korrents

world models

Internal representations of an environment that an agent builds and revises, and whether machines can build them fast.

FilterEveryone, all time17 from 11 people
Group All Technologists 3Scientists 1Economists & investors 2Artists & entertainers 1Writers & journalists 2
Who Everyone FCFrançois Chollet 6 SLSergey Levine 2 CRCharan Ranganath 1 DHDemis Hassabis 1 FLFei-Fei Li 1 KKKevin Kelly 1 LWLilian Weng 1 NBNat Bullard 1 PBPeter Bevelin 1 RDRay Dalio 1 THTim Harford 1
When All time Past 1 yearPast 3 yearsPast 5 years
  1. TH

    Tim Harford quoted

    Economist and journalist

    Our readingPeople's behavior is generally predictable rather than random.

    Their wordsNo supercomputing was required: Shannon's machine had just 16 bits of memory. We humans are not hard to read.

    ↗Random is as random doestimharford.com 1st of 3 in this piece

  2. 3 weeks earlier
  3. FL

    Fei-Fei Li quoted

    Sequoia Professor of Computer Science at Stanford

    Our readingThe scaling law for world models was a conviction held in advance, not a discovery; the devils were always in the architecture choices and the data mixture.

    scaling laws

    Their wordsI think three of us have total conviction about the scaling law. That that I think we do. I do think the exact architecture choices and data mixtures is where the the devils are in the details.

    ↗Why Fei-Fei Li Is Betting on Spatial Intelligenceyoutube.com 7th of 13 in this recording

  4. 4 weeks earlier
  5. KK

    Kevin Kelly quoted

    Founding executive editor of Wired and author of What…

    Our readingThe companies that already control the largest troves of real-world video data will be first to build successful world models.

    startups

    Their wordsMany startups are racing to create foundational world models, but I think the first ones to succeed will likely be the platforms in control of this data bank.

    ↗Worldbuilding with Spatial Intelligencekk.org 1st of 3 in this piece

  7. FC

    François Chollet quoted

    Creator of the Keras deep-learning library and the ARC-AGI benchmark

    Our readingCoding is not just another application domain for AI; it is the meta-skill that lets AI generate its own training material and start the recursive self-improvement loop.

    recursive self-improvement

    Their wordsCoding isn't yet another application domain -- it's the meta-skill required for AI to automatically develop its own training material, via symbolic world models. That's how the RSI loop actually kicks off.

    ↗@fchollet on Xx.com

  8. 11 months earlier
  9. SL

    Sergey Levine quoted

    Professor of computer science at UC Berkeley

    Our readingSynthetic experience can only rehearse what a system already knows; information about the world has to be injected from outside.

    Their wordsSo here's what I would say that deep down at a very fundamental level the synthetic experience that you create yourself doesn't allow you to learn more about the world. It allows you to rehearse things. It allows you to consider counterfactuals but somehow information about the world needs to get injected into the system.

    ↗Fully autonomous robots are much closer than you think – Sergey Levineyoutube.com 20th of 24 in this recording

  11. SL

    Sergey Levine quoted

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

    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.

    ↗Fully autonomous robots are much closer than you think – Sergey Levineyoutube.com 21st of 24 in this recording

  12. 7 weeks earlier
  13. DH

    Demis Hassabis quoted

    Co-founder and CEO of Google DeepMind

    Our readingA model that can predict the next frames of a video coherently understands the world, in the only sense of the word that is doing any work.

    Their wordsI think to the extent that it can predict the next frames in a coherent way, that is a form of understanding, not in the anthropomorphic version of, it's not some kind of deep philosophical understanding of what's going on, I don't think these systems have that, but they certainly have modeled enough of the dynamics, put it that way

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 4th of 20 in this recording

  14. 6 days earlier
  15. FC

    François Chollet quoted

    Our readingHuman adaptability is fundamentally linked to the efficient construction and refinement of internal world models

    Their wordsWe argue that this profound adaptability is fundamentally linked to the efficient construction and refinement of internal representations of the environment, commonly referred to as world models, and we refer to this adaptation mechanism as world model induction.

