Gary Marcus

Rank 13 of 47

Score 84

Update 2: Looking more carefully at this paper, I think the gist of what it says is correct -- LLMS are prone to various inconsistencies -- but the actual experimental materials here aren't compelling. Having seen examples like this before in my own informal experiments

6/27/2025, 8:49:27 PM

In reply to:

Gary Marcus

@GaryMarcus

229d

Update: o3 appears to me (baed on informal testing) to be less likely to succumb to simple “potemkins” but not immune.

Here (left) it affirms what a haiku is, but tells me (wrongly) that “asphalt shimmers” fits the requirement that the last line of haiku should be five

Gary Marcus

@GaryMarcus

229d

BREAKING: Explosive new paper from MIT/Harvard/UChicago.

Things just got worse — a lot worse — for LLM’s and the myth that they can understand and reason.

The paper documents a pattern they called Potemkins, a kind of reasoning inconsistency (see figure below). They show that

The statement engages in a critical analysis of a paper discussing the limitations of large language models (LLMs), specifically their inconsistencies. The tone is analytical and somewhat skeptical, as the speaker acknowledges the paper's main point but questions the strength of its experimental evidence. This reflects a constructive approach to academic critique, focusing on the validity of the research methods rather than dismissing the findings outright.

Principle 1:
I will strive to do no harm with my words and actions.
The statement does no harm and maintains a respectful tone while critiquing the paper. [+1]
Principle 3:
I will use my words and actions to promote understanding, empathy, and compassion.
The statement promotes understanding by acknowledging the paper's main point about LLM inconsistencies. [+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 by questioning the experimental materials rather than attacking the authors personally. [+1]
Principle 5:
I will acknowledge and correct my mistakes.
The statement implicitly acknowledges the possibility of error in the paper's methods, aligning with the principle of acknowledging and correcting mistakes. [+1]