Pundit ClaimsAI claims by Gary Marcus and Ed Zitron, with the evidence

Gary Marcus Observation First said

“This symmetry fail is mighty big, a mighty persistent error that has endured for decades.”

Elegant and powerful new result that seriously undermines large language models, paragraph 24. The full statement is at the source.

Our reading, the sentence that was checked:

The symmetry failure is a big, persistent error in neural networks that has endured for decades.

Assessment: Unresolved Evidence as of

Where and when Gary Marcus wrote it

Marcus on AI, Elegant and powerful new result that seriously undermines large language models, paragraph 24.

Said
Source observed
live page, full text, public
Evidence as of
the cutoff for this assessment
Assessment recorded
assessment marcus-elegant-and-powerful-new-result-that-59f0e0-a1

We recorded 8 claims from this post. See every claim from this post.

The wording above is the version this site observed on 2026-10-29. The source may have been edited after 2023-09-22. This page does not claim the wording was captured then.

What the claim commits to

Kind
Observation
Deadline, as the speaker wrote it
None stated. This site does not add one.
Deadline state when assessed
No horizon
Deadline state on 2026-10-29
No horizon (the release’s reference date)

From the speaker

  • This symmetry fail
  • a mighty persistent error
  • endured for decades

Our interpretation

  • The checkable core is recurrence of comparable symmetry or reversal-generalization failures across neural-network generations separated by at least twenty years.
  • Persistence does not require every intervening model to fail or an unchanged underlying mechanism.
  • The historical period begins with the 1980s discussion named in the surrounding text and ends at the post.
  • Mighty big is an assessment of importance, not a supplied quantitative threshold.

Counts as supported if: Independent primary records establish comparable relational reversal failures in neural networks at dates at least twenty years apart, including the contemporary endpoint.

Counts as contradicted if: The historical records alleged to demonstrate the earlier failure instead establish that it did not occur or concerned a materially different phenomenon, with the claimed continuity thereby disproved.

Retrospective: these criteria were written after the outcome was public.

Current assessment

Was it true for the period described?

Outcome: Unresolved

The record was checked, and the evidence available by the cutoff does not settle it.

Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.

Evidence as of
Deadline state when assessed
No horizon
Assessment recorded
Policy version
2026-10-28.7

Rationale

The records establish an old debate about systematicity and a modern direction-sensitive generalization failure. They do not establish that the same symmetry error persisted across decades. Treating every systematicity concern as the same failure would decide the claim through an interpretation Marcus did not specify.

Limitations

  • Failure to locate a matching historical experiment is not evidence that none exists.
  • The broader claim that systematicity has been debated for decades is easier to establish than this claim about a persistent error.

Evidence (2 items)

  • Context only: Cognition; Fodor and Pylyshyn, author-text mirror,

    There are, however, no proposals for such a mechanism.

    Why it matters here: The passage challenges how connectionist architectures guarantee systematic representations, while acknowledging compatibility with systematicity. It is not an empirical demonstration of failed inverse-fact recall; availability uses the conservative last day of the reported publication month.

    Publisher
    Cognition; Fodor and Pylyshyn, author-text mirror
    Kind of source
    Peer-reviewed paper
    Describes events of
    Public from
  • Context only: arXiv; Grosse et al.,

    This trend is consistent with different model sizes.

    Why it matters here: Reversing the order of otherwise equivalent training content sharply reduces its measured influence. This supplies a contemporary endpoint, but no evidence of a matching failure decades earlier.

    Publisher
    arXiv; Grosse et al.
    Kind of source
    Preprint, not peer reviewed
    Describes events of
    Public from

What would settle it

An independent primary experiment or precise formal result from 2003 or earlier demonstrating a comparable failure to infer an inverse relation from a learned relation, with enough methodological detail to compare it fairly with the 2023 tests.

Where it was said (1)

Said once in the sources this register covers.

  • : Marcus on AI, Elegant and powerful new result that seriously undermines large language models

    This symmetry fail is mighty big, a mighty persistent error that has endured for decades.

    Published
    2023-09-22
    Source observed
    , live page
    Access
    full text, public

History

Assessment versions

Recorded events

  1. Assessment publishedBy this site

    First assessment published after both reviews and the final-text check passed.

    About assessment marcus-elegant-and-powerful-new-result-that-59f0e0-a1.