“The reality, though is that these systems have never been been stable.”
Our reading, the sentence that was checked:
Generative AI systems have never been stable.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, ChatGPT has gone berserk, paragraph 9.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-chatgpt-has-gone-berserk-74297d-a1
We recorded 8 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 2024-02-21. 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
- these systems
- never
- stable
Our interpretation
- In context, these systems means generative AI systems, especially conversational language models.
- The ordinary reading of stable is reliably coherent and dependable in ordinary operation. It does not require identical answers or flawless performance.
- Never makes this a claim about the systems' history through publication. No starting date is supplied, so the historical start remains unspecified.
- Individual incidents can establish failures but cannot alone establish that the whole class never achieved stable operation.
Counts as supported if: Broad historical evidence establishes persistent operational unreliability across the relevant systems, sufficient to support the claim that stable operation had never been achieved.
Counts as contradicted if: Credible longitudinal evidence demonstrates sustained, reliably coherent operation by relevant systems under representative ordinary use before publication.
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
The records establish real reliability problems, including limitations acknowledged before this incident. They also record restored normal operation. Neither fact settles whether generative AI systems had never been stable. Supporting that sweeping statement from selected failures would overstate the evidence; rejecting it from isolated successes would do the same.
Limitations
- Stable has an ordinary meaning, but its boundary is not specified.
- The evidence concerns particular models and episodes rather than the entire history of generative AI.
Evidence (2 items)
Context only: OpenAI via arXiv,
not fully reliable
Why it matters here: The report establishes acknowledged reliability problems in GPT-4 before the incident. Imperfect reliability in one model does not establish that generative AI systems have never been stable.
- Publisher
- OpenAI via arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: OpenAI,
ChatGPT is operating normally.
Why it matters here: OpenAI reported restoration of normal operation. This weighs against treating the episode as uninterrupted malfunction, but a recovery announcement is not a longitudinal demonstration of stability.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
What would settle it
Representative longitudinal reliability evidence from before publication, assessed against a justified interpretation of stable operation. A clearer statement of the intended systems, duration, and tolerance for failures would make a firm assessment possible.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, ChatGPT has gone berserk
The reality, though is that these systems have never been been stable.
- Where
- paragraph 9
- Published
- 2024-02-21
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-chatgpt-has-gone-berserk-74297d-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.