“For example, OpenAI’s o1 is able to create and execute Python code that can do geometric calculation, but because it does not understand how objects are embedded in space and interact in space or, indeed, what spatial relationships actually mean, it often makes very basic mistakes.”
Our reading, the sentence that was checked:
OpenAI's o1 often makes very basic spatial mistakes because it does not understand how objects are embedded and interact in space.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, AI still lacks “common” sense, 70 years later, paragraph 19.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-ai-still-lacks-common-sense-70-years-da3ee2-a2
We recorded 15 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 2025-01-05. 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
- OpenAI’s o1
- able to create and execute Python code
- because it does not understand how objects are embedded in space and interact in space
- often makes very basic mistakes
Our interpretation
- The claim includes an explanation for the errors, not merely an observation that errors occur.
- Often means recurring across varied elementary spatial tasks; no numerical frequency is supplied.
- Spatial understanding is read behaviorally as representing and applying spatial relationships consistently.
- The current-state context is late 2024 through 2025-01-05.
- Results for o1-preview are relevant context but are not automatically results for the released o1 model. Code generation and actual execution must also be distinguished.
Counts as supported if: Controlled evaluations of the identified o1 version show recurring elementary spatial errors, and interventions or diagnostic tests connect those errors to deficient spatial representations despite effective geometric calculation.
Counts as contradicted if: Representative evaluations show dependable elementary spatial performance, or controlled evidence establishes that the cited errors arise from another mechanism rather than the asserted spatial deficit.
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
A study available September 30, 2024 reports spatial-planning limitations in o1-preview alongside improved constraint-following over GPT-4 [e1]. Davis's October 11 report describes September 2024 o1-preview tests that found improved performance but continuing spatial difficulties [e2]. These preview-model results do not establish how often released o1 made elementary spatial errors during October 1, 2024 through January 5, 2025, or whether deficient spatial representations caused such errors [e1, e2].
Limitations
- The cited studies do not establish the behavior of released o1 during the target period.
- The tasks do not establish a representative frequency of very basic spatial mistakes.
- The evidence does not isolate spatial representation from perception, task interpretation or calculation as the cause.
- Actual Python execution by the relevant o1 configuration remains unverified.
Evidence (2 items)
Context only: arXiv; Kevin Wang and colleagues,
spatially complex tasks.
Why it matters here: The study finds spatial planning limitations in o1-preview while also reporting improved constraint-following over GPT-4. It concerns a different version, predates the target interval and does not isolate the claimed cause.
- Where
- Abstract
- Publisher
- arXiv; Kevin Wang and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv; Ernest Davis,
spatial reasoning
Why it matters here: Davis reports improved but imperfect performance on 105 mathematics and science problems using o1-preview in September. This is an associated researcher's report, not independent confirmation of elementary-error frequency or the mechanism asserted for released o1.
- Where
- Abstract
- Publisher
- arXiv; Ernest Davis
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
Reproducible tests of the identified o1 version during the target period, documenting tool access and actual Python execution, measuring repeated elementary spatial errors, and using controls or interventions to distinguish deficient spatial representations from alternative causes.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, AI still lacks “common” sense, 70 years later
For example, OpenAI’s o1 is able to create and execute Python code that can do geometric calculation, but because it does not understand how objects are embedded in space and interact in space or, indeed, what spatial relationships actually mean, it often makes very basic mistakes.
- Where
- paragraph 19
- Published
- 2025-01-05
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-ai-still-lacks-common-sense-70-years-da3ee2-a2: 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-ai-still-lacks-common-sense-70-years-da3ee2-a2.