Gary Marcus

Rank 21 of 47
|
Score 52
6/28/2025, 9:04:03 PM
X
In reply to:
Oliver Habryka
@ohabryka
·
54d
@justanotherlaw @GaryMarcus But o3 had finished training in 2024! The evals were released in 2024. It was available to some people in a research preview. If you have this much wiggle room with your dates you can avoid falsification in any given year for frontier progress.
Lawrence Chan
@justanotherlaw
·
54d
@GaryMarcus @ohabryka I mean, yeah, that's my point later in my thread: you seem to have intended the claim very literally/narrowly, but it’s easy for people to see it as claiming that we won’t see any large advances in 2024. (Also, I’m pretty sure that if you hid the CoT and showed o1 to people in
Gary Marcus
@GaryMarcus
·
54d
@justanotherlaw @ohabryka and because even when i say unambigously no GPT-5 and GPT-5 doesn’t show up people find something to gripe about.
Lawrence Chan
@justanotherlaw
·
54d
@GaryMarcus @ohabryka (In large part because he doesn’t make specific, unambiguously resolvable forecasts, as this attempt to resolve his forecasts is clearly demonstrating.)
Lawrence Chan
@justanotherlaw
·
54d
@GaryMarcus @ohabryka I do think it’s quite unfair to say that Gary is “exceptionally bad”! He’s doing better than the horde of people who confidently asserted that AIs wouldn’t be able do particular things that they did.
Lawrence Chan
@justanotherlaw
·
54d
@GaryMarcus @ohabryka Paul was pretty correct about “prosaic” AI alignment (where you assume there are no fundamental changes from the deep learning paradigm) being the thing to work on back in 2015-6, while most of the AI(S) community was on the “deep learning won’t work” train.

The statement 'it wasn’t available to any independent scientists' is part of a broader discussion about the availability and transparency of AI research and evaluations. This conversation touches on public discourse as it involves the accessibility of scientific research and the implications for independent verification and progress in AI development.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not appear to cause harm with words or actions. It is a factual claim about the availability of research data.
  2. Principle 2:
    I will respect the privacy and dignity of others and will not engage in cyberbullying, harassment, or hate speech.
    The statement respects the privacy and dignity of others, as it does not engage in cyberbullying, harassment, or hate speech.
  3. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement could promote understanding by highlighting issues of transparency in AI research, but it does not explicitly promote empathy or compassion.
  4. 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 is part of a dialogue and does not engage in personal attacks or ad hominem arguments. It contributes to a constructive discussion about the availability of research data. [+1]