Gary Marcus

Rank 19 of 47
|
Score 64
6/15/2025, 2:58:33 AM
X
In reply to:
FeltSteam0
@FeltSteam
·
20d
@GaryMarcus @flowersslop @kimmonismus I am not convinced the human mind uses neurosymbolic systems. I looked at your studies but they do have their flaws. Your 1999 infant rule-learning paper failed a four-lab replication, and I think your 95' paper conflates novelty with default-rule use.
Gary Marcus
@GaryMarcus
·
20d
@flowersslop @kimmonismus 1. don’t build anything safety- or mission-critical on LLMs.
2. intelligence is multidimensional
3. LLMs are not a replacement for classical techniques; they are different set of tools with different strengths. we need both (hence my lifelong advocacy of neurosymbolic AI).
Flowers ☾ ❂
@flowersslop
·
20d
@GaryMarcus @kimmonismus Ok so we agree that both are completely different systems that work fundamentally different and have different purposes. But what is the takeaway here? Is a pocket calculator smarter than a brain because it can calculate faster? Is Stockfish more intelligent than Magnus Carlsen?
Gary Marcus
@GaryMarcus
·
20d
@flowersslop @kimmonismus neural networks lack reliable representations whereas pocket calculators and chess computers are built around them.
Flowers ☾ ❂
@flowersslop
·
20d
@GaryMarcus @kimmonismus And pocket calculators can calculate with 0% error rate while a probabilistic LLMs cant do that? So what is the takeaway here? That neural networks arent deterministic algorithms?

And the average to good chess player for SURE cant play blindfolded.
Gary Marcus
@GaryMarcus
·
20d
@flowersslop @kimmonismus many grandmasters can play blindfolded, and zillions of classical
AI chess systems can play from text.

The statement is part of a public discourse on the topic of neurosymbolic systems and AI, engaging in a debate about the validity of certain studies and their implications for AI development.

  1. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement attempts to promote understanding by referencing a meta-analysis with a larger sample size, which could provide more robust evidence. [+1]
  2. 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 objection to a study and providing additional evidence without resorting to personal attacks. [+1]