“Returns are starting to diminish.”
Our reading, the sentence that was checked:
Returns from scaling generative AI are starting to diminish.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, The need for a President that speaks AI natively, paragraph 5.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-the-need-for-a-president-that-speaks-6091ac-a1
We recorded 11 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2024-06-28. 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
- Returns
- are starting to diminish
Our interpretation
- The preceding sentence makes this a claim about capability returns from scaling, rather than company profits or investment returns.
- The ordinary test is whether recent increases in training resources produced smaller capability gains per additional unit of resources than earlier increases.
- Use 2023 through June 28, 2024 as the comparison period for the asserted emerging trend, with earlier results serving as baseline context.
- A selected benchmark or an extrapolated scaling curve is insufficient by itself to establish the broad contemporary trend.
- Do not require proof of a sudden breakpoint or complete cessation of progress.
Counts as supported if: Comparable evaluations and training-resource measurements show shrinking marginal capability gains across successive scales or generations around publication, with sufficient breadth to support the general claim.
Counts as contradicted if: Comparable measurements show stable or increasing marginal capability gains across the relevant scales and period, rather than a broad diminishing trend.
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
CSET provides relevant evidence of diminishing returns on a programming measure, so the claim has empirical support in a limited setting. But that analysis does not establish that capability returns generally were shrinking across frontier systems around publication. The GPT-4 report shows predictable gains, which are compatible with diminishing returns. The available comparisons do not settle the broader statement.
Limitations
- A narrower claim about the particular programming curve would be easier to support.
- Training-budget assumptions, training loss and broad capability are different measures.
- Unresolved does not mean the evidence shows accelerating returns.
Evidence (2 items)
Context only: Georgetown University Center for Security and Emerging Technology,
diminishing marginal returns
Why it matters here: CSET's reanalysis supports diminishing returns on a programming measure using an assumed training-budget conversion. Its limited task coverage and extrapolations do not establish a broad trend across frontier releases around June 2024.
- Publisher
- Georgetown University Center for Security and Emerging Technology
- Kind of source
- Independent measurement
- Describes events of
- Public from
Context only: OpenAI,
predictable scaling
Why it matters here: The report documents predictable scaling results for GPT-4, while withholding important training details. Predictable improvement can coexist with diminishing marginal returns, so this does not contradict the claim or establish its asserted contemporary trend.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
What would settle it
A consistent series of model evaluations and training-resource measurements covering successive scales through June 2024, showing whether marginal gains shrank across a representative set of capabilities.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, The need for a President that speaks AI natively
Returns are starting to diminish.
- Where
- paragraph 5
- Published
- 2024-06-28
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-the-need-for-a-president-that-speaks-6091ac-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-the-need-for-a-president-that-speaks-6091ac-a1.