“It is precisely because Sora is not a physics engine that a large fraction of Sora's videos are inconsistent with physics or biology, in very basic ways.”
Our reading, the sentence that was checked:
Because Sora is not a physics engine, a large fraction of its videos are inconsistent with physics or biology in basic ways.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, No, Sora has not “learned physics”, paragraph 10.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-no-sora-has-not-learned-physics-4b17c9-a1
The wording above is the version this site observed on 2026-10-29. The source may have been edited after 2024-03-02. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Causal claim
- 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
- precisely because Sora is not a physics engine
- a large fraction of Sora's videos
- inconsistent with physics or biology
- in very basic ways
Our interpretation
- The target is the original Sora preview discussed in the post.
- A large fraction means a substantial share of representative outputs, beyond isolated failures. The speaker gives no numerical cutoff.
- Basic inconsistencies include unintended violations of object continuity, ordinary physical interactions, or anatomical structure.
- The claim requires evidence of both prevalence and causation. A collection of failures does not establish either the population rate or its cause.
- The target period is the public preview through the post date because no separate observation window is stated.
Counts as supported if: A representative evaluation finds a substantial share of basic physical or biological failures in the original Sora, and controlled evidence attributes those failures to the absence of physics-engine computation.
Counts as contradicted if: A representative evaluation finds only isolated basic failures, or controlled evidence rejects the claimed causal explanation.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Does the evidence support the cause?
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
Basic failures are documented, and a contemporaneous study also finds geometric consistency in selected clips. Neither establishes what fraction of the original Sora's outputs violated physics or biology. The evidence also does not isolate the absence of a physics engine as the cause. Examples support the existence of failures, but not the full prevalence-and-cause claim.
Limitations
- Public demonstrations are selected samples.
- Geometric consistency measures do not test every physical or biological constraint.
- Later Sora evaluations cannot be substituted for an evaluation of the original preview.
Evidence (2 items)
Context only: OpenAI,
spontaneously appear
Why it matters here: This documents a basic continuity failure in the original preview. It supplies neither the fraction of outputs affected nor evidence that lacking a physics engine causes the failure.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: Xuanyi Li and colleagues, arXiv,
stability and consistency
Why it matters here: The authors found geometric consistency in selected publicly available Sora clips. This counterbalances a wholly failure-based account, but the selected sample and geometric measures cannot determine the overall physical or biological failure rate.
- Publisher
- Xuanyi Li and colleagues, arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
A representative evaluation of the original Sora with a clear error rubric and denominator, accompanied by controlled evidence testing whether physics-engine computation explains the failures.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, No, Sora has not “learned physics”
It is precisely because Sora is not a physics engine that a large fraction of Sora's videos are inconsistent with physics or biology, in very basic ways.
- Where
- paragraph 10
- Published
- 2024-03-02
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-no-sora-has-not-learned-physics-4b17c9-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-no-sora-has-not-learned-physics-4b17c9-a1.