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

    A definition of 'taste' draws a heated reaction

    David Pfau objected to the definition, while Dylan Hadfield-Menell called the evaluation metric optimization.

    David PfauDP
    Andreas Kirsch 🇺🇦AK
    Dylan HadfieldMenellDH
    5 Sources, ,

    TLDR

    David Pfau reacted angrily to a linked definition of “taste.” In a separate response to a claim that AI research taste is doubling every three months, Dylan Hadfield-Menell argued the evaluation measured metric optimization instead. Neil Chowdhury called the related plot the most important since METR’s time-horizon chart.

    Combined views

    4.8K

    5 Sources, first seen 2h ago

    Combined views

    4.8K

    5 Sources, first seen 2h ago

    159 likes
    2h ago
    first seen 2h ago
    159 likes
    9 comments
    19 saves
    7 reposts
    9 comments
    19 saves
    7 reposts

    David Pfau reacted sharply on X to a linked definition of “taste,” calling it almost calculated to send him into “an apoplectic rage.”

    Featured Source

    A separate post focused on a claim that AI research taste is doubling every three months. In a meme-format response, Dylan Hadfield-Menell contrasted research taste with “hill-climbing a metric” and argued that the linked evaluation measured metric optimization instead.

    Neil Chowdhury linked to the same post as Hadfield-Menell and called its plot the “most important” one since METR’s time-horizon chart. Chowdhury did not elaborate in his post on what he believed the plot demonstrated.

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    #4

    Today's Rank

    #4

    David Pfau

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    5 Sources

    David Pfau@pfauThis definition of "taste" seems almost calculated to drive me into an apoplectic rage.2h
    Dylan HadfieldMenell@dhadfieldmenell>eval says research taste doubles every 3 months >ask if it measures research taste or metric optimization >they don't understand >pull out diagram explaining taste vs hill-climbing a metric >they laugh and say "it's a good eval sir" >read the paper >it's metric optimization1h
    Neil Chowdhury@ChowdhuryNeilMost important plot since METR time horizon:1h
    Andreas Kirsch 🇺🇦@BlackHC@pfau They have great taste for making definitions that are uniquely suited to trigger researchers 😂39m

    5 Sources

    David Pfau@pfauThis definition of "taste" seems almost calculated to drive me into an apoplectic rage.2h
    Dylan HadfieldMenell@dhadfieldmenell>eval says research taste doubles every 3 months >ask if it measures research taste or metric optimization >they don't understand >pull out diagram explaining taste vs hill-climbing a metric >they laugh and say "it's a good eval sir" >read the paper >it's metric optimization1h
    Neil Chowdhury@ChowdhuryNeilMost important plot since METR time horizon:1h
    Andreas Kirsch 🇺🇦@BlackHC@pfau They have great taste for making definitions that are uniquely suited to trigger researchers 😂39m