Gary Marcus

Rank 13 of 47
|
Score 115
6/18/2024, 11:26:41 PM
X
In reply to:
Tamay Besiroglu
@tamaybes
·
453d
@iamtrask @RyanPGreenblatt @GaryMarcus It is evidence against Chollet's view that the getting the symbolic part right is the more fundamental challenge, and in favor of the view that system 2 reasoning can straightforwardly be added once LLMs are scaled up far enough. In that sense the meme seems apt.
Andrew Trask
@iamtrask
·
454d
Useful stuff. @RyanPGreenblatt is a smart guy....

....but about this #meme.

This meme throws shade at neurosymbolic AI folks (@GaryMarcus / etc.)

It implies that "drawing more samples" proves that LLMs don't need symbolic stuff

... but then the article proceeds to do a…
Buck Shlegeris
@bshlgrs
·
454d
ARC-AGI’s been hyped over the last week as a benchmark that LLMs can’t solve. This claim triggered my dear coworker Ryan Greenblatt so he spent the last week trying to solve it with LLMs. Ryan gets 71% accuracy on a set of examples where humans get 85%; this is SOTA.

The statement engages in a technical discussion about the capabilities and limitations of AI systems, specifically addressing the integration of System 2 reasoning into large language models (LLMs) and the challenges of eliminating hallucinations. The tone is analytical and aims to contribute to the ongoing debate about AI development strategies.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement strives to do no harm by focusing on a technical discussion rather than personal attacks. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement promotes understanding by discussing the complexities of AI development and the limitations of current approaches. [+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 by addressing the limitations of System 2 reasoning in AI without resorting to personal attacks. [+1]