One is the physical 3D world. Another one is the digital 3D world. And we never really were able to live in in between. But spatial intelligence now can be a unifying technology that makes sense of both 3D grounded world as well as digital 3D world.

• 10:03Aria Finger?

  So when I think about sort of the the promise of spatial intelligence, you know, if you went back to 1880, horse-drawn carriages, unpaved roads, like you would be like, this is a totally different world. But if you go back to 1980, okay, like, people are driving different cars, but they're living in the same buildings. They're still driving cars, sort of like the the mechanics of this sort of real world are pretty much the same. Do you think that there sort of this other half of intelligence will change that over the coming decades? That we'll actually see the great transformation in the real world that we've seen in the digital world over the past few few years?

• 10:38Speaker 00

  I think so, Aria, and I think that the line between real and digital will start to blur. For example, I I think about me driving on the highway and if I have a flat tire. I have a feeling I'm gonna, despite the fact that I'm a technologist, I have a feeling I'm gonna have a a hard problem with that flat tire. But if I can wear a glass or even point my phone to the car I'm having the flat tire and just collaborate with that potential application to to guide me through the process of changing that tire,

• 11:20Speaker 00

  whether it's through visual guidance or some dialogue or mixture.

• 11:26Speaker 00

  I think that is a very mundane daily life example that really breaks the boundary of physical 3D world and digital 3D world. And that image of technology like this empowering people, whether it's changing a flat tire or doing a cardiology surgery is a really exciting imagery for me.

• 11:51Aria Finger?

  And so you say, you know, you use LLMs all the time to teach yourself things, which I always think like should be inspiring. Like my kids, they're always like, oh, I don't, I'm so good at math. I don't need to learn anymore. And I can be like, you know, Fei-Fei Li is using LLMs to learn. I think you have some, you know, you have some more to go. But what do you, what do you see when you're talking about large world models versus LLMs? Like, how do you explain that difference to people? And how do you think that's going to play out in the future?

• 12:18Speaker 00

  Well, fundamentally, like I said, one is about saying things. The other one is about seeing and doing things. So, so they are very fundamentally different modality. The large language models, the basic unit are lexicons, whether it's a letter or a word. And in our model, the world models, we use the basic units are pixels or voxels. So they are very different language. I almost feel like language is the language of human. 3D is the language of nature. We really want to get to a point that AI algorithms allow people to interact with the pixel world, whether it's virtual or physical.

• 13:02Reid Hoffman?

  Your answer is reminding me of another quote you've used, which is quoting sociobiologist Edward O. Wilson saying, "We have paleolithic emotions, medieval institutions and godlike technology, and it is terrifically dangerous." So given, you know, kind of reasoning, language of nature, education of people.

• 13:23Reid Hoffman?

  How do you invert that? And what is the opportunity upon us for humanity in the age of AI?

• 13:29Speaker 00

  Yes, I still believe that. And because I believe that you and I and our friends started the Human Center AI Institute. So if I were to reverse it, I almost would reverse the sentence. Humans are capable of creating godlike technology so that we can improve our medieval institutions and raise above our paleolithic emotions or channel our paleolithic emotions into creativity, productivity, and benevolence.

• 14:04Reid Hoffman?

  What would you say is the key thing in how we build a technology to help us realize our aspirations? Is it a focus on compassion? Is it a question of kind of the human centered and the symbiosis of the interaction? What's the thing that you would build as the next step on this having technology and AI help us realize our better selves?

• 14:33Speaker 00

  I can see why you're a Sims's major, Reid. The combination of philosophy and technology in you. I agree. And, you know, just the previous quote, we almost use paleolithic as a negative word, but it's actually not a negative word. It's a very neutral word. The human emotions are a human self-image of who we are and is deeply rooted in evolution, in our DNAs. And we're not going to change that. And the world is simultaneously beautiful and messy because of that. So thinking about the technology and the future of technologies' relationship with human, I think we need to respect that.

• 15:17Speaker 00

  We need to respect some of the most fundamental, truly paleolithic roots of who we are. A couple of things that technology development really need to respect. And the more we respect it, the better we are. One is the respect of human agency. I really think one of the Public communication issues of AI is we too often use AI as a subject in a sentence as if we're taking away human agency. Or sentences like, AI will cure cancer. I'm sometimes even guilty of saying that.

