“Large Language Models — which are trained by feeding them massive amounts of training data and then reinforcing their understanding through further training runs — are hitting the point of diminishing returns.”
Our reading, the sentence that was checked:
LLMs are hitting diminishing returns from scaling data and training.
Assessment: Unresolved Evidence as of
Where and when Ed Zitron wrote it
Where's Your Ed At, There Is No AI Revolution, paragraph 79.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment wyea-wheres-the-money-a86f32-a1
We recorded 42 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 2025-02-24. 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
- Large Language Models
- hitting the point of diminishing returns
- more compute and more data while pretraining
Our interpretation
- The ordinary reading is that increasing pretraining resources was producing smaller additional improvements in model performance. It does not assert that improvements had stopped.
- The surrounding discussion concerns recent frontier training runs, including December 2024 reporting. I assess the latest training cycle described, from the final quarter of 2024 through the post date.
- A longstanding mathematical pattern of diminishing marginal gains would not alone establish the suggested recent change in frontier progress.
Counts as supported if: Comparable scaling experiments from the described training cycle show progressively smaller performance improvements from additional pretraining compute and data across representative LLMs.
Counts as contradicted if: Comparable experiments show that marginal performance gains remained stable or increased across the relevant frontier scales.
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
Contemporary reporting supports the existence of concerns and disappointing runs. It does not establish a general measured decline in returns across LLMs. Earlier public experiments found predictable scaling, which also does not disprove diminishing marginal gains. The available records do not settle the claim at its stated breadth.
Limitations
- Frontier training details were not supplied in the cited reporting.
- The interpretation of diminishing returns materially affects the test.
Evidence (2 items)
Context only: arXiv,
accurately predict
Why it matters here: These experiments found predictable scaling with additional training. Predictability can coexist with diminishing marginal gains, and these earlier, smaller experiments do not settle the contemporary frontier claim.
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: Reuters, reproduced by China Intellectual Property Lawyers Network,
“Scaling the right thing matters more now than ever.”
Why it matters here: The report describes Sutskever's concerns about pretraining progress and accounts of disappointing training runs. It supplies contemporary testimony, not the resource and performance measurements required to establish a general scaling trend.
- Publisher
- Reuters, reproduced by China Intellectual Property Lawyers Network
- Kind of source
- News report
- Describes events of
- Public from
What would settle it
Comparable training-resource and performance measurements for the frontier runs discussed, with a stated definition of marginal return.
Where it was said (1)
Said once in the sources this register covers.
: Where's Your Ed At, There Is No AI Revolution
Large Language Models — which are trained by feeding them massive amounts of training data and then reinforcing their understanding through further training runs — are hitting the point of diminishing returns.
- Where
- paragraph 79
- Published
- 2025-02-24
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment wyea-wheres-the-money-a86f32-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.