Gary Marcus

Rank 14 of 47
|
Score 83
5/18/2025, 2:39:11 AM
X
In reply to:
Andrew Prock
@andrewprock
·
244d
@littmath But they do have zero understanding of the world. They can interpolate within their training set, but give it an opportunity to demonstrate trivial understanding by having it step outside that training set results in catastrophic failures.
Gary Marcus
@GaryMarcus
·
244d
You have got to be kidding me.

I first wrote about how (certain kinds) of neural nets had trouble with concepts like a pair and three of kind in … 1998.

Here’s o3, today:
Daniel Litt
@littmath
·
244d
The specific claims “zero abstracting ability” and “zero understanding of the world” are also clearly false.
Daniel Litt
@littmath
·
244d
I don’t know how you can use frontier models and hold onto beliefs like this. Not sure about “conscious,” but for any reasonable meaning of “intelligent,” LLMs are already intelligent.

The statement is part of a discussion on the capabilities and limitations of large language models (LLMs). It critiques the assertion that LLMs have 'zero understanding' by suggesting a 'superficial understanding' might be more accurate. The tone is analytical and aims to refine the argument rather than dismiss it outright.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not cause harm and seeks to clarify a point, aligning with the principle of doing no harm. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    It promotes understanding by offering a nuanced perspective on the capabilities of LLMs. [+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.
    The statement engages in constructive dialogue by refining the argument rather than attacking the original claim. [+1]