korrents

← People and their mental models

Sergey Levine's mental models

2 claims Sergey Levine made fit 2 mental models. Most often: Emergence, Thought experiments. Everything they said here.

Models we see in what they say

Our reading: their claim applies the idea without naming it. The claim is theirs; filing it here is ours.

Emergence

Whole systems do things none of their parts do, and nobody has to design that order for it to appear.

Others who think with this model

Adam SmithCharity MajorsEric RiesNeri OxmanSean Carroll

Open the model page

Emergent abilities come from generalisation turning compositional, not merely from a training set containing a lot of stuff.

  1. Sergey Levine Professor of computer science at UC Berkeley Yeah. So there's a subtlety here. Emerging capabilities don't just come from the fact that internet data has a lot of stuff in it. They also come from the fact that generalization once it reaches a certain level becomes compositional. Fully autonomous robots are much closer than you think – Sergey Levineyoutube.com · 12 Sept 2025 · 0:39:23 into the videoAll korrents from this video
    Their wordsYeah. So there's a subtlety here. Emerging capabilities don't just come from the fact that internet data has a lot of stuff in it. They also come from the fact that generalization once it reaches a certain level becomes compositional.

    Where this was said

    Watch from 0:39:23 plays here

    ↗Fully autonomous robots are much closer than you think – Sergey Levineyoutube.com

    12 Sept 2025 · video · 1h 28m · spoken · machine transcript

Thought experiments

Test an idea in your head first, by imagining the case that would break it.

Others who think with this model

Shane ParrishDax RaadJensen Huang

Open the model page

The goal was never good simulation; it was answering counterfactuals, and a value function does that job as well as a simulator does.

  1. Sergey Levine Professor of computer science at UC Berkeley It tells us that the key is not necessarily to do really good simulation. The key is to figure out how to answer counterfactuals. Fully autonomous robots are much closer than you think – Sergey Levineyoutube.com · 12 Sept 2025 · 1:09:16 into the videoAll korrents from this video
    Their wordsIt tells us that the key is not necessarily to do really good simulation. The key is to figure out how to answer counterfactuals.

    Where this was said

    Watch from 1:09:16 plays here

    ↗Fully autonomous robots are much closer than you think – Sergey Levineyoutube.com

    12 Sept 2025 · video · 1h 28m · spoken · machine transcript