“Now? Here’s what I got out of Imagen 4, showing failure at the feature “binding problem” that some of us have been harping on for three decades:”
Our reading, the sentence that was checked:
Imagen 4 still fails at the feature binding problem on the stacked colored blocks prompt.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Image generation: Still crazy after all these years, paragraph 27.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-image-generation-still-crazy-after-3a9c1a-a1
We recorded 4 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2025-06-25. 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
- DeWeese lab block-stacking prompt (qualifier, stated by the speaker)
From the speaker
- Now
- what I got out of Imagen 4
- showing failure at the feature “binding problem”
- A red conical block on top of a grey cubic block on top of a blue cylindrical block, with a green cubic block nearby
Our interpretation
- This reports a failure in the particular displayed output, rather than asserting that every possible generation fails.
- Binding is tested by matching each specified color to its shape and each object to its requested position.
- Success requires a red cone above a grey cube above a blue cylinder, plus a separate green cube nearby.
- The target period is the June 24–25 testing episode described in the surrounding passage.
Counts as supported if: Independent inspection of the authenticated reported output identifies at least one clear mismatch in a specified color-shape pairing or spatial relationship.
Counts as contradicted if: Independent inspection establishes that the reported output satisfies all specified color-shape pairings and spatial relationships.
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 prompt provides a clear checklist, so this claim is checkable. However, the retrieved text supplies only Marcus's description of the outcome. General evidence that Imagen 4 struggles with composition cannot establish an error in this particular image. The available records therefore do not settle the claim.
Limitations
- The decisive artifact is an image that the retrieved plain text does not make inspectable.
- A fresh successful or unsuccessful generation would not determine what the reported historical output showed.
Evidence (0 items)
No evidence item settled this claim by the cutoff.
What would settle it
An independent inspection of an authenticated copy of the reported block image, scored against the prompt's colors, shapes, and spatial relationships.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Image generation: Still crazy after all these years
Now? Here’s what I got out of Imagen 4, showing failure at the feature “binding problem” that some of us have been harping on for three decades:
- Where
- paragraph 27
- Published
- 2025-06-25
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-image-generation-still-crazy-after-3a9c1a-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-image-generation-still-crazy-after-3a9c1a-a1.