“[LLMs] have always been prone to hallucinations and errors.”
Our reading, the sentence that was checked:
LLMs have always been prone to hallucinations and errors.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Is AGI the right goal for AI?, paragraph 3.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-is-agi-the-right-goal-for-ai-b23dd5-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 2025-10-16. 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
- have always been prone
- hallucinations and errors
Our interpretation
- The ordinary reading is persistent susceptibility across successive LLM generations through publication, rather than errors in every response or an unchanged error rate.
- The historical period ends on 2025-10-16. Its beginning is unspecified because Marcus gives no starting date or definition of the first LLM.
- Hallucinations means plausible but false factual statements; errors also includes incorrect reasoning.
Counts as supported if: Evaluations across early and later LLM generations document recurring false statements or reasoning errors, including models current before publication.
Counts as contradicted if: Broad evaluations establish that a subsequent generation had reliably eliminated these problems before publication, undermining the assertion of persistent susceptibility.
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
Evaluations and contemporaneous disclosures document factual errors in early model families, hallucinations and reasoning errors in GPT-4, and continued hallucinations in GPT-5 before the post. This supports persistent susceptibility. The same sources document improvement, so the finding does not imply unchanged error rates or errors in every answer.
Limitations
- Benchmark error rates do not directly measure ordinary usage.
- The historical evidence is representative rather than exhaustive.
Evidence (3 items)
Supports the claim: arXiv; Stephanie Lin, Jacob Hilton and Owain Evans,
truthful on 58% of questions
Why it matters here: The evaluation tested GPT-2, GPT-3, GPT-Neo/J and a T5-based model on questions designed to elicit common falsehoods. It documents substantial factual-error susceptibility across several early model families, without estimating everyday error rates.
- Where
- Abstract
- Publisher
- arXiv; Stephanie Lin, Jacob Hilton and Owain Evans
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Supports the claim: OpenAI,
makes reasoning errors
Why it matters here: The GPT-4 release disclosure explicitly identifies both hallucinations and reasoning errors. The same section reports improved factuality relative to GPT-3.5, which limits any reading that error rates had not improved.
- Where
- Limitations
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Supports the claim: OpenAI,
but they still occur.
Why it matters here: This confirms that hallucinations persisted in a contemporary generation shortly before Marcus's post. Its explicit acknowledgment of fewer hallucinations is material qualification against an interpretation of no progress.
- Publisher
- OpenAI
- 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, Is AGI the right goal for AI?
[LLMs] have always been prone to hallucinations and errors.
- Where
- paragraph 3
- Published
- 2025-10-16
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-is-agi-the-right-goal-for-ai-b23dd5-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-is-agi-the-right-goal-for-ai-b23dd5-a1.