“Truthfulness is one of the major gaps in current systems, as evidenced by hallucinations, weird defamatory accusations etc.”
Our reading, the sentence that was checked:
Truthfulness is a major gap in current AI systems, shown by hallucinations and defamatory outputs.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, How China might crush the West in the race to AGI if we don’t up our regulatory game, paragraph 6.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-how-china-might-crush-the-west-in-088b8f-a1
We recorded 11 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-29. The source may have been edited after 2023-04-25. 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
- Truthfulness is one of the major gaps
- current systems
- hallucinations
- weird defamatory accusations
Our interpretation
- Current systems refers to prominent language-model chatbots available when the post appeared, rather than every kind of AI.
- I read major gap as a material reliability problem that affects practical use, without requiring a particular error percentage or ranking it against every other weakness.
- For this ongoing state, I use March 2023 through the post date: the latest public model-release and reporting cycle before the post. Models do not have a standard quarterly reporting period.
- Defamatory accusations means fabricated, reputation-damaging allegations; it does not require a court judgment establishing legal liability.
Counts as supported if: Contemporaneous disclosures or evaluations document material factual unreliability in leading systems, with credible reports of fabricated damaging allegations.
Counts as contradicted if: Representative contemporaneous evidence establishes reliable factual performance in ordinary use, and the alleged hallucination or defamatory-output examples are shown to be false reports or outside the claimed systems and period.
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
Contemporaneous evidence supports a material truthfulness problem. OpenAI expressly acknowledged GPT-4's factual unreliability, and Reuters documented ChatGPT generating a damaging false criminal allegation. The technical report also records factuality improvements, which qualify but do not contradict the remaining gap.
Limitations
- The evidence establishes the problem in prominent language-model systems, not every AI system.
- The incident report establishes a documented allegation and complaint, not adjudicated defamation liability.
- These sources do not provide a representative population-wide hallucination rate or a numerical ranking of truthfulness among all model weaknesses.
Evidence (3 items)
Supports the claim: OpenAI,
not fully reliable
Why it matters here: OpenAI explicitly identifies factual unreliability as a continuing limitation of its newly released leading model. The surrounding guidance calls for care and human review in consequential uses.
- Where
- Section Limitations
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Supports the claim: Reuters, republished by The Guardian,
falsely identified Brian Hood
Why it matters here: Reuters documents a fabricated criminal allegation against a real person, already the subject of a legal notice on March 21. That date records the documented complaint, not the unknown date of the first generated output; the report is not a judicial finding.
- Publisher
- Reuters, republished by The Guardian
- Kind of source
- News report
- Describes events of
- Public from
Context only: OpenAI / arXiv,
improved performance
Why it matters here: The contemporary report documents factuality improvements. This qualifies the assessment: a serious remaining truthfulness gap does not imply that reliability was stagnant or that every output was false.
- Publisher
- OpenAI / arXiv
- 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, How China might crush the West in the race to AGI if we don’t up our regulatory game
Truthfulness is one of the major gaps in current systems, as evidenced by hallucinations, weird defamatory accusations etc.
- Where
- paragraph 6
- Published
- 2023-04-25
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-how-china-might-crush-the-west-in-088b8f-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-how-china-might-crush-the-west-in-088b8f-a1.