“But the anti-neurosymbolic tradition is limited; its most visible manifestation, LLMs, has brought fame and money, but no robust solution to solving any particular problem with great reliability.”
Our reading, the sentence that was checked:
The anti-neurosymbolic tradition is limited. Its most visible manifestation, LLMs, has brought fame and money but no robust solution to solving any particular problem with great reliability.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Two Nobel Prizes for AI, and Two Paths Forward, paragraph 30.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-two-nobel-prizes-for-ai-and-two-paths-f71b30-a1
We recorded 7 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 2024-10-09. 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
- its most visible manifestation, LLMs
- no robust solution
- any particular problem
- with great reliability
Our interpretation
- Assess the substantive universal claim that LLMs had delivered no highly reliable solution to even one particular task.
- The ordinary reading allows narrow practical tasks; it does not require general intelligence or perfect performance.
- Robust reliability means consistently high task success on representative inputs, including ordinary variations, rather than one favorable example.
- Prompting and fine-tuning are eligible. A separate symbolic solver doing the essential work would not directly test the claim about the anti-neurosymbolic tradition.
- 'Has brought' describes accumulated achievements through 9 October 2024. Later model capabilities cannot refute this historical observation.
- 'Fame and money' is background characterization, not a specified company financial metric.
Counts as supported if: Broad evidence establishes that the claimed class had produced no qualifying reliable task solution by the post date. Failures on selected tasks alone are insufficient to establish this universal negative.
Counts as contradicted if: At least one LLM-based solution available by the post date demonstrates consistently high success on a defined practical task under representative conditions and ordinary variations.
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
Contemporaneous studies document real reliability problems, but those findings do not prove Marcus's universal negative. Other pre-post studies report classification accuracy around 97–99%, with some adaptation and consistency checks. These are substantial counterevidence. The retrieved records do not settle whether a candidate meets the stronger requirement of robust performance under representative operating variations. A categorical verdict would depend on an added reliability threshold.
Limitations
- High accuracy is evidence of useful capability and should not be dismissed merely because errors remain.
- Benchmark success and dependable deployment are different evidentiary claims.
- The candidate counterexamples are author-reported preprints.
- Post-October-2024 model improvements were not used to judge the historical claim.
Evidence (3 items)
Context only: Nature,
may have become less reliable.
Why it matters here: The study finds persistent errors and sensitivity to natural variations in questions across the tested model families. This supports concern about reliability, but cannot establish that no particular reliable application existed.
- Where
- Abstract
- Publisher
- Nature
- Kind of source
- Peer-reviewed paper
- Describes events of
- Public from
Context only: arXiv,
outperforms fine-tuned RoBBERTa
Why it matters here: CARP reports 97.39% accuracy on SST-2 and 98.78% on R8; its adaptation experiment retains 96.29% on SST-2 when demonstrations come from Yelp. These are material counterexamples to a weaker claim of no high-accuracy task performance, but do not fully establish operational robustness.
- Where
- Sections 5.2 and 5.4
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv,
more reliable.
Why it matters here: The study reports roughly 99% SMS-classification accuracy after fine-tuning and improved output consistency. This is a substantial candidate counterexample, but the retrieved evidence does not establish sustained reliability on representative deployment traffic and ordinary changes in that traffic.
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
A pre-post evaluation or deployment record establishing high, repeatable success on one defined practical task under representative variations, with an error tolerance appropriate to that task. Supporting the universal negative would require much broader coverage than the failure studies provide.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Two Nobel Prizes for AI, and Two Paths Forward
But the anti-neurosymbolic tradition is limited; its most visible manifestation, LLMs, has brought fame and money, but no robust solution to solving any particular problem with great reliability.
- Where
- paragraph 30
- Published
- 2024-10-09
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-two-nobel-prizes-for-ai-and-two-paths-f71b30-a1: 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-two-nobel-prizes-for-ai-and-two-paths-f71b30-a1.