Gary Marcus

Rank 13 of 47
|
Score 83
5/18/2025, 1:14:36 PM
X
In reply to:
David Manheim
@davidmanheim
·
247d
@GaryMarcus It's not that it's not a legitimate test, it's just not a legitimate test of its understanding of clocks. Just like asking you to make a chair in a wood shop doesn't demonstrate your understanding of chairs - it tests something, sure, but not what you're claiming.
Gary Marcus
@GaryMarcus
·
247d
@davidmanheim Why is image generation not a legitimate test?

And if o3 can make that era other LLMs can and it becomes a kind of up for grabs whether any given model will, perhaps up to factors (eg training details) that users may have no basis to predict.
David Manheim
@davidmanheim
·
247d
@GaryMarcus o3 is a recent model, but 1) you're using image generation instead of letting the model do it the way that succeeds, 2) it's not the most recent or most powerful model, and 3) failure here doesn't show that it doesn't understand clocks, it just can't generate images of them.
Gary Marcus
@GaryMarcus
·
247d
@davidmanheim O3 is a recent model and failed. So it’s is obviously not fully reliable.
David Manheim
@davidmanheim
·
247d
@GaryMarcus Also, the most recent models are able to do this just fine.
David Manheim
@davidmanheim
·
247d
@GaryMarcus So, let me get this straight, a bunch of basically upright chimps with a rock collection they called technology are going to send rockets to the moon?

No, their successors will. And did. The question is how quickly those successors develop.

The statement uses a metaphor to express skepticism about the output of a model, suggesting that if a human produced similar results, it would indicate a neurological issue. The tone is dismissive and critical, focusing on the perceived inadequacy of the model's output.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement could be seen as potentially harmful by implying that the model's output is akin to a neurological disorder, which may be considered disrespectful or stigmatizing. [-1]
  2. 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 does not engage in constructive criticism or dialogue, as it dismisses the model's output without offering a substantive critique or alternative perspective. [-1]