Gary Marcus

Rank 13 of 47

Score 115

@flowersslop @WilKranz there have been several recent, careful academic articles that have made this case; dismissing them as midwits etc is beneath you.

engage in the actual arguments.

9/6/2025, 4:54:00 AM

In reply to:

Flowers ☾

@flowersslop

@WilKranz its just midwit self importance. dumb and intelligent people know it can think

William Kranz

@WilKranz

@flowersslop it's genuinely idiotic to claim they can't/aren't reasoning. it's the most obvious thing in the world to me that they are. when a human child makes an honest mistake, do you accuse them of not having a brain? so dumb.

Flowers ☾

@flowersslop

People who chant “LLMs can’t reason; they just predict the next token” are the real stochastic parrots:

> learned about LLMs in late 2022 (knowledge cutoff)
> claims to be an LLM expert (hallucination)
> read “next-token prediction” once and won’t shut up (overfitting)
> doesn’t

The statement engages in a discussion about the reasoning capabilities of large language models (LLMs), which is a topic of public interest in the context of AI development and its implications. The tone is corrective and encourages engagement with academic arguments rather than dismissive language.

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 encourages constructive criticism and dialogue by urging others to engage with academic arguments rather than resorting to dismissive language. It avoids personal attacks and promotes a more informed discussion. [+2]