“The study Mollick linked to doesn’t actually show what he claims. If you read it carefully, it says literally nothing about capabilities improving.”
Our reading, the sentence that was checked:
The study Mollick cited does not show that LLM capabilities are improving; it only addresses compute efficiency.
Assessment: Mixed Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Evidence that LLMs are reaching a point of diminishing returns - and what that might mean, paragraph 2.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-evidence-that-llms-are-reaching-a-458f77-a1
We recorded 20 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 2024-04-13. 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
- doesn’t actually show what he claims
- literally nothing about capabilities improving
- computational resources they require to get to given level of performance
Our interpretation
- Assess the study available when Marcus wrote, including its distinction between compute efficiency and performance.
- Separate two assertions: the study does not establish capability doubling every 5–14 months; the study says nothing about capability improvement.
- The target period covers the identified study's publication through the post date, rather than the years covered by its dataset.
Counts as supported if: The reported doubling concerns compute efficiency at fixed performance, without establishing a corresponding doubling of capabilities. Full support additionally requires that the study neither discusses nor analyzes capability or performance improvements.
Counts as contradicted if: The study explicitly discusses or analyzes improvements in language-model performance or capabilities, or actually establishes the capability-doubling result attributed to it.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Mixed
Part of the claim holds and part does not, and the record says which part is which.
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
Marcus correctly distinguishes compute-efficiency doubling from capability doubling. The study does not establish Mollick's stated doubling rate for capabilities. But 'literally nothing about capabilities improving' overstates the criticism: the authors explicitly analyze performance gains and describe their work as quantifying progress in language-model capabilities.
Limitations
- Language-model performance is narrower than general capability.
- Reading 'literally nothing' solely as shorthand for 'no evidence of capability doubling' would produce a more favorable assessment.
Evidence (2 items)
Supports the claim: Anson Ho and colleagues, arXiv,
We want to estimate the rate at which newer language models are able to achieve a certain level of performance more efficiently than older models.
Why it matters here: The study measures efficiency at a given performance level. Its compute-halving estimate does not establish that capabilities themselves doubled at the same rate.
- Publisher
- Anson Ho and colleagues, arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Goes against the claim: Epoch AI, study authors,
Nevertheless, our work provides a rigorous quantitative analysis of the remarkable rate of progress in language modeling capabilities over the past decade.
Why it matters here: The authors explicitly describe an analysis of improving capabilities. Their summary also explains the study's decomposition of performance gains into compute scaling and algorithmic contributions, contradicting the assertion that it addresses only efficiency.
- Publisher
- Epoch AI, study authors
- Kind of source
- Secondary summary
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Evidence that LLMs are reaching a point of diminishing returns - and what that might mean
The study Mollick linked to doesn’t actually show what he claims. If you read it carefully, it says literally nothing about capabilities improving.
- Where
- paragraph 2
- Published
- 2024-04-13
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-evidence-that-llms-are-reaching-a-458f77-a1: Current. Outcome: Mixed. 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-evidence-that-llms-are-reaching-a-458f77-a1.