• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
  • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    Researcher Describes 2022 Pivot to Language Models

    Dimitris Papailiopoulos recounts pivoting his group after early exposure to davinci.

    JL
    DP
    KC
    4 Sources, 29d ago, first seen 29d ago

    TLDR

    Dimitris Papailiopoulos, Principal Researcher at Microsoft Research AI Frontiers and professor on leave from UW-Madison, posted that shortly before starting the Looped TFs paper in fall 2022 he grew interested in AGI after using davinci on the OpenAI playground. He told his group to drop ongoing work and study language models instead, which he found remarkable and hard to grasp. The post is presented as an evergreen quote from the researcher.

    Combined views

    64.5K

    4 Sources, first seen 29d ago

    Combined views

    64.5K

    4 Sources, first seen 29d ago

    586 likes
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    586 likes
    8 comments
    451 saves
    62 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    8 comments
    451 saves
    62 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    4 Sources

    @DimitrisPapailEvergreen. Shortly before we started the Looped TFs paper in fall 2022, I was getting AGI pilled by davinci on the openai playground. I then told my group we were dropping whatever we were doing to try and understand language models. They seemed like something out of this world that I could not wrap my head around.. Looped TFs was a first attempt to understand how a transformer can implement general-purpose algorithms of arbitrary depth and be programmed through its prompts to do it. Been a few years since and I'm still not sure I understand transformers and language models (let alone agents), yet I've never been more excited and obsessed. And now, with agents, the object of study has become the instrument used to probe the thing that you want to understand.. So cool :)
    @jasondeanleeLooped transformers https://arxiv.org/abs/2301.13196 @DimitrisPapail
    @KyleCranmerRT @DimitrisPapail: Evergreen. Shortly before we started the Looped TFs paper in fall 2022, I was getting AGI pilled by davinci on the ope…
    @CarinaLHongCongrats to @AngelikiGiannou (Axiom MTS, work done during PhD with @DimitrisPapail)!

    4 Sources

    @DimitrisPapailEvergreen. Shortly before we started the Looped TFs paper in fall 2022, I was getting AGI pilled by davinci on the openai playground. I then told my group we were dropping whatever we were doing to try and understand language models. They seemed like something out of this world that I could not wrap my head around.. Looped TFs was a first attempt to understand how a transformer can implement general-purpose algorithms of arbitrary depth and be programmed through its prompts to do it. Been a few years since and I'm still not sure I understand transformers and language models (let alone agents), yet I've never been more excited and obsessed. And now, with agents, the object of study has become the instrument used to probe the thing that you want to understand.. So cool :)
    @jasondeanleeLooped transformers https://arxiv.org/abs/2301.13196 @DimitrisPapail
    @KyleCranmerRT @DimitrisPapail: Evergreen. Shortly before we started the Looped TFs paper in fall 2022, I was getting AGI pilled by davinci on the ope…
    @CarinaLHongCongrats to @AngelikiGiannou (Axiom MTS, work done during PhD with @DimitrisPapail)!