• 15:59Speaker 00

  The truth is, humans will use AI to cure cancer. It's not AI curing cancer. Or AI will solve fusion. The truth is, human scientists and engineers will use AI as a tool to solve fusion. And what is even more dangerous is AI will take away your job. You know, and, and I think we really need to recognize this technology has so much more opportunity to create opportunities and jobs to empower human agency. And that, that is a very important first principle I, I care about.

• 16:41Speaker 00

  And, and a second important first principle is the respect of humans. Every individual wants to be healthy, wants to be productive, wants to be respected members of society. And no matter how we develop or use AI, we cannot lose sight of that. And losing sight of that is dangerous, is counterproductive. And I think these two things alone are critical in guiding our development of this technology.

• 17:16Reid Hoffman?

  Fei Fei, I love you zooming in on the concept of human agency, which as you know, is at the center of my new book, Superagency. Can you talk a bit more about agency as well as the importance of making AI human-centered? What does human-centered AI mean? And how should technologists and companies be thinking about this?

• 17:37Speaker 00

  Yeah, Reid, I mean, look, you and I started on this journey together before the establishment of a human-centered AI institute at Stanford. Talking about this is, is really rooted in this deep belief that the sole what of any technology, any innovation, is to be human benevolent. And that is the arc of human civilization, is that every time we created a tool, we wanted to use this tool for good. Of course, it's a double edged sword. We could misuse the tool. We would have bad actors to use the tool. So even looking at the dark side of

• 18:22Speaker 00

  technology and tools, it pushes us to even wanna double down on how to make it better, how to, you know, make it human centered. And and that was really truly the the fundamental principle of Human Center AI Institute is that we see AI, you and I and our friends at Stanford, see AI as such a powerful tool. It's a civilizational tool. We better put a framework around it as early as possible that put human and human benefits in the middle of of this. And one of the absolute most critical aspect of human centered AI

• 19:05Speaker 00

  and how I believe it should be guiding every company, every developer, is really this concept of empowering people.

• 19:13Aria Finger?

  You have been working in AI for such a long time in many different capacities. I feel like some people are just getting up to speed on AI now. Like, what do you think about this moment in AI innovation? Both in terms of like where we are, what developers are facing? Like, what do you think we need to do to get to the next level of solving these problems today?

• 19:35Speaker 00

  It is a phenomenal moment. The reason I I think this is absolutely the the inflection of a a revolution is because of the application. The AI now can be used by everyday people and businesses. And many of the dreams that us, the early AI pioneers, have thought of in our early phase of the career have been realized or almost realized. For example, The rhetoric, general public phrase Turing test is pretty much a solved problem. Now Turing test itself, I wouldn't call that the be all and end all test of intelligence.

• 20:19Speaker 00

  But it was a yardstick that was so hard that it was a legitimate yardstick. And it's a solved problem. And cars driving by themselves, right? It's not fully solved, but it's so much more solved than two thousand and six. So so I think because of the power of these models being productionized into the hands of everyday people in business. I think this is a phenomenal uh phase of the AI revolution. But I'm also keenly aware, Ariel, we live in a Silicon Valley bubble because I still think the entire global population still are being brought up to speed of where AI is.

• 21:01Speaker 00

  But we do see the future and where future is going.

• 21:06Aria Finger?

  So I think so many possible listeners can resonate with what you're saying about this could be an enormous human amplification, it could be enormous positive. And like, we do have to worry about sort of the negative consequences. We want to be able to steer it in the right direction. Like, what do you think, like, from a development perspective? What what do we need to do to make sure that AI is going into this positive direction? And if there's, you know, government or cross-sector collaboration that you think is needed, I would I would love to hear about it.

• 21:35Speaker 00

  Honestly, I think there's a lot we can do. And I think we should be doing it yesterday and it's not late. We should just really commit to do this. One thing that I think we should do is base all this on science, not science fiction.

• 21:50Aria Finger?

  Hmm.

• 21:50Speaker 00

  There has been so much hype rhetoric on, you know, the extinction of humanity because of AI or, you know, world peace because of AI. Either side is more science fiction than science. So when we think about how we approach AI policy, AI governance, basing it on data, basing it on scientific facts, basing it on scientific methodology is so important. Now, the second is, I really believe, just like many other technology and tools, putting guardrails around.

