Gary Marcus

Rank 25 of 47
|
Score 22
6/15/2025, 8:42:43 PM
X
In reply to:
Sevak Avakians
@SevakAvakians
·
63d
@SaiSunkam @GaryMarcus @Will83756176 His point is that LLMs aren't good at chess. He's not talking about regular chess playing algorithms that are able to beat top players.
Sai Sunkam
@SaiSunkam
·
63d
@GaryMarcus @Will83756176 Doesn’t every chess engine just enumerate the possible valid moves and calculate the lines and choose the best one, how do you end up with an invalid move?
Gary Marcus
@GaryMarcus
·
64d
@Will83756176 makes illegal moves, so same story
Will
@Will83756176
·
64d
@GaryMarcus Chess?
Gary Marcus
@GaryMarcus
·
64d
Terence Tao reporting on the same problem we have seen endlessly in other domains: LLM’s produce output that “looks” correct but is often deeply wrong and even stupid on careful inspection.

I know of no domain in which this is NOT the case.

And yet people seem surprised over
vitrupo
@vitrupo
·
64d
Terence Tao says today's AIs pass the eye test -- but fail miserably on the smell test.

They generate proofs that look flawless. But the mistakes are subtle, and strangely inhuman.

“There's a metaphorical mathematical smell.. it's not clear how to get AI to duplicate that.”

The conversation discusses the limitations of large language models (LLMs) in tasks like chess, highlighting their tendency to produce outputs that appear correct but are flawed. This is part of a broader discourse on AI capabilities and limitations.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not cause harm and contributes to a factual discussion about AI limitations. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement promotes understanding by discussing the specific limitations of LLMs in chess. [+1]
  3. Principle 4:
    I will engage in constructive criticism and dialogue with those in disagreement and will not engage in personal attacks or ad hominem arguments.
    The statement engages in constructive dialogue by addressing a specific point about AI capabilities without personal attacks. [+1]