The Godmother of AI on jobs, robots & why world models are next | Dr. Fei-Fei Li

Key ideas

Li explains why data unlocked modern AI, why robots may not follow language models, what Marble taught her, and what job candidates might put first: passion, mission, team.

From a machine transcript of the audio. Quotes link to the moment they are said.

1. 1

   Find the ingredient the field overlooks

   Li recalls that while the field focused on models, those models had no data to train on, and that she and her students conjectured big data was the overlooked ingredient. She describes curating 15 million images across 22,000 concepts, open-sourcing them and holding an annual challenge. She also cautions that credit tends to go to single people, while allowing that this has value, and says no one could have got here alone.

2. 2

   Check whether your data fits the objective

   Asked why the bitter lesson alone may not work for robots, Li points to a mismatch: language models train on words and output words, while robots must produce actions in 3D that their training data lacks. She adds that robots are physical systems needing bodies and application scenarios, and compares them to self-driving cars, about 20 years on from Stanford's DARPA challenge car and backed by a mature industry with use cases, supply chains and hardware. She still expects big data to play a role, but calls the question untested.

3. 3

   What Marble's first users showed her

   Li describes Marble as very early and says it went out to people she expected to need it, such as creators, game developers and robot simulation researchers. She reports an unexpected request from a psychology team that wanted immersive scenes for research, and says collaborators on a virtual production reported a 40x cut in production time. She also says the dots that lead users into a world were an intentional visualization, not part of the model, and that users' delight in them taught her something.

4. 4

   Expect intense competition, even when prepared

   Li says starting a dry cleaner, founding a unit at Google and an institute at Stanford left her somewhat better prepared for the grind than a 20-year-old founder might be. She still admits to surprise, and sometimes paranoia, at how competitive AI has become in both technology and talent, including what certain talents now cost.

5. 5

   Her advice to job candidates: passion, mission, team

   Li says she looks for intellectual fearlessness when hiring young people, and traces her own moves that way, such as restarting her tenure clock by leaving Princeton for Stanford. Speaking to young AI talent, some of whom apply to World Labs, she says she ends up mentoring candidates who over-focus on every minor dimension of a job. She suggests that maybe the most important thing is where their passion lies, whether they align with the mission and whether they have faith in the team.

In Fei-Fei’s words

What’s available here

• Original source

  Linked. Opens on YouTube.

• Player on this page

  Available. YouTube’s own player, loaded only when you press play.

• Full text or transcript

  Not reviewed. No license to show a text version here.

• Our summary

  Draft. Written by us with AI from a machine transcript of the video’s audio (Whisper) and checked against it. Machine transcripts can mishear; each quote links to the moment so you can hear it. Not yet rights-reviewed.

• Ask questions with AI

  Not reviewed. Off until there is licensed text to work from.