Gary Marcus

Rank 14 of 47
|
Score 83
12/22/2024, 3:08:46 AM
X
In reply to:
David Balfour
@hal_incandenza4
·
377d
@GaryMarcus @ZggyPlaydGuitar @fchollet @doomie @GoogleDeepMind I guess the language was ambiguous to a lot of the audience. I think Altman said that directly to the sponsor assuming he knew what he meant, in reaction to his colleague’s choice of words. It wasn’t pre-scripted and I doubt he intended to mislead
Gary Marcus
@GaryMarcus
·
377d
@hal_incandenza4 @ZggyPlaydGuitar @fchollet I believe that @doomie from @GoogleDeepMind (Veo project inter alia) had much the same reaction I did.

I certainly didn’t hear “no targeting” it as “no cheating”; i heard it as no teaching to the test, and 750 training items feels like teaching to the test when i know humans…
David Balfour
@hal_incandenza4
·
377d
@GaryMarcus @ZggyPlaydGuitar @fchollet He meant they didn’t do any of the things not allowed by competition (private datasets, task leakages, task specific hacks etc). Any good faith listener can surmise this. He’s not dumb enough to lie to literal sponsor of the challenge that they didn’t use the public training set
Gary Marcus
@GaryMarcus
·
377d
@ZggyPlaydGuitar @hal_incandenza4 @fchollet i wish it had been clearer in the video. and that sam had not said we didn’t target.

but then people wouldn’t have started whispering about AGI, either.
Zggy
@ZggyPlaydGuitar
·
377d
@GaryMarcus @hal_incandenza4 @fchollet it's in the first sentence
Gary Marcus
@GaryMarcus
·
377d
@hal_incandenza4 As I said before, it’s not about how the test was conducted, it’s about what people took from the video and Sam’s “we didn’t target” remark.

In my informal poll , over half of listeners assumed that o3 had done what @fchollet called the “more impressive” thing of taking the…

The statement engages in a discussion about the clarity and honesty of communication regarding AI testing results. It acknowledges previous concerns about honesty and appreciates the recognition of ambiguous language, suggesting a focus on transparency and accountability.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement aims to clarify misunderstandings, aligning with the principle of doing no harm by promoting transparency. [+1]
  2. Principle 2:
    I will respect the privacy and dignity of others and will not engage in cyberbullying, harassment, or hate speech.
    It respects the privacy and dignity of others by not engaging in personal attacks, though it does question honesty based on past concerns.
  3. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement promotes understanding by acknowledging ambiguity and seeking clarity. [+1]
  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.
    It engages in constructive criticism by addressing the issue of ambiguous language without resorting to personal attacks. [+1]