• 22:33Speaker 00

  The application where rubber meets the road, where where humans are being impacted is the right place to focus on our governance energy rather than stopping upstream development. Uh huh. Think about the early days of car, it wasn't very safe, you know, there was no seatbelt, there was the beginning didn't even have doors, there was no, you know, speed limit and all that. And then we did have lessons, you know, that cost human lives. But what happened is we didn't go to Ford and G M and say shut down the the factory. We created regulatory framework for seatbelts, for speed limits and all that.

• 23:17Speaker 00

  So today, A I is similar. It it is a profoundly empowering technology, but it comes with its harms. So what we should look at is when A I is applied in medicine, how do we update our F D A regulatory measures? Uh, when A I is applied to, you know, finance, how do we put the regulatory guardrails? So the application is where we should focus governance energy on, that's a second thing. Last but not the least to me is that we need to understand that a positive future of A I comes from a positive ecosystem. And that ecosystem, it needs private sector.

• 24:02Speaker 00

  I think private sector, both in terms of big company as well as private sector in terms of entrepreneurship, uh, is very important. But we also need public sector. Yeah. Because public sector produces public goods. And in my opinion, there are two forms of public goods. The one form of public goods are those curiosity-driven innovations and new knowledge, whether it's using A I for fusion or using A I for curing diseases, using A I for for empowering our teachers, all these different ideas, a lot of them come from public sector. Image that came from pu- public sector, you know? Yeah. Then another form of public goods are people.

• 24:47Speaker 00

  We need to educate more and more youth and the public about this technology and public sector shoulders from K twelve to higher education shoulders.

• 25:00Speaker 00

  The bulk of the societal duty in education. Yeah. So I think these are the different aspect of AI governance and policy that I care a lot about.

• 25:10Reid Hoffman?

  Actually, what you just spoke about, I think one of the things you should also emphasize a bit of the AI for all. Because one of the other things you've been doing is this work to make sure that AI is not just the province of amazing, you know, professors from Stanford with PhDs in physics from Caltech. Right. But everybody else, you know, say a little bit about the AI for all and what the the mission and contribution is there.

• 25:34Speaker 00

  So AI for All is a nonprofit organization I co-founded with my former students and colleagues. And the mission there is really to give opportunity for K-12 students from diverse backgrounds to getting to AI. It's through summer programs in universities and and internships. The idea is to try to get to the public good, you know, education part of AI is we know AI will change the world. But who will change AI? We want more diverse group of people to come and be inspired by using this technology, developing this technology for all kinds of great causes.

• 26:18Speaker 00

  So we we have been focusing on women and students from rural or inner city or or different just historically underrepresented communities and and backgrounds to participate in these summer programs. And it's just so inspiring to see these young people using AI or studying AI and improving ambulance dispatch algorithms to using AI to assess water quality in rural communities. It's still a small effort, but I hope it keeps growing because this very goal of including more and diverse people in AI is a very important one.

• 27:09Reid Hoffman?

  So one of the things that you've also worked on is healthcare. And you know, one of the things that I think people, the areas that people should also track about the elevation of the, of humanity, of the human condition, is what AI can do with healthcare. So say a little bit about that and say what some of the work you've done and some of the work you're, you know, kind of hopeful for, for AI and healthcare, kind of looking forward.

• 27:34Speaker 00

  Yeah, really, as my book also describes, I am passionate about AI's application in healthcare for many reasons. You know, healthcare is absolutely the very core of human centeredness. So healthcare is a very vast industry. It goes from basic biosciences of drug discovery or diagnostics, all the way to the clinical diagnosis, clinical treatment, healthcare delivery and public health, right? So this is, the exciting thing is every point of this system, we're seeing AI could be of great help. My own area I, I love and focus on is healthcare delivery.

• 28:18Speaker 00

  That's where humans are helping humans. We have far fewer nurses in America than our patients need. The job is galling. We have strong nurse attrition. There are some stunning statistics. For example, in one shift, some nurses average more than four miles of walking just by fetching medicines and, and equipments and all that. In one shift, our nurses can do up to 150, 180 different tasks. In the meantime, we have patients that fall from the hospital bed because they are lacking enough care.

