Gary Marcus

Rank 11 of 47
|
Score 83
9/4/2024, 3:26:32 PM
X
In reply to:
Daniel Litt
@littmath
·
513d
@GaryMarcus @ChrSzegedy @colin_fraser @Lang__Leon OK, that seems way out of reach to me. E.g. formalizing this paper of mine would require formalizing a substantial amount of non-abelian Hodge theory first, which would be a monumental task: https://annals.math.princeton.ed…
Gary Marcus
@GaryMarcus
·
513d
@littmath @ChrSzegedy @colin_fraser @Lang__Leon “given an[y] article 95% of the time” was the intent, though the wording could have been slightly better. i am sure one could find one paper post hoc that some system could succeed on. that’s not the interest here.
Daniel Litt
@littmath
·
513d
@GaryMarcus @ChrSzegedy @colin_fraser @Lang__Leon Actually sorry, I am not sure I understand the bet. Is the bet that there will exist *one* paper in the Annals where an AI can do this? Or that it can do it for 95% of papers? If the latter, no way.
Daniel Litt
@littmath
·
513d
@GaryMarcus @ChrSzegedy @colin_fraser @Lang__Leon Ah gotcha. I’d be very impressed if this happens by 2029.
Gary Marcus
@GaryMarcus
·
513d
@littmath @ChrSzegedy @colin_fraser @Lang__Leon Here’s the bet:
Gary Marcus
@GaryMarcus
·
833d
@ChrSzegedy @MCHammer @Grady_Booch Great - theorem proving bet back on! Here are some proposed rules from @ErnestSDavis. He offers $200, which I am doubling, for a total $400 on our side, with @MCHammer and @Grady_Booch as our witnesses :)
Daniel Litt
@littmath
·
513d
@GaryMarcus @ChrSzegedy @colin_fraser @Lang__Leon I find what @ChrSzegedy is saying here reasonably plausible (though I don't know about your bet) -- LLMs alone clearly can't do math well but I think there's now some evidence that with formalization they can be valuable.

The statement and the preceding conversation constitute public discourse as they engage in a substantive discussion about the capabilities of AI in theorem proving, a topic of public interest in the fields of mathematics and artificial intelligence. The conversation involves multiple participants and references a specific bet about the future capabilities of AI, which has implications for the broader discourse on AI's role in academic and scientific research.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement and the preceding conversation strive to do no harm. They are focused on a technical and academic discussion about AI capabilities, without any personal attacks or harmful language. [+2]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The conversation promotes understanding and empathy by engaging multiple experts in a constructive dialogue about the challenges and possibilities of AI in theorem proving. It encourages the sharing of knowledge and perspectives. [+2]
  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 participants engage in constructive criticism and dialogue, discussing the feasibility of AI achieving certain tasks. The conversation is respectful and focused on the technical aspects of the debate, without resorting to personal attacks or ad hominem arguments. [+2]