“My own view, which I stand by, is that we need “neurosymbolic” contributions from both traditions if we are to get to trustworthy AI.”
Our reading, the sentence that was checked:
In Marcus's stated view, which he stands by, getting to trustworthy AI requires neurosymbolic contributions from both the neural network and classical symbolic traditions.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Breaking: Marcus weighs in (mostly) for LeCun, paragraph 10.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-breaking-marcus-weighs-in-mostly-749268-a2
We recorded 6 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2026-01-06. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Causal claim
- 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)
Conditions
- my own view, which I stand by (qualifier, stated by the speaker)
- if we are to get to trustworthy AI (qualifier, stated by the speaker)
From the speaker
- we need
- “neurosymbolic” contributions from both traditions
- if we are to get to trustworthy AI
Our interpretation
- The plain reading is that dependable and safe AI requires contributions from both neural networks and classical symbolic AI.
- My own view identifies the speaker's position; it does not remove the substantive necessity claim.
- Trustworthiness must be evaluated across intended uses. One high benchmark score is insufficient.
- The surrounding discussion concerns broadly capable AI. A narrow, dependable symbolic tool is relevant counterevidence to an unrestricted reading but does not automatically settle that broader claim.
- The statement is a causal requirement, not a forecast with an unstated deadline.
Counts as supported if: Evidence establishes that both traditions are necessary for the relevant level and scope of trustworthiness, rather than merely useful in successful examples.
Counts as contradicted if: A transparently documented AI system satisfies appropriate independent trustworthiness evaluations across the relevant intended uses while lacking contributions from one of the two traditions.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Does the evidence support the cause?
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
NIST's January 26, 2023 framework treats trustworthiness as several characteristics balanced for the system's context of use; it does not establish an architectural requirement [e1]. The preprint revised April 11, 2024 reports that symbolic Wu's method alone solves 15 geometry problems, but does not evaluate broad AI trustworthiness [e2]. These sources establish neither that both traditions are necessary nor that a system lacking one meets independent trustworthiness requirements across the broader intended uses [e1, e2].
Limitations
- The evidence provides no agreed application scope, failure tolerance, or boundary for what counts as a contribution from either tradition.
- NIST supplies guidance, not experimental evidence of architectural necessity.
- The symbolic geometry result is a preprint covering a narrow task, without comprehensive trustworthiness evaluation.
Evidence (2 items)
Context only: National Institute of Standards and Technology,
the AI system’s context of use.
Why it matters here: NIST treats trustworthiness as a combination of characteristics assessed in context. This official framework is guidance, not an experiment or a finding that any particular architecture is necessary.
- Publisher
- National Institute of Standards and Technology
- Kind of source
- Official statistics
- Describes events of
- Public from
Context only: arXiv; Sinha and colleagues,
can solve 15 problems
Why it matters here: The paper reports substantive geometry results from a symbolic method without neural guidance. That challenges an unrestricted claim that both traditions are needed for every dependable reasoning application, but it does not evaluate broad AI trustworthiness.
- Where
- Abstract
- Publisher
- arXiv; Sinha and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
Define the intended uses, trustworthiness thresholds, and what counts as a contribution from each tradition. Then provide either evidence establishing that both are necessary or architecture documentation and independent evaluations showing that a system meets those requirements without one tradition.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Breaking: Marcus weighs in (mostly) for LeCun
My own view, which I stand by, is that we need “neurosymbolic” contributions from both traditions if we are to get to trustworthy AI.
- Where
- paragraph 10
- Published
- 2026-01-06
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-breaking-marcus-weighs-in-mostly-749268-a2: Current. Outcome: Unresolved. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment marcus-breaking-marcus-weighs-in-mostly-749268-a2.