• 29:02Speaker 00

  We have a lot of issues triaging very sick patients, but not so sick patients. So healthcare delivery needs a lot of help, let alone our elderly's, uh, living at home. Home Alone, you know, on fall risks or dementia deterioration. So my work in the past ten years plus has been looking at using smart cameras. Cameras that are non-invasive and non uh no contact to uh help our caretakers to pay attention to our patients. If they're in the hospital bed, pay attention to their movement to prevent fall. If they're at home, pay attention to their behaviors

• 29:46Speaker 00

  or loneliness or nutrition intake so that we can track what's going on. If they're in surgery room, pay attention to the instruments that the nurses have to count every single minute so we don't lose them in the body of the patients. This kind of smart camera technology we call ambient intelligence is intended to be helpful to our doctors and healthcare workers so that we can collectively improve the quality of care for our patients.

• 30:20Reid Hoffman?

  Amen. Um. So now, you know, AGI is a term that's bandied about a lot. I think you might have said somewhere, I'm not even sure what AGI means. Because obviously so many people mean it, it's kind of their own Rorschach test. So say a little bit about why, what this AGI discussion is, maybe what it should mean. Right. And what what would make this a little bit more rational versus a set of scattered? You know, it's great, it's terrible, it's it's gonna destroy all jobs, it's gonna help all humanity, etc.

• 30:50Speaker 00

  I know, Reid, I mean, this is this is both a fun but also a frustrating conversation. I I genuinely don't know what AGI means because I think the term came from a commercial side of the world around ten years ago. When AI is coming of age and there is a lot more commercial interest. What the initial term, which I I respect that intention, is to add the word G, general, into AI to really Emphasize that the future of AI is more generalizable capabilities rather than the very narrow one.

• 31:32Speaker 00

  For example, today's self-driving car is much more generalizable than just a camera that detects trees. Right? So so that that narrow focus of one specific task versus a a powerful technology that can do a suite of tasks is real. The reason I feel I have always not been a hundred percent clear is if I go back to history, to the founding fathers of AI, the John McCarthy and Marvin Minsky of AI. If you go to their their dreams and hopes starting in the nineteen fifty six summer,

• 32:11Speaker 00

  that has been their dreams all along is to make machines that can think and help people to make decisions. They eventually can even do things. Nobody in their original dream of AI said we dream of extremely narrow AI tasks that detect trees. The birth of this field as artificial intelligence is thinking machines. So from that point of view, well, we share the same dream. We share the same scientific curiosity. We share the same quest, which is machines that can perform extremely intelligent tasks.

• 32:54Speaker 00

  So from that point of view, I don't know if I should call it AI or AGI. To me, it's the same thing.

• 33:00Aria Finger?

  Well, in thinking about sort of AI that can do things, like you said, it feels like recently with the new improvements in voice and thinking about agentic AI, we're sort of creeping closer where you're like, oh, I'm just having a normal conversation with my AI and we're getting close to it sort of doing things for you. Are there ways that you use agents now in your life that you think are particularly helpful? Or sort of the promise of agentic AI and voice in the next few years you think are going to change things?

• 33:30Speaker 00

  I definitely think the natural language way of sharing knowledge, allowing people to be able to, whether it's search or just ideate or learn something, is a very powerful tool. Even for myself, I use L L Ms to try to understand a concept, to try to understand a paper, to try to just ask a number of questions that I didn't know. What really excites me the most is seeing people and kids using it as a tool to improve their own learning. I do wanna focus that no matter what,

• 34:11Speaker 00

  let's make sure we keep the agency, the self-agency in people and give them good tools to learn and to empower them. And I think as we deepen the powerful tools, and I myself is working on that, we'll see more and more that collaborative capability to allow humans to use these tools to do things more precisely. And I would be excited to see those happen.

• 34:42Aria Finger?

  I think it's not only important because of course it's the right thing to do, but I think also you then get the narrative that, oh, these people developing A I are trying to replace humans and get rid of them. And I don't wanna look at a screen for 10 hours a day. And listen, there's no one who wants to look at a screen for 10 hours a day less than me. I think human interactions are so critical and so important. And they're so important for everything, for teaching, for community, for empathy. And, you know, one story you told in your truly beautiful book, The Worlds I See, was about a math teacher. You had a high school math teacher, Mr. Sibella, which just showed that it is important to have human interaction. So can you say a little bit more about that and perhaps the memorable advice that he gave you?