    ↗Assessing Adaptive World Models in Machines with Novel Games (with 13 co-authors)arxiv.org All 5 from this paper

  17. FC

    François Chollet quoted

    Our readingCurrent AI evaluation of world models is too narrow, favoring static representations learned from massive datasets over efficient learning through interaction

    Their wordsHowever, current understanding and evaluation of world models in artificial intelligence (AI) remains narrow, often focusing on static representations learned from training on massive corpora of data, instead of the efficiency and efficacy in learning these representations through interaction and exploration within a novel environment.

    ↗Assessing Adaptive World Models in Machines with Novel Games (with 13 co-authors)arxiv.org All 5 from this paper

  19. FC

    François Chollet quoted

    Our readingRapid world model induction will be critical for achieving robust, general AI that functions in complex, fast-changing real and human worlds

    Their wordsWe expect that building and evaluating AI systems capable of this kind of rapid world model induction will be critical for achieving robust, general AI capable of functioning effectively in the complex and fast-changing real world, and especially in human worlds - the environments that human beings have evolved in, created, and are continually changing and re-creating.

    ↗Assessing Adaptive World Models in Machines with Novel Games (with 13 co-authors)arxiv.org All 5 from this paper

  21. FC

    François Chollet quoted

    Our readingAn adaptive agent's world model cannot be a static representation learned once and fixed

    Their wordsFor adaptation to complex and changing environments, an agent's world models cannot be static representations learned once and fixed.

    ↗Assessing Adaptive World Models in Machines with Novel Games (with 13 co-authors)arxiv.org All 5 from this paper

  23. FC

    François Chollet quoted

    Our readingGames are particularly rich and controlled environments uniquely well-suited for evaluating rapid world model induction in AI

    Their wordsWe contend that games provide particularly rich and controlled environments uniquely well-suited for systematically evaluating rapid model adaptation and the process of world model induction.

    ↗Assessing Adaptive World Models in Machines with Novel Games (with 13 co-authors)arxiv.org All 5 from this paper

  24. 4 weeks earlier
  25. NB

    Nat Bullard quoted

    Energy and decarbonisation analyst

    Our readingHow you measure the present strongly shapes what can be extrapolated about the near future.

    Their wordsExtrapolating from the present day into the near future is highly dependent upon how we measure the present day.

    ↗Growth Measurementsnathanielbullard.com 1st of 3 in this piece

  26. 13 months earlier
  27. CR

    Charan Ranganath quoted

    Professor of psychology and neuroscience at the University of California

    Our readingThe brain does not record experience continuously; it lays down an episodic memory at moments of surprise, uncertainty and prediction error, because that is where the useful information is.

    Their wordsis this idea that we form these internal models at particular points of high prediction error or points of, I believe also points of uncertainty, points of surprise or motivationally significant periods. And those points are when it’s maximally optimal to encode an episodic memory.

    ↗Charan Ranganath: Human Memory, Imagination, Deja Vu, and False Memories | Lex Fridman Podcast #430youtube.com 3rd of 20 in this recording

  28. 6 weeks earlier
  29. LW

    Lilian Weng quoted

    Machine-learning researcher

    Our readingVideo generation requires more world knowledge than image generation because of the need for temporal consistency across frames.

    Their wordsIt has extra requirements on temporal consistency across frames in time, which naturally demands more world knowledge to be encoded into the model.

    ↗Diffusion Models for Video Generationlilianweng.github.io 1st of 2 in this piece

  30. 18 months earlier
  31. RD

    Ray Dalio quoted

    Founder of Bridgewater Associates

    Our readingSurprise usually comes from something that has not happened in one's own lifetime but did happen before, so it is not seen or thought about.

    Their wordswhenever I was surprised and I think whenever most people it's surprised it's because of something that never happened in their lifetime before but it happened before so it doesn't happen in their context but it happens so you don't see it so you don't think about it

    ↗Ray Dalio: Principles for a Life Well Lived | Knowledge Project Podcast Archiveyoutube.com 5th of 12 in this recording

  32. 15 years earlier
  33. PB

    Peter Bevelin quoted

    Swedish investor and author of Seeking Wisdom: From Darwin to Munger

    Our readingLearning starts with observing reality and learning the core concepts that account for it

    Their wordsLook around you – observe reality. What can explain this? Learn some core concepts that account for reality.

    ↗Interview with Peter Bevelin, author of Seeking Wisdom (Value Investing World, 2007)valueinvestingworld.com 2nd of 7 in this piece