“Examples like these, in which ChatGPT confidently generates an image that does not actually meet requirements, are nicely complement the visual hallucinations that Ernest Davis and I reported here in October, in which visual ChatGPT misinterpreted various images:”
Our reading, the sentence that was checked:
In examples like those shown, ChatGPT confidently generated images that did not actually meet requirements, and in examples reported in October, visual ChatGPT misinterpreted various images.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Where’s Waldo? The Elephant in the Room, paragraph 6.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-wheres-waldo-the-elephant-in-the-78471f-a1
We recorded 9 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 2024-01-25. 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)
Conditions
- examples like these (scope, stated by the speaker)
- various images reported in October (scope, stated by the speaker)
From the speaker
- examples like these
- ChatGPT confidently generates an image that does not actually meet requirements
- reported here in October
- visual ChatGPT misinterpreted various images
Our interpretation
- This concerns the particular examples referenced, not ChatGPT's average performance.
- Both parts require verification: the January image-generation failures and the October 2023 image-interpretation failures.
- Confidence means an affirmative presentation of success or an unqualified incorrect description, not a measured internal probability.
- The target period covers the two sets of examples identified in the post.
Counts as supported if: Authenticated prompts, images and responses establish the stated failures in both sets of examples.
Counts as contradicted if: Complete original records establish that the alleged failures did not occur, that outputs met the relevant requirements, or that the examples were materially misrepresented.
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 searches located the referenced October article and a link to Fraser's January post. The readable records do not expose the original visual exchanges sufficiently to verify either part independently. Similar failures elsewhere would not prove that these particular examples were accurately described.
Limitations
- This is an access and verification gap, not evidence that Marcus's descriptions were false.
- The assessment does not estimate how frequently ChatGPT made such errors.
Evidence (0 items)
No evidence item settled this claim by the cutoff.
What would settle it
Accessible original conversations or authenticated captures containing the prompts, input images, generated images and complete responses for both sets of examples.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Where’s Waldo? The Elephant in the Room
Examples like these, in which ChatGPT confidently generates an image that does not actually meet requirements, are nicely complement the visual hallucinations that Ernest Davis and I reported here in October, in which visual ChatGPT misinterpreted various images:
- Where
- paragraph 6
- Published
- 2024-01-25
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-wheres-waldo-the-elephant-in-the-78471f-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-wheres-waldo-the-elephant-in-the-78471f-a1.