• 35:25Speaker 00

  The book really acknowledged my early days as an immigrant kid who came to New Jersey at the age of 15 and landed in a public high school not speaking English. And that was the beginning of my journey. And very lucky for me, very soon I met a math teacher called Bob Sibella, who Really treated me with that sense of respect and unconditional support. And not only was my math teacher, but also became my friends in a very difficult teenage years as a new immigrant and throughout my life till his own passing away. But what has he taught me is not through words.

• 36:09Speaker 00

  He never sat me down and say, hey, Fei Fei, AI is gonna take over the world. Let me tell you to be human centered with AI. Like, I don't think that word even, you know, was was in our dictionary. Sure. He just taught me through action is that at the end of the day, the meaning of our society, of our lives are the kind of positive things we do to each other, for each other, and the kind of beliefs we we hold and the beacon of light we chase. And in through his action, I came to appreciate respecting and lifting other humans is a

• 36:50Speaker 00

  beautiful thing, even if that's a clueless kid who who didn't speak English and didn't know what she was doing in a new country. So so I think that kind of generosity, fundamental kindness and compassion is the core of what being human is about. And that, you know, the biggest learning from him for me is put humans in the center. It's beautiful. Thank you.

• 37:17Reid Hoffman?

  Yeah, beautiful.

• 37:24Reid Hoffman?

  So rapid fire. Is there a movie, song, or book that fills you with optimism for the future?

• 37:32Speaker 00

  My Neighbor Totoro is one of my all time favorite movie. Yes, I can hear the music. I'm not gonna attempt to sing because I'm lousy, but yes. It's so simple, so beautiful. I I'm a sucker of beautiful pixels and and yet so so profound. I have the excuse to watch it with kids, but honestly, I don't even care how they feel. I just love watching it.

• 38:06Aria Finger?

  That's awesome. That's awesome. So Fei Fei, what is a question that you wish people would ask you more often?

• 38:13Speaker 00

  I wish people would ask me more about how can I use AI to help people? Because I can talk for hours about that and I can, I can think about so many of my amazing colleagues at Stanford worldwide who are doing that. I can, if I don't know what they're doing, because they're just their own genius um expert in, in, in the discipline, at least I can point people to their work.

• 38:41Aria Finger?

  Absolutely. There's so many people doing amazing things right now and we, we need to inspire more people to do the same.

• 38:47Reid Hoffman?

  Where do you see progress or momentum outside of your industry that inspires you?

• 38:54Speaker 00

  I actually think our humanities focus on energy uh really does inspire me because I think, again, I, I guess I just cannot completely untether myself from AI. Even AI's development is pushing against this very real question of energy, right? Electrical power. And I think that the changes of the environment, the also the democratization of having energy for the global population is so critical. And also, you know, we cannot always rely on fossil fuel.

• 39:33Speaker 00

  So a lot of these advances and, and just the, the global movement in the field of energy is, is uh exciting.

• 39:42Aria Finger?

  So for our final question always, uh can you leave us with a final thought on what you think is possible in the next fifteen years if everything breaks humanities way? And what's the first step to get there?

• 39:54Speaker 00

  I wanna see a global increase of knowledge, well-being and productivity, especially with the emphasis on shared prosperity. The reason I wanna emphasize on that is On the technology side, I am an optimist. I know that technology can help people. I know if we use it right, it can discover new knowledge. It can help us to innovate. It can increase our well-being. But I think it's so important that we are learning that lesson over and over again

• 40:35Speaker 00

  is that when that happens, we need to recognize, we need to share that prosperity. We need to democratize that benefits. Absolutely. I hope so too.

• 40:47Reid Hoffman?

  Faithhe, thank you as always.

• 40:49Speaker 00

  Thank you.

• 40:54Reid Hoffman?

  Possible is produced by Wonder Media Network. It's hosted by Aria Finger and me, Reid Hoffman. Our showrunner is Sean Young. Possible is produced by Katie Sanders, Edie Allard, Sarah Schlied, Vanessa Handy, Aliyah Yates, and Paloma Moreno Jimenez. Jane Kaplan is our executive producer and editor.

• 41:15Aria Finger?

  Special thanks to Jessica Williams, Aki Sheraltz, Harini Shropathy, Russell Wald, and Little Monster Media Company.