“Sooner or later, the exponential greed for data will exceed what is available.”
Our reading, the sentence that was checked:
The exponentially growing demand for training data will eventually exceed the data available.
Assessment: Unresolved No deadline stated on 2026-10-29 Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Breaking news: Scaling will never get us to AGI, paragraph 5.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-breaking-news-scaling-will-never-c46103-a2
We recorded 6 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-29. The source may have been edited after 2024-04-08. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Forecast
- Deadline, as the speaker wrote it
- “Sooner or later”
- Deadline state when assessed
- No deadline stated
- Deadline state on 2026-10-29
- No deadline stated (the release’s reference date)
From the speaker
- Sooner or later
- the