“But it’s also pretty prone to surprising stumbles, like this uncomprehending reply to “a red cube on top of a blue cube”.”
Our reading, the sentence that was checked:
DALL-E 2 is fairly prone to uncomprehending errors such as failing to render 'a red cube on top of a blue cube'.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, The New Science of Alt Intelligence, paragraph 19.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-the-new-science-of-alt-intelligence-d724f8-a1
We recorded 20 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 2022-05-14. 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
- pretty prone to surprising stumbles
- a red cube on top of a blue cube
- DALL-E 2
Our interpretation
- The ordinary reading is that the original DALL-E 2 repeatedly mishandles simple compositional instructions, including assigning the requested colors to objects.
- Uncomprehending describes the observable mismatch between request and output; it does not establish an internal mental state.
- Pretty prone supplies no numerical failure threshold and is assessed as a recurring failure mode.
- For this ongoing observation, I use the latest public evaluation cycle, April 1 through May 14, 2022, surrounding the original model's introduction.
Counts as supported if: Original-model evaluations document recurring compositional or attribute-binding failures, including the cube example or an equivalent test.
Counts as contradicted if: Representative original-model tests establish that these failures are rare and show that the cited example was misidentified or did not reflect the model's behavior.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Supported
The dated evidence matches what the claim committed to, under the criteria frozen before grading.
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 DALL-E 2 paper independently documents the exact cube example and identifies recurring difficulty assigning attributes to the correct objects. That supports the ordinary qualitative reading of Marcus's claim. It does not establish a numerical error rate or prove anything about the model's internal understanding.
Limitations
- Support concerns a recurring compositional failure mode, not a measured failure rate across all prompts.
- The evidence is the developers' technical evaluation.
Evidence (1 item)
Supports the claim: arXiv; OpenAI,
binding attributes to objects
Why it matters here: The authors discuss the exact red-cube/blue-cube prompt in Figure 14 and report recurring object–attribute mix-ups in reconstructions in Figure 15. This supports a recurring compositional weakness in the original model, without establishing a population-wide error rate.
- Publisher
- arXiv; OpenAI
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, The New Science of Alt Intelligence
But it’s also pretty prone to surprising stumbles, like this uncomprehending reply to “a red cube on top of a blue cube”.
- Where
- paragraph 19
- Published
- 2022-05-14
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-the-new-science-of-alt-intelligence-d724f8-a1: Current. Outcome: Supported. 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-the-new-science-of-alt-intelligence-d724f8-a1.