MIT EI seminar, Hyung Won Chung from OpenAI. "Don't teach. Incentivize."

Поділитися
Вставка
  • Опубліковано 9 лис 2024

КОМЕНТАРІ • 36

  • @hengfun
    @hengfun Місяць тому +49

    "teach him taste of fish and make him hungry"🤣

  • @buoyrina9669
    @buoyrina9669 Місяць тому +6

    I enjoyed every bit of this talk.

  • @BuddhaMedam
    @BuddhaMedam 14 днів тому

    Thank you very much for this goldly video sir .Was Definitely worth spending 35 mins on it .

  • @JimSlattery
    @JimSlattery Місяць тому +5

    Thanks for the insightful talk!
    I love the clarity at 18:50 of seeing the LLM going through training, with so many skills implicitly demanded by the next token prediction task.

  • @vrushankdesai715
    @vrushankdesai715 Місяць тому +1

    such an incredibly information dense talk, thank you!

  • @wwkk4964
    @wwkk4964 Місяць тому +3

    One of the best talks on UA-cam right now!

  • @JimSlattery
    @JimSlattery Місяць тому +7

    I like your concept of scaling:
    1) identify the modeling assumption or inductive biases that bottlenecks further scaling
    2) replace it with a more scalable one.
    Example: letting the model learn it's own representations is a more available approach.

  • @KeldonLee
    @KeldonLee Місяць тому +1

    very insightful, thanks.

  • @JohnSeong
    @JohnSeong Місяць тому +2

    좋은 강의 무료로 올려주셔서 감사합니다!

  • @windmaple
    @windmaple Місяць тому +10

    Somehow Hyung Won Chung's talk is always very abstract and purely at a meta level. He doesn't talk about specific LLM techniques or anything like that, but goes all into the fundamental intuition behind the scaling 👍

    • @swyxTV
      @swyxTV Місяць тому +11

      probably cause if he would get into trouble otherwise haha

  • @sughoshkaushik7261
    @sughoshkaushik7261 Місяць тому

    This talk is gold

  • @honon-cs2wl
    @honon-cs2wl Місяць тому +5

    형원님, 정말 멋지십니다. 구독하고 항상 응원할게요!! 앞으로도 많은 영상 부탁드려요

  • @YucheolChoi
    @YucheolChoi Місяць тому +1

    멋지십니다 형원님!!😊

  • @chenjus
    @chenjus Місяць тому +5

    "No amount of bananas can incentivize monkeys to do mathematical reasoning" lol

  • @SONJOGYO
    @SONJOGYO Місяць тому

    좋은 강의 감사합니다. 끊임없는 배움의 해체(unlearning)에 대해 이야기하신 것이 세상을 보는 눈을 트이게 한 느낌이네요. 잘못된 공리를 바탕으로 구축된 직관과 아이디어가 해체되어야 한다는 이야기를 크게 생각해본 적이 없었으니까요.

  • @JimSlattery
    @JimSlattery Місяць тому +3

    Great point: The Bitter Lesson article is the single most important writing in the field of AI. 😳

  • @wangqis
    @wangqis Місяць тому

    谢谢这个分享~

  • @tiendatnguyen7527
    @tiendatnguyen7527 Місяць тому

    thank you for sharing!

  • @andrewthomas9433
    @andrewthomas9433 Місяць тому +1

    referenced talk:
    John Schulman - Reinforcement Learning from Human Feedback: Progress and Challenges
    ua-cam.com/video/hhiLw5Q_UFg/v-deo.html

  • @조바이든-r6r
    @조바이든-r6r Місяць тому

    멋있어요! 일이 많아서 힘드실거라고 생각이 듭니다. 건강도 챙기시길 바래요

  • @-mwolf
    @-mwolf Місяць тому

    yep - we should follow this principle in architecture too!

  • @roro5179
    @roro5179 Місяць тому +2

    my goat

  • @optimaiz
    @optimaiz Місяць тому

    cool, thx u for sharing

  • @hsuai6584
    @hsuai6584 Місяць тому

    It is suprised to know that you majored in mechanical engineering of your phd degree. How can you make such a big move?

  • @ThomasBrown2
    @ThomasBrown2 Місяць тому +1

    ++

  • @madrooky1398
    @madrooky1398 Місяць тому +1

    "MIT EI seminar" reads like "With egg seminar" when you are German, super weird title.... Is it about breakfast? 😂

  • @DistortedV12
    @DistortedV12 Місяць тому

    No Q/A?

  • @deter3
    @deter3 Місяць тому +8

    very shallow perspective .

  • @jashvira9487
    @jashvira9487 Місяць тому +1

    Propagandic