“It’s fun to compare the two—they have so many differences—even though neither can be trusted.”
Our reading, the sentence that was checked:
GPT cannot be trusted.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, A Tale of Two Liars, paragraph 7.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-a-tale-of-two-liars-8c6c20-a2
We recorded 3 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 2023-11-22. 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
- neither can be trusted
Our interpretation
- Only the supplied GPT proposition is assessed; the comparison with George Santos is not graded.
- In this discussion of factual falsehoods, the ordinary reading is that GPT’s factual answers require independent checking.
- The claim does not mean that every answer is wrong or that GPT has no useful applications.
- For this ongoing state, the target is the latest reportable month-to-date before publication: November 1–22, 2023. GPT has no fixed factual-reliability reporting cycle.
Counts as supported if: Contemporaneous measurements show recurring factual errors in GPT outputs sufficient to make unchecked reliance unsafe, including when source information is supplied.
Counts as contradicted if: Broad, credible evaluations show that the relevant GPT systems provide factual information reliably enough to accept without independent checking across the uses discussed.
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
Vectara’s November 1, 2023 results reported hallucinations in 3.0% of GPT-4 summaries and 3.5% of GPT-3.5 Turbo summaries, using an automated model to detect unsupported content when source documents were supplied [e6]. These recurring errors support the interpretation that GPT’s factual outputs required independent checking during the November 1–22 assessment period [e6]. Vectara’s November 6 announcement reported that OpenAI’s models performed best among those tested, but that comparative lead did not eliminate the observed errors [e6, e7].
Limitations
- The conclusion concerns unchecked factual reliance; the summarization results do not establish that every GPT answer is false or that GPT has no useful applications [e6].
- These task-specific rates do not measure factual reliability across all domains or GPT deployments [e6].
- The evaluation used an automated hallucination detector, rather than establishing each classification through independent human verification [e6].
Evidence (2 items)
Supports the claim: Vectara,
introduces hallucinations
Why it matters here: The preserved table reports hallucinations in 3.0% of GPT-4 summaries and 3.5% of GPT-3.5 Turbo summaries, despite providing source documents. These recurring errors support independent checking; the percentages apply to this summarization test, not all GPT answers.
- Publisher
- Vectara
- Kind of source
- Benchmark result
- Describes events of
- Public from
Context only: Vectara via GlobeNewswire,
the strongest performance
Why it matters here: OpenAI’s models performed comparatively well in the same evaluation. This limits any interpretation that GPT was generally incapable or uniquely unreliable, but does not remove the observed errors.
- Publisher
- Vectara via GlobeNewswire
- 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, A Tale of Two Liars
It’s fun to compare the two—they have so many differences—even though neither can be trusted.
- Where
- paragraph 7
- Published
- 2023-11-22
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-a-tale-of-two-liars-8c6c20-a2: 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.