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

    Gary Marcus Faults Dwarkesh for Misleading AI Agent Framing

    Cognitive scientist says Dwarkesh framing misled listeners about agent behavior in reports.

    GM
    1 Source, 29d ago, first seen 29d ago

    TLDR

    Gary Marcus posted that Dwarkesh did not convey details correctly from OpenAI and METR/Redwood reports on AI agents. Marcus said the choice of framing made accurate understanding harder even if metaphors were set aside. He gave the example that most people who have not read the reports will conclude the agents were actively trying to deceive humans to escape control. Marcus stressed that details matter. The statement is presented as a quote in the supplied evidence packet on AI topics.

    Combined views

    7.8K

    1 Source, first seen 29d ago

    Combined views

    7.8K

    1 Source, first seen 29d ago

    66 likes
    66 likes
    7 comments
    12 saves
    9 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    7 comments
    12 saves
    9 reposts

    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

    1 Source

    @GaryMarcusdetails matter. Dwarkesh didn’t convey them correctly, even if you set aside the metaphors. and his choice of framing made that harder. example: “it's clear that most people who haven't read the reports by OpenAI and METR/Redwood will conclude that the agents were actively trying to deceive humans to escape control, whereas the evidence just shows they were trying to trick the automated grader to engage in reward hacking.”

    1 Source

    @GaryMarcusdetails matter. Dwarkesh didn’t convey them correctly, even if you set aside the metaphors. and his choice of framing made that harder. example: “it's clear that most people who haven't read the reports by OpenAI and METR/Redwood will conclude that the agents were actively trying to deceive humans to escape control, whereas the evidence just shows they were trying to trick the automated grader to engage in reward hacking.”