“(Not one of my predictions about GPT-4 proved to be incorrect; every flaw I predicted would persist persisted.)”
Our reading, the sentence that was checked:
Marcus says none of his predictions about GPT-4 proved incorrect, and every flaw he predicted would persist in GPT-4 did persist.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, GPT-5 and irrational exuberance, paragraph 8.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-gpt-5-and-irrational-exuberance-d8c80f-a1
We recorded 17 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 2023-04-08. 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
- Not one of my predictions about GPT-4 proved to be incorrect
- every flaw I predicted would persist persisted
Our interpretation
- This is a retrospective assertion about predictions made before GPT-4 and its documented behavior through the post date.
- The target period is GPT-4's release on March 14, 2023 through April 8, 2023, rather than later GPT-4 updates.
- Support requires checking the relevant prediction inventory, not merely finding some accurate predictions.
- A future prediction still awaiting its outcome is not an incorrect prediction.
- Solving particular old examples does not refute a prediction that a broader failure category remains unreliable.
Counts as supported if: A sufficiently complete inventory shows no relevant prediction falsified by April 8, 2023, and contemporaneous evidence documents persistence of every predicted GPT-4 flaw.
Counts as contradicted if: At least one relevant prediction was demonstrably false under its original conditions by April 8, 2023, or a flaw predicted to persist was shown to have been eliminated.
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
Contemporaneous OpenAI documentation supports Marcus's predictions that hallucinations and reasoning errors would persist. His earlier essay contains additional claims whose thresholds and outcomes require separate checks. The records retrieved do not establish that every relevant prediction survived and every predicted flaw persisted. Nor do they establish a qualifying counterexample. Pending long-term forecasts are not counted as errors.
Limitations
- Confirming several central predictions cannot establish a universal track-record claim.
- Later GPT-4 improvements cannot decide what was true by April 8, 2023.
- Marcus's earlier writing is used only to identify his commitments.
Evidence (2 items)
Supports the claim: OpenAI,
makes reasoning errors
Why it matters here: OpenAI confirms persistence of factual hallucinations and reasoning errors at GPT-4's release. This supports those predictions but does not verify the universal assertion about every prediction and flaw.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: Gary Marcus, Marcus on AI,
critical, unsolved problem.
Why it matters here: This establishes an original prediction, not its success. The essay also covers reasoning, hallucinations, guardrails, downstream integration, general intelligence, and longer-term architectural change.
- Publisher
- Gary Marcus, Marcus on AI
- Kind of source
- Secondary summary
- Describes events of
- Public from
What would settle it
An exhaustive dated inventory and contemporaneous outcome evidence for each relevant prediction. One matched, established counterexample would contradict the universal claim; support requires coverage of every predicted persistent flaw.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, GPT-5 and irrational exuberance
(Not one of my predictions about GPT-4 proved to be incorrect; every flaw I predicted would persist persisted.)
- Where
- paragraph 8
- Published
- 2023-04-08
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-gpt-5-and-irrational-exuberance-d8c80f-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-gpt-5-and-irrational-exuberance-d8c80f-a1.