“Of course, the LLM companies have armies of humans; in my own brief experiments a couple hours later, OpenAI had apparently already patched this particular problem.”
Our reading, the sentence that was checked:
LLM companies have armies of humans, and when Marcus ran his own brief experiments a couple of hours later, OpenAI had apparently already patched this particular problem.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Why ChatGPT can’t be trusted with breaking news, paragraph 11.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-why-chatgpt-cant-be-trusted-with-17ff45-a1
We recorded 6 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 2026-01-03. 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)
Conditions
- apparently (qualifier, stated by the speaker)
- in my own brief experiments (scope, stated by the speaker)
- a couple hours later (qualifier, stated by the speaker)
From the speaker
- LLM companies have armies of humans
- in my own brief experiments
- a couple hours later
- apparently
- patched this particular problem
Our interpretation
- The specific event is the apparent correction Marcus says he observed on January 3, 2026, a few hours after Barrett's testing.
- Apparently qualifies the inference of a patch. It does not establish that Marcus knew of an internal deployment or human intervention.
- The minimum observable commitment is that his later tests appeared to acknowledge the capture correctly.
- Armies of humans means substantial human workforces, without a supplied numerical threshold. Earlier staffing records are context for the January 3 claim.
- The target day comes from the described episode, not from the dates of available staffing evidence.
Counts as supported if: Contemporaneous, independently verifiable test records show the error appeared corrected within the stated few-hour window, with records corroborating substantial human staffing. A claim of an actual targeted patch would additionally require deployment evidence.
Counts as contradicted if: Reliable records of Marcus's stated tests show that they still made the same denial, or deployment records directly disprove a specifically identified patch. An unrelated failing session would not refute his limited observation.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
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
A primary source corroborates substantial OpenAI staffing. It does not verify the later responses Marcus describes or establish that OpenAI changed the system within hours. The word apparently makes this a tentative inference, so lack of a patch announcement does not contradict it.
Limitations
- Private experiments may have occurred without a public record.
- Correct output would establish apparent improvement, but would not by itself identify a targeted patch or human intervention.
- The available staffing evidence is contextual and predates the target day.
Evidence (1 item)
Context only: OpenAI Forum,
past 3,000 globally.
Why it matters here: An OpenAI recruiter describes a workforce exceeding 3,000. This corroborates substantial staffing before the episode, but does not identify people correcting this error or establish a January 3 patch.
- Publisher
- OpenAI Forum
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
What would settle it
Timestamped records of the later tests, including prompts and configurations, would test the observed correction. Deployment or incident records would be needed to establish its cause.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Why ChatGPT can’t be trusted with breaking news
Of course, the LLM companies have armies of humans; in my own brief experiments a couple hours later, OpenAI had apparently already patched this particular problem.
- Where
- paragraph 11
- Published
- 2026-01-03
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-why-chatgpt-cant-be-trusted-with-17ff45-a1: Current. Outcome: Unresolved. 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-why-chatgpt-cant-be-trusted-with-17ff45-a1.