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

    User reports a possible new information theory result after 200-plus hours of agent work

    The user describes three weeks of “fighting with” astra, fable and sol, with 120 pages of work still to review and simplify.

    Alex DimakisAD
    Dimitris PapailiopoulosDP
    will brownWB
    15 Sources, ,

    TLDR

    A user says they think they have a new information theory result after 200-plus hours of agent work and three weeks of “fighting with” astra, fable and sol. Their next task, they say, is to review and simplify 120 pages of work.

    Combined views

    98.9K

    15 Sources, first seen 26d ago

    Combined views

    98.9K

    15 Sources, first seen 26d ago

    828 likes
    26d ago
    first seen 26d ago
    828 likes
    32 comments
    269 saves
    25 reposts
    32 comments
    269 saves
    25 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    15 Sources

    Dimitris Papailiopoulos@DimitrisPapailok after 200+ hours of agent work, many hours of GPU certificates, 3 weeks of fighting with astra, fable & sol, I think I finally have a reasonably cool new info theory result, and have to now review and simplify 120 pages of work (lol). See ya on the other side in a week.26d
    Demetri Spanos@DemetriSpanosA very interesting discussion of experience, both practical and personal, exploring an open information theory problem with AI, at the scale of a single researcher with "only" a few thousand dollars. I remember this problem myself from graduate school.22d
    Alex Dimakis@AlexGDimakisThe capacity of the deletion channel within one hundredth of a bit. This is one of the most famous open problems in information theory, so its good to see some progress by our AI overlords, prompted by Dimitris.22d
    Eric Scrivner@etscrivnerStarting to see how mathematics with AI will likely become more like open source software: Results will become like snippets of code to extend and improve. Exact attribution will become less important. Quote from the article that got me thinking: "Could experts in the area have produced something like this themselves? Quite possibly. But if making it available saves them time and compute or gives them a better initializer for their next result, then that seems as non-trivially valuable."22d
    will brown@willcb@DimitrisPapail @llm_wizard yeah this is also how it's felt for a lot of my research work lately, no real math but certainly leveraging old lessons for agents-juggling-agents-juggling-agents experiments + builds22d
    Dmytro Mishkin@ducha_aikiThat’s a great story of agentic research, read it, it’s worth it.20d

    15 Sources

    Dimitris Papailiopoulos@DimitrisPapailok after 200+ hours of agent work, many hours of GPU certificates, 3 weeks of fighting with astra, fable & sol, I think I finally have a reasonably cool new info theory result, and have to now review and simplify 120 pages of work (lol). See ya on the other side in a week.26d
    Demetri Spanos@DemetriSpanosA very interesting discussion of experience, both practical and personal, exploring an open information theory problem with AI, at the scale of a single researcher with "only" a few thousand dollars. I remember this problem myself from graduate school.22d
    Alex Dimakis@AlexGDimakisThe capacity of the deletion channel within one hundredth of a bit. This is one of the most famous open problems in information theory, so its good to see some progress by our AI overlords, prompted by Dimitris.22d
    Eric Scrivner@etscrivnerStarting to see how mathematics with AI will likely become more like open source software: Results will become like snippets of code to extend and improve. Exact attribution will become less important. Quote from the article that got me thinking: "Could experts in the area have produced something like this themselves? Quite possibly. But if making it available saves them time and compute or gives them a better initializer for their next result, then that seems as non-trivially valuable."22d
    will brown@willcb@DimitrisPapail @llm_wizard yeah this is also how it's felt for a lot of my research work lately, no real math but certainly leveraging old lessons for agents-juggling-agents-juggling-agents experiments + builds22d
    Dmytro Mishkin@ducha_aikiThat’s a great story of agentic research, read it, it’s worth it.20d