The Godmother of AI on jobs, robots & why world models are next | Dr. Fei-Fei Li
The Godmother of AI on jobs, robots & why world models are next | Dr. Fei-Fei Li
The Godmother of AI on jobs, robots & why world models are next | Dr. Fei-Fei Li
Dr. Fei-Fei Li, a pioneering force in artificial intelligence, reflects on the field’s transformative journey and the pivotal role she played in shaping its modern era. From foundational research to real-world applications, her insights reveal how AI evolved from an academic pursuit into a societal catalyst—driven not just by technology, but by vision and responsibility.
Dr. Fei-Fei Li recounts how ImageNet, her groundbreaking dataset of 15 million images, catalyzed the deep learning revolution by enabling neural networks to excel at visual recognition when paired with GPUs. This breakthrough laid the foundation for today’s AI systems. She critiques the notion of artificial general intelligence (AGI) as overhyped, emphasizing that current AI lacks human-like reasoning, creativity, and emotional depth. Instead, she champions 'world models'—AI with spatial intelligence—that can interact with 3D environments, a key step toward embodied cognition. Her company, World Labs, launched Marble, a text-to-3D-world application transforming industries from film production to robotics and psychology. Unlike language models, robotics faces data scarcity due to physical world complexity, requiring synthetic or teleoperated datasets. Li stresses the importance of human-centered AI, exemplified by Stanford’s HAI institute, which promotes ethical development across disciplines. She advocates for inclusive AI participation, urging professionals from all fields to engage in shaping its future. Ultimately, she believes AI should augment human agency, not replace it, calling for mission-driven innovation rooted in collaboration and responsibility.
08:10
08:10
AI is not artificial. It's made by people and impacts people.
13:02
13:02
Human learning and evolution are big-data learning processes.
18:30
18:30
Big data, neural networks, and GPUs are the three pillars of modern AI.
26:54
26:54
Current AI can't achieve what Newton or Einstein did
33:32
33:32
Spatial intelligence is key to advancing AI beyond language models.
42:30
42:30
Robots lack the abundant, perfectly aligned training data that language models benefit from.
48:03
48:03
Marble is the world's first app that can generate 3D worlds through prompts.
51:00
51:00
Marble cuts virtual film production time by 40x with Sony.
1:08:10
1:08:10
Focus on people, mission, and passion, not potential downsides.
1:12:30
1:12:30
HAI involves faculty from all eight Stanford schools.
1:14:24
1:14:24
AI is about people, and every profession has a role to play in its ethical development.
