Gary Marcus

Rank 13 of 47
|
Score 84
6/27/2025, 2:25:39 PM
X
In reply to:
David Manheim
@davidmanheim
·
227d
@GaryMarcus It might also be a vindication of those claiming that in practice, even non-symbolic methods can end up with human-level capabilities once scaled and extended, whether directly or not - and that there may be no barrier to improving far past the point of human capability.
Gary Marcus
@GaryMarcus
·
227d
new result (when correctly interpreted): LLMs on their own can’t reason. They need external tools, typically built with (often-derided) classic symbolic techniques, in order to reason reliably.

HUGE vindication for those of long favoring neurosymbolic approaches.
Wes Roth
@WesRothMoney
·
228d
Pfizer researchers argue that what looks like a collapse in AI reasoning may actually be an agentic gap—models failing not in thought, but in action.

When given tools, the same models crushed tasks they had just failed.
The problem isn’t thinking—it’s interface.

A must-read

The statement is part of a technical discussion on AI capabilities, specifically addressing the role of symbolic methods in AI reasoning. It clarifies a previous argument about AI's potential to exceed human capabilities, emphasizing the necessity of symbols in the process.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not cause harm and contributes to a technical discussion. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    It promotes understanding by clarifying a position on AI capabilities. [+1]
  3. 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.
    Engages in constructive dialogue by addressing a specific point in the conversation without personal attacks. [+1]