“Gary Marcus is astonished that some people still haven’t recognized that generative AI has struggled mightily when it comes to world models, high level reasoning and factuality, even as evidence for that conclusion continues to pour in, even with data of ever-greater magnitude and sophistication..”
Our reading, the sentence that was checked:
Generative AI has struggled mightily with world models, high-level reasoning and factuality, even with ever more data.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Sora Can’t Handle the Truth, paragraph 21.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-sora-cant-handle-the-truth-cfa367-a1
We recorded 11 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 2024-02-17. 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
- even with ever-greater data (qualifier, stated by the speaker)
From the speaker
- generative AI has struggled mightily
- world models, high level reasoning and factuality
- even with data of ever-greater magnitude and sophistication
Our interpretation
- The plain reading is that substantial, repeatable difficulties remained across these capabilities despite growth and refinement in training data.
- The claim allows substantial improvements and successful tasks; it does not say that every model always fails or that scaling provides no benefit.
- World-model difficulties include failures to represent ordinary objects and spatial relationships coherently; high-level reasoning includes composing multiple steps correctly.
- This is a retrospective observation through February 17, 2024. No starting date is supplied, so the target period has an open beginning.
- Despite more data describes coexistence, not a controlled causal claim that additional data cannot help.
Counts as supported if: Evaluations available for the historical period document substantial recurring failures in world consistency, multistep reasoning, and factuality, alongside evidence of increased or refined training data.
Counts as contradicted if: Broad, credible evaluations establish that the specified difficulties had been substantially overcome across representative generative systems by publication.
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
Contemporary studies document recurring spatial-consistency and multistep-reasoning failures, while model reports acknowledge persistent factual errors. Training data had also grown and been refined. This supports the ordinary reading that substantial difficulties remained. Documented improvements prevent extending that finding into a claim that scaling achieved nothing or that generative AI lacked every relevant capability.
Limitations
- The severity term is qualitative, so the outcome depends partly on the stated reading.
- Evidence covers selected models and tasks, not every generative system.
- The assessment concerns capabilities before the post, not their status in 2026.
Evidence (4 items)
Supports the claim: arXiv; Gokhale and colleagues,
poor spatial
Why it matters here: The evaluation reports substantial failures to generate requested object relationships despite strong visual quality. This supports difficulties with spatial world consistency, without proving the absence of internal world models.
- Publisher
- arXiv; Gokhale and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Supports the claim: arXiv; Dziri and colleagues,
failures on surprisingly trivial problems
Why it matters here: Experiments on multiplication, logic puzzles, and dynamic programming document sharply worsening performance with task complexity. These are repeatable multistep reasoning difficulties, rather than isolated anecdotes.
- Publisher
- arXiv; Dziri and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Supports the claim: OpenAI,
makes reasoning errors
Why it matters here: OpenAI acknowledges continuing factual and reasoning errors in GPT-4. The same section reports improved factuality relative to GPT-3.5, showing that persistent limitations coexisted with progress.
- Where
- Limitations
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: Meta; Touvron and colleagues, arXiv,
with a larger size
Why it matters here: The report documents 40% more pretraining tokens, revised data selection, and remaining hallucination risks. It also reports improvements, which limit any interpretation that greater scale or refinement brought no benefit.
- Publisher
- Meta; Touvron and colleagues, arXiv
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Sora Can’t Handle the Truth
Gary Marcus is astonished that some people still haven’t recognized that generative AI has struggled mightily when it comes to world models, high level reasoning and factuality, even as evidence for that conclusion continues to pour in, even with data of ever-greater magnitude and sophistication..
- Where
- paragraph 21
- Published
- 2024-02-17
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-sora-cant-handle-the-truth-cfa367-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-sora-cant-handle-the-truth-cfa367-a1.