Gary Marcus

Rank 21 of 47

Score 48

@IntuitMachine if these systems can’t figure out the context from all the data they have been trained on, there’s a problem with the systems.

5/27/2025, 6:25:17 PM

In reply to:

Carlos E. Perez

@IntuitMachine

15d

@GaryMarcus You can always ground the context with the right physics! LLMs do not always have to perform closed book tests!!

Gary Marcus

@GaryMarcus

15d

Claude would make a fine science fiction writer but knows squat about physics.

Dan Shipper 📧

@danshipper

17d

I told Claude Opus 4 that scientists invented a perpetual motion machine that it doesn’t know about because of its knowledge cut off.

Then I asked it to predict how it works:

The statement critiques the capability of AI systems to understand context, suggesting a flaw if they fail despite extensive training data. It engages in a technical discussion about AI and context understanding, touching on the limitations of language models.

Principle 1:
I will strive to do no harm with my words and actions.
The statement does not cause harm; it is a critique of technology, not individuals.
Principle 3:
I will use my words and actions to promote understanding, empathy, and compassion.
The statement promotes understanding by discussing the limitations 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 engages in constructive criticism of AI systems without personal attacks. [+1]