
Aug 10, 2026 · 2h 9m
Fei-Fei Li explains how spatial intelligence will augment human capability
Using AI to Increase Your Intelligence & Enrich Humanity | Dr. Fei-Fei Li
Understanding the intersection of computer vision and human cognition is essential for building ethical AI tools that enhance, rather than replace, human intelligence and agency.
- 1Vision acted as an evolutionary catalyst for biological intelligence and remains the cornerstone for advanced machine learning.
- 2True machine creativity and intuition require understanding physical laws and spatial dimensions rather than just static internet data.
- 3Human-centered AI design ensures technology acts as a collaborative tool to enhance human agency instead of replacing it.
Don't miss
Fei-Fei Li explains how training AI on video data allows it to learn physical laws and predict future frames, unlocking spatial intelligence.
The brief
Stanford computer science professor Fei-Fei Li joins neuroscientist Andrew Huberman to explore the deep connection between biological vision and the evolution of intelligence, tracing the roots of active sensing back 540 million years to the Cambrian explosion.
They trace how computer vision reached a major inflection point in 2012. By combining massive datasets like ImageNet with neural networks and GPU computing, researchers trained algorithms to recognize objects, eventually surpassing human performance.
The conversation shifts to the limits of machine learning. While AI can simulate intuitive decision-making using reinforcement learning, the guests debate whether systems trained purely on internet data can ever capture the true depth of human cognition.
Li highlights the rise of spatial intelligence and her startup, World Labs, which aims to help AI understand physical laws and three-dimensional space. This technology is designed to empower human creators and supercharge artistic industries.
Ultimately, Li argues for a human-centered approach to technology. Through Stanford's Human-Centered AI Institute, she advocates for building tools that augment human capabilities in medicine and education rather than replacing human agency.
What was said on this episode
65 statements · 41 positive · 17 negative · 2 mixed · 5 neutral
Humanity has generally progressed toward better conditions over history
“by and large, we advance for the better”
Listen at 0:17
Humanity fundamentally advances toward improvement despite atrocities and setbacks.
“Fundamentally, I'm an army optimist in humanity.”
Listen at 0:26
Society and Silicon Valley are failing to adequately support teachers and parents.
“What I worry about are teachers and some parents, because I think our society today, and especially Silicon Valley are not doing them a service.”
Listen at 0:39
Visual function involves approximately half of human cortical activity
“It is estimated half of the cortical activities in human brain is involved in visual function.”
Listen at 6:43
By age six, humans can learn tens of thousands of object categories.
“Humans by age 6 can learn tens of thousands of different object categories.”
Listen at 10:32
Limited training data was a major cause of slow AI progress.
“So we conjectured that the lack of data was a huge part of the reason, that's the lack of progress in AI.”
Listen at 10:52
Insufficient training data substantially contributed to slow AI progress
“we conjectured that the lack of data was a huge part of the reason, that's the lack of progress in AI.”
Listen at 10:53
ImageNet contains 15 million images.
“Because Imagenet is a collection of 15 million images.”
Listen at 11:31
ImageNet contains 15 million images
“Imagenet is a collection of 15 million images.”
Listen at 11:32
GPU computing, mature neural networks, and big data triggered the modern AI revolution.
“These convergence of these three things really set off the revolution of AI.”
Listen at 14:49
Image-recognition algorithms surpassed humans on 1,000-object naming by 2016
“it still took another three years, I remember by 2012, 2016 for the algorithm to beat humans in naming 1000 objects.”
Listen at 17:24
Transformers are more powerful than early ImageNet-era neural-network algorithms.
“We have an even more powerful neural network algorithm called transformer.”
Listen at 22:53
OpenAI and Google rapidly advanced transformer-based AI technology
“companies like OpenAI and Google quickly rallied beyond this very important technology.”
Listen at 23:10
Massive training datasets enabled current AI systems to recognize contextual patterns reliably.
“This current era when the huge data that these algorithms have learned, let's take Gemini or GPT have learned, really created the capability in the machine's learned space.”
Listen at 26:51
Children identify cats from far fewer examples than current AI systems use
“They likely have seen three cats, 10 cats at most, but yet they're able to identify that tail as a cat tail instead of a foxtail”
Listen at 28:08
Adding video training data around 2023 enabled major advances in AI video generation.
“around 2023, very shortly after ChatGPT moment, multiple research teams start to put video into the training data.”
Listen at 30:36
Video-generation algorithms do not model a cat's muscle structure explicitly
“I would say the algorithm doesn't.”
Listen at 32:45
Current video-generation algorithms do not understand a cat’s muscle structure.
“the algorithm doesn't.”
Listen at 32:45
AI generates plausible cat motion from statistical patterns in large datasets
“It's the statistics, it's the large amount of data that showed you what is the plausible generation of cat movements.”
Listen at 33:25
Current AI cannot access uncaptured, highly nuanced human cognition
“today's AI doesn't have a way to do that.”
Listen at 38:01
The Internet is the largest multimodal collection of recorded human behavior.
“Internet is the biggest collection of human behavior in multimodal forms.”
Listen at 38:17
Unrecorded, highly personalized thoughts remain unavailable to Internet-trained AI.
“That thought is not captured, therefore it's not on the Internet, therefore AI has not seen it.”
Listen at 41:21
AI can combine learned information in highly creative ways
“AI has learned so many things. It can combine information in highly creative way.”
Listen at 41:39
AlphaGo’s Move 37 represents a form of AI creativity.
“Move 37 has symbolized AI's creativity.”
Listen at 41:50
Human-AI collaboration may solve problems requiring unprecedented mathematical creativity
“My current conjecture is hybrid, is that humans working alongside AI would help us to solve these problems whose solutions have yet to be invented.”
Listen at 44:32
Current AI cannot access highly personal memories underlying individual reactions
“no matter how mighty AI is today cannot access that.”
Listen at 45:23
Current AI cannot access deeply personal memories that were never digitally recorded.
“That's where AI cannot access.”
Listen at 45:46
AI can improve communication by learning a user's writing patterns
“AI learning your patterns of writing can already help you to be a better communicator, a more effective communicator, a more efficient communicator.”
Listen at 48:38
AI should augment rather than remove human agency
“we need to think about AI as a tool that helps us in our agency. It should not take away our agency.”
Listen at 49:20
AI should support human agency rather than remove it.
“It should not take away our agency.”
Listen at 49:27
People should retain the choice to use or avoid emerging technologies.
“But they should have the choice.”
Listen at 50:28
People should learn to use AI to empower learning, work, and expression
“learn about how you can use it yourself to empower yourself, your learning or your work or your expression.”
Listen at 52:29
AI can retain vast information and synthesize knowledge across disciplines
“we have a new tool whose brain that can retain humongous amount of information, can help us synthesize knowledge, can go across disciplines in ways that you and I cannot go.”
Listen at 1:01:27
Global liver-surgery data may be insufficient to train a fully autonomous surgical algorithm.
“even if you aggregate the world's liver patient surgeries, you might not have enough data to train this algorithm.”
Listen at 1:05:51
Human-robot collaboration is safer than autonomous surgery with insufficient training data
“having a human collaborating with the robot is way better than a underlearned robot doing the surgery by itself.”
Listen at 1:06:18
AI may help people without immediate doctors when common diagnostic patterns are well represented
“AI has learned that. So we can then now take advantage of that for people who don't have immediate access to doctors.”
Listen at 1:07:13
Her father's robotic laparoscopic liver surgery reduced blood loss tenfold
“he lost 10x less blood than a typical surgery thanks to the laparoscopic capability of a robot surgery.”
Listen at 1:07:26
No current technology can access completely unexpressed individualized intuition
“There's no technology that can do that.”
Listen at 1:13:12
Current machines do not possess human cognitive states such as motivation or fear
“And do machines have it today? No, let's make it very clear.”
Listen at 1:17:51
Current machines do not possess human cognitive states such as motivation, fear, or love.
“Do machines have it today? No, let's make it very clear.”
Listen at 1:17:51
AI technology for generating human-like visual communication already exists
“The technology is There.”
Listen at 1:21:34
AI development should adopt professional norms and ethics education
“AI has to go through the same, right? We need to have our professional norms, we need to have education.”
Listen at 1:26:02
AI governance problems should be solved collaboratively by multiple stakeholders
“these are multi stakeholder problems to solve together.”
Listen at 1:27:24
AI tools could harm young people by removing learning agency and motivation
“The absolute bad outcome is that our young generation, their agency and human level motivation of learning and living is taken away by tools.”
Listen at 1:29:48
Poorly used AI can undermine young people’s agency and motivation to learn.
“If AI is not used right, the agency and motivation is taken away.”
Listen at 1:31:09
Guided AI use could enable students to learn more deeply than before
“with the proper agency, proper motivation, proper ways of using this tool, we can go a lot deeper with AI than we have ever learned.”
Listen at 1:31:38
Properly supported AI use could make future generations substantially more capable
“Then this generation, this coming generation and many generations to come will be way smarter than us because they are superpowered.”
Listen at 1:33:00
K–12 schools should teach students how to prompt AI systems
“I would love to see our schools K12 teaching prompting.”
Listen at 1:34:32
The next AI frontier is embodied, nonlinguistic intelligence beyond language.
“The next frontier of AI as I have been saying is beyond language because again humans develop first pre verbal evolution took you know, 500 million years without verbal communication.”
Listen at 1:36:52
Robots could improve society when deployed in a human-helping manner
“the world would in the right version would be a lot better place with robots helping humans.”
Listen at 1:37:11
Robots should eventually assist people in everyday society
“I would love to see robots being part of our society helping us.”
Listen at 1:38:26
Healthcare systems face a shortage of caretakers
“we don't have an excess of caretakers. We have a shortage of caretakers.”
Listen at 1:40:05
Humanity should collectively decide how future robotics develops
“Humanity should have the agency to decide how we imagine this.”
Listen at 1:43:11
Many AI technologists are motivated by empowering people rather than purely advancing technology.
“I don't feel despair, Andrew. I'm an educator, I'm a builder, I'm a technologist. I see many people around me, including my entire startup. These brilliant young technologists could join any startup or company they want, but they come to world Labs because they want to empower people.”
Listen at 1:47:48
Public AI discourse should move between extreme pessimism and utopianism toward pragmatic collective guidance.
“I think we should come to the middle and talk about what this technology is, how to use it, how we can collectively have that agency to guide the future.”
Listen at 1:48:47
Spatial and physical intelligence are the next major frontier of AI.
“unlocking spatial and physical intelligence is really the next chapter.”
Listen at 1:50:55
World Labs aims to generate 3D and 4D environments for creative, robotic, and architectural applications.
“generating 3D, 4D world that are deeply useful for creators, for robot training, for architecture design, or to enable those interactive environments.”
Listen at 1:51:21
AI can increasingly generate high-quality video shots from scripts
“technology has advanced enough that taking scripts and generating shots, video shots, is getting really good.”
Listen at 1:55:16
Human emotion and perspective remain central to storytelling despite AI-generated production.
“every part of storytelling and story creation, there are humans behind it with their unique emotion, story technique, how they see the world, how they move the cameras, how they characterize characters.”
Listen at 1:55:48
AI should enhance creators' jobs and creative capabilities
“Technology should be making their jobs better, superpowering their creativity.”
Listen at 1:58:47
Workers can reskill and upskill during technological industry changes
“People also get reskilled, upskilled.”
Listen at 1:59:27
Technological change creates both opportunities and losses for workers.
“moments of change is moment of both opportunity and loss.”
Listen at 1:59:41
Teachers and parents should receive support and resources for AI adoption
“They are the most important people in our society. We should be talking to them, we should be uplifting them, we should be supporting them, we should be providing resources to them.”
Listen at 2:02:25
Society should educate, support, and resource teachers and parents navigating AI.
“We should be talking to them, we should be uplifting them, we should be supporting them, we should be providing resources to them.”
Listen at 2:02:28
Supporting teachers and parents is necessary to help children adapt to AI
“we should all remember our teachers and help our teachers and parents so that we can help our kids.”
Listen at 2:04:01
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Books & mentions
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.
