Gary Marcus

Rank 22 of 47
|
Score 17
1/6/2025, 4:44:47 AM
X
In reply to:
Gary Marcus
@GaryMarcus
·
319d
6. The presentation of o3 was problematic, as discussed below, and no clean test of out-of-domain generalization was provided, suggesting that we may see the kinds of problems with distribution shift as ever, outside of semi-closed domains where synthetic data can readily be…
Gary Marcus
@GaryMarcus
·
319d
5. In o1’s best case, there may have been haeavy data augmentation. In domains where that’s not possible, o1 may not even be that much better than GPT-4 (e.g on some language tasks), as you can see if you read OpenAI’s reports carefully: https://openai.com/index/learnin…
Gary Marcus
@GaryMarcus
·
319d
4. o1 probably isn’t as robust as people might have hoped. Even on certain benchmarks, results may be fragile or difficult to replicate:
Alejandro Cuadron
@Alex_Cuadron
·
319d
Surprising find: OpenAI's O1 - reasoning-high only hit 30% on SWE-Bench Verified - far below their 48.9% claim. Even more interesting: Claude achieves 53% in the same framework. Something's off with O1's "enhanced reasoning"... 🧵1/8
Gary Marcus
@GaryMarcus
·
319d
3. Commonsense reasoning is still dodgy, as it has long been, review here, earlier today: https://garymarcus.substack.com/…
Gary Marcus
@GaryMarcus
·
319d
2. A recent paper on Putnam math problems showed a similar failure on distribution shift, showing 30% decement by o1 on math problems with minor variations eg on variable names.

Same old story, different set of problems.

https://openreview.net/forum?id=…
Gary Marcus
@GaryMarcus
·
319d
1. The problem of distribution shift, well-tested in Apple’s 2024 reasoning paper, which goes back to my own 1998 work, continues.

LLMs, like their predecessors generalize well to similar items but struggle with reliability in unfamiliar regimes, even minor variations sometimes…

The statement engages in a technical discussion about the limitations of large language models (LLMs) and the challenges of scaling, referencing specific benchmarks and performance issues. It reflects on the broader implications of AI development and the potential plateau in progress.

  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 constructive discussion about AI development challenges. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    It promotes understanding by discussing technical aspects of AI and acknowledging the complexity of the field. [+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 criticism of AI models and their limitations without resorting to personal attacks. [+1]