“They are merely empirical generalizations that held for a certain period time, when there was enough fresh data and compute.”
Our reading, the sentence that was checked:
Scaling laws are merely empirical generalizations that held for a limited period while enough fresh data and compute were available.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, An AI rumor you won’t want to miss, paragraph 2.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-an-ai-rumor-you-wont-want-to-miss-db6cf3-a1
We recorded 6 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 2024-11-10. 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
- when there was enough fresh data and compute (qualifier, stated by the speaker)
From the speaker
- merely empirical generalizations
- held for a certain period time
- when there was enough fresh data and compute
Our interpretation
- I assess the status of scaling research through the post date, 2024-11-10. The wording supplies no starting date.
- The ordinary reading is that scaling laws describe observed relationships within tested resource and data conditions, without guaranteeing their indefinite continuation.
- I do not treat this sentence as an unqualified declaration that scaling had already stopped working. The surrounding text describes that possibility as a suspicion.
- Empirical does not mean that no theoretical explanation exists.
Counts as supported if: Published research identifies scaling relationships through empirical measurements over bounded ranges and documents limits associated with data or compute conditions.
Counts as contradicted if: Research establishes that the relevant relationships are unconditional guarantees beyond their tested regimes, rather than empirical regularities.
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
The research supports the empirical and conditional character of scaling laws. Kaplan and colleagues describe measured loss relationships; data-constrained experiments show that repeated data can exhaust the value of additional compute. Support applies to this methodological reading. It does not establish that scaling had already ended, or that the relationships lacked theoretical explanations.
Limitations
- The result depends on reading the sentence in light of the surrounding hedges about scaling having stopped.
- Loss scaling does not establish equivalent improvements in reasoning or factual reliability.
- Theoretical explanations existed before the post.
Evidence (3 items)
Supports the claim: arXiv; Kaplan et al.,
empirical scaling laws
Why it matters here: The authors explicitly characterize their loss relationships as empirical. Their reported measurements cover finite ranges of model size, data and training compute.
- Publisher
- arXiv; Kaplan et al.
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Supports the claim: arXiv; Muennighoff et al.,
the value of adding compute
Why it matters here: Controlled experiments show limits to additional compute when training repeatedly on constrained data. They support the importance of data conditions, without establishing an industry-wide date when scaling ceased.
- Publisher
- arXiv; Muennighoff et al.
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv; Sharma and Kaplan,
can be explained
Why it matters here: Scaling laws already had proposed theoretical explanations. This qualifies merely, although a conditional explanation does not turn measured scaling relationships into unconditional guarantees.
- Publisher
- arXiv; Sharma and Kaplan
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, An AI rumor you won’t want to miss
They are merely empirical generalizations that held for a certain period time, when there was enough fresh data and compute.
- Where
- paragraph 2
- Published
- 2024-11-10
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-an-ai-rumor-you-wont-want-to-miss-db6cf3-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-an-ai-rumor-you-wont-want-to-miss-db6cf3-a1.