“As Ernie Davis, Scott Aaronson, and I showed, a week or so after the release, it’s easily fooled, especially with complex sentences and words like “similar” that would require a deep understanding of language to be interpreted correctly”
Our reading, the sentence that was checked:
DALL-E 2 is easily fooled, especially by complex sentences and words requiring deep language understanding.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Horse rides astronaut, paragraph 18.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-horse-rides-astronaut-61347d-a1
We recorded 10 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 2022-05-28. 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
- especially complex sentences and words like 'similar' (qualifier, stated by the speaker)
From the speaker
- a week or so after the release
- it’s easily fooled
- especially with complex sentences and words like “similar”
Our interpretation
- The context explicitly places these observations in April 2022. The target is the original DALL-E 2.
- The plain reading is that complex captions repeatedly produce substantial errors in objects, attributes or relationships, with particular difficulty interpreting relational language.
- “Similar” is an illustrative example. No numerical failure rate or precise meaning of “deep understanding” is supplied.
- Correctness concerns whether images satisfy the caption, rather than their visual quality.
Counts as supported if: Independent records of the April 2022 model show repeated substantive failures on complex relational captions, including evidence that linguistic complexity or relational expressions create particular difficulty.
Counts as contradicted if: Comparable controlled tests of the April 2022 model show reliable interpretation of complex relational captions, without the claimed particular difficulty.
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 original model paper documents attribute-binding problems, so there is evidence for a narrower weakness. The retrieved independent evidence does not settle the stronger claim about frequent failures, especially on complex sentences and relational expressions, during April 2022. Marcus’s own study is excluded by the evidence rules. Later evaluations cannot automatically establish the earlier state.
Limitations
- Unresolved does not mean the claim was false.
- The evidence supports some compositional weaknesses without establishing the full comparison Marcus makes.
Evidence (1 item)
Context only: OpenAI authors, published through arXiv,
worse at binding attributes to objects
Why it matters here: The original model paper documents difficulty assigning properties to objects. This supports a narrower weakness, but does not establish frequent failures specifically associated with complex syntax or expressions such as “similar.”
- Publisher
- OpenAI authors, published through arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
Independent April prompt-output records, or a documented evaluation of the preserved April model, testing complex relational captions against simpler controls with all outputs retained.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Horse rides astronaut
As Ernie Davis, Scott Aaronson, and I showed, a week or so after the release, it’s easily fooled, especially with complex sentences and words like “similar” that would require a deep understanding of language to be interpreted correctly
- Where
- paragraph 18
- Published
- 2022-05-28
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-horse-rides-astronaut-61347d-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.