“We have no way of guaranteeing that large language models will reliably stick to facts, a point so thoroughly established by now I don’t think it needs a reference.”
Our reading, the sentence that was checked:
There is no way to guarantee that large language models reliably stick to facts.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, A CERN for AI and the Global Governance of AI, paragraph 10.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-a-cern-for-ai-and-the-global-governance-87d6e2-a1
We recorded 13 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 2023-10-26. 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
- We have no way of guaranteeing
- large language models
- will reliably stick to facts
- by now
Our interpretation
- The ordinary reading is that no established method assured factual reliability across general-purpose, open-ended LLM use. This is not a claim that constrained applications cannot provide any guarantees.
- The statement describes the research position in October 2023. I use October 1 through the post date, October 26, as the current reporting window; research has no standard quarterly reporting period.
- Support requires a contemporary synthesis of the unresolved problem, rather than treating isolated mistakes as proof that every possible method fails.
Counts as supported if: Contemporary research identifies general factual reliability as unresolved after considering available mitigation methods, with no established general assurance method found.
Counts as contradicted if: A method available by the post date provided a substantiated factual-reliability guarantee covering general-purpose, open-ended LLM answers.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Supported
The dated evidence matches what the claim committed to, under the criteria frozen before grading.
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 October 2023 survey treats factual reliability as unresolved after reviewing methods including retrieval and further training. That supports Marcus's ordinary claim that general-purpose LLM factuality could not be assured. The finding does not mean that factuality improvements were absent or that guarantees for restricted tasks were impossible.
Limitations
- Support rests on a contemporary research synthesis, not an exhaustive proof of nonexistence.
- The result depends on reading the claim as about general-purpose use.
- No numerical reliability threshold was supplied.
Evidence (1 item)
Supports the claim: arXiv; Cunxiang Wang and colleagues,
not without limitations
Why it matters here: The contemporary survey considers factuality evaluation and enhancement for standalone and retrieval-augmented models, and describes fully factual LLMs as an outstanding research goal. This supports the ordinary research-state reading, although it does not prove a universal impossibility.
- Publisher
- arXiv; Cunxiang Wang and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, A CERN for AI and the Global Governance of AI
We have no way of guaranteeing that large language models will reliably stick to facts, a point so thoroughly established by now I don’t think it needs a reference.
- Where
- paragraph 10
- Published
- 2023-10-26
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-a-cern-for-ai-and-the-global-governance-87d6e2-a1: Current. Outcome: Supported. 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-a-cern-for-ai-and-the-global-governance-87d6e2-a1.