“This should be a piece of cake, but for currently popular systems, it’s not.”
Our reading, the sentence that was checked:
Currently popular vision-language systems cannot reliably tell apart captions that use the same words in a different order.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Did GoogleAI Just Snooker One of Silicon Valley’s Sharpest Minds?, paragraph 27.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-did-googleai-just-snooker-one-of-c24cf3-a1
We recorded 12 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 2022-09-14. 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
- currently popular systems
- This should be a piece of cake
- it’s not
- both captions contain a completely identical set of words/morphemes, only in a different order
Our interpretation
- Assess the factual claim that contemporary vision-language systems did not reliably match images to captions when the distinction depends on word order.
- The context explicitly names Winoground, so its evaluated models and matching task provide the principal test.
- Reliably means consistently successful matching well above chance, not occasional correct examples.
- Use the 2022 year-to-post evaluation period for the contemporary models discussed. There is no common quarterly reporting period.
- Should be a piece of cake is rhetorical emphasis; the checkable part concerns actual model performance.
Counts as supported if: Contemporary Winoground evaluations show widespread failures or near-chance performance on matching captions and images whose distinctions depend on word order.
Counts as contradicted if: Representative contemporary models consistently match these caption-image pairs correctly, with strong performance across the dataset rather than isolated examples.
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 Winoground results directly support the claim in its stated context. Tested contemporary systems failed to reliably complete matching tasks requiring sensitivity to differently ordered captions. Every model was below chance on the complete group metric. Some performed above chance on caption selection, so the finding establishes unreliability rather than a total inability to distinguish word order.
Limitations
- The result applies to the evaluated models and task.
- Winoground failures can also involve visual perception and image-text alignment.
- Above-chance text scores should not be confused with reliable complete matching.
Evidence (1 item)
Supports the claim: arXiv; Thrush and colleagues,
below random chance
Why it matters here: The group score requires all relevant caption-image comparisons to be correct; every tested model scored below its random baseline. Some models exceeded chance on the less demanding text score, but remained far behind humans and did not reliably solve the complete matching task.
- Publisher
- arXiv; Thrush and colleagues
- 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, Did GoogleAI Just Snooker One of Silicon Valley’s Sharpest Minds?
This should be a piece of cake, but for currently popular systems, it’s not.
- Where
- paragraph 27
- Published
- 2022-09-14
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-did-googleai-just-snooker-one-of-c24cf3-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-did-googleai-just-snooker-one-of-c24cf3-a1.