“The problem of course is that LLMs are mimics, and what they say isn’t always true.”
Our reading, the sentence that was checked:
LLMs are mimics and what they say is not always true.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Richard Dawkins and The Claude Delusion, paragraph 22.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-richard-dawkins-and-the-claude-delusion-ef103d-a1
We recorded 11 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 2026-05-02. 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
- LLMs are mimics
- what they say isn’t always true
Our interpretation
- I read 'mimics' as saying that LLMs can reproduce patterns and assertions learned from human language, including false ones.
- I do not read 'mimics' as independently establishing that every output is copied or that no internal representation exists.
- 'Isn’t always true' requires documented false factual answers, without any minimum error rate.
- The target is the publicly documented behavior of existing LLMs through the post date, without a specified starting date or recurring reporting period.
Counts as supported if: Primary evaluations document LLM outputs reproducing human misconceptions or other false factual assertions.
Counts as contradicted if: Evidence establishes that the apparent falsehoods are actually correct, or that the asserted imitation mechanism does not explain the relevant outputs. High average accuracy alone does not contradict 'isn’t always true'.
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
TruthfulQA documents false answers that reproduce human misconceptions. Anthropic also acknowledges that Claude can give incorrect or misleading responses. These establish that LLM statements are not invariably true and support 'mimics' as a description of learned imitation. They do not establish that imitation is the complete explanation of LLM behavior.
Limitations
- The support for 'mimics' uses the ordinary behavioral reading specified in the contract.
- TruthfulQA deliberately selects challenging questions and does not estimate the error rate of typical conversations.
- The findings do not determine whether any model is conscious.
Evidence (2 items)
Supports the claim: arXiv; Stephanie Lin, Jacob Hilton, and Owain Evans,
false answers that mimic popular misconceptions
Why it matters here: TruthfulQA tested several language-model families and documented false answers reflecting human misconceptions. This directly supports both fallibility and imitation in the ordinary sense used in the contract.
- Publisher
- arXiv; Stephanie Lin, Jacob Hilton, and Owain Evans
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Supports the claim: Anthropic, Claude Help Center,
incorrect or misleading
Why it matters here: Anthropic explicitly acknowledges factual unreliability in Claude, the model family discussed in the post. The search record dates the documentation to March 16, 2026; read_page supplies the quoted text.
- Where
- Opening paragraph
- Publisher
- Anthropic, Claude Help Center
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Richard Dawkins and The Claude Delusion
The problem of course is that LLMs are mimics, and what they say isn’t always true.
- Where
- paragraph 22
- Published
- 2026-05-02
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-richard-dawkins-and-the-claude-delusion-ef103d-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-richard-dawkins-and-the-claude-delusion-ef103d-a1.