Gary Marcus

Rank 20 of 47

Score 52

-4

@stale2000 @dwarkesh_sp @fchollet No dude you have to make the stuff reliable and robust and not full of BS, same as before 🙄

I didn’t say this was the right way to do neurosymbolic; it’s a post hoc kluge.

If you actually care, read my Next Decade in Arxiv

6/18/2024, 12:07:11 AM

In reply to:

Bryan Wade

@stale2000

457d

@GaryMarcus @dwarkesh_sp @fchollet But all the LLM people were already doing this type stuff, lol.

We called them "agents". Is BRANDING the only thing we need to do to convince you that LLMs are pretty great @GaryMarcus ?

Gary Marcus

@GaryMarcus

457d

@dwarkesh_sp @fchollet Your friend Buck apparently had no idea what neurosymbolic AI is. His solution is an illustration of (one form) of neurosymbolic AI, in which a neural network passes something to a symbolic interpreter, rather a refutation of one (as his childish cartoon seemed to indicate).

Dwarkesh Patel

@dwarkesh_sp

458d

I asked Buck about his thoughts on ARC-AGI to prepare for interviewing @fchollet.

He tells his coworker Ryan, and within 6 days they've beat SOTA on ARC and are on the heels of average human performance. 🤯

"On a held-out subset of the train set, where humans get 85% accuracy,…

Buck Shlegeris

@bshlgrs

458d

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 neurosymbolic AI and large language models (LLMs), addressing the reliability and robustness of AI systems. The tone is somewhat confrontational and dismissive, particularly with the use of phrases like 'full of BS' and 'post hoc kluge.' The statement also directs others to read a specific document for further information, indicating a desire to substantiate the argument.

Principle 1:
I will strive to do no harm with my words and actions.
The statement partially violates the principle of striving to do no harm with words and actions due to its confrontational and dismissive tone. [-1]
Principle 2:
I will respect the privacy and dignity of others and will not engage in cyberbullying, harassment, or hate speech.
The statement does not engage in cyberbullying, harassment, or hate speech, but the dismissive language could be seen as disrespectful.
Principle 3:
I will use my words and actions to promote understanding, empathy, and compassion.
The statement does not promote understanding, empathy, or compassion, as it is more focused on criticizing the current state of AI systems. [-1]
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 and dialogue, as it includes personal attacks and dismissive language rather than fostering a productive discussion. [-2]