“It’s not just OpenAI either; it’s the whole field, as I discussed in my own recent Fortune article:”
Our reading, the sentence that was checked:
Diminishing returns from scaling affect the whole AI field, not just OpenAI.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, GPT 4.5 is no GPT-5, paragraph 11.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-gpt-45-is-no-gpt-5-20ad04-a1
We recorded 14 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 2025-02-26. 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
- scaling just isn’t producing the fruit that it once did
- not just OpenAI
- the whole field
Our interpretation
- The ordinary reading is that additional pretraining scale was producing smaller capability gains across major frontier developers.
- The surrounding discussion concerns frontier generative models, rather than every application of AI.
- I interpret 'whole field' as a widespread industry trend, not a claim that every laboratory experienced identical results.
- The comparison concerns the current frontier-model development cycle, operationalized as November 2024 through publication. This reporting window is evaluator-supplied.
- Disappointing expectations or delayed releases alone do not establish declining returns per unit of scale.
Counts as supported if: Comparable results or well-substantiated reporting establish that additional pretraining scale yielded smaller capability gains across several major developers than in their preceding model generations.
Counts as contradicted if: Comparable evidence across major developers establishes that returns to additional pretraining scale remained broadly comparable to, or exceeded, those of preceding generations.
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
There were credible reports of difficulties beyond OpenAI. That supports the narrower statement that several labs faced disappointing results. It does not establish the claimed field-wide decline in returns from scaling. Company disagreement and DeepSeek's technical progress also prevent treating the reports as conclusive. These competing signals do not establish a mixed outcome because they do not measure the same scaling relationship.
Limitations
- Important underlying experiments were confidential.
- Reporting is used because comparable primary measurements were not found.
Evidence (2 items)
Context only: TIME,
leading AI companies are experiencing diminishing returns
Why it matters here: This corroborates that slowdown reports extended beyond OpenAI. The same article records Anthropic's disagreement and uncertainty about whether high expectations explain the reported disappointments; it supplies no comparable scaling curves.
- Publisher
- TIME
- Kind of source
- News report
- Describes events of
- Public from
Context only: DeepSeek-AI / arXiv,
performance comparable to leading closed-source models
Why it matters here: The report documents substantial contemporary progress outside the named US labs. It combines architectural and training changes, so it neither isolates pretraining scale nor establishes whether its marginal returns had declined.
- Where
- Abstract
- Publisher
- DeepSeek-AI / arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
Comparable training-resource and capability results for successive generations at several major labs, or sufficiently detailed independent reporting of those experiments.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, GPT 4.5 is no GPT-5
It’s not just OpenAI either; it’s the whole field, as I discussed in my own recent Fortune article:
- Where
- paragraph 11
- Published
- 2025-02-26
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-gpt-45-is-no-gpt-5-20ad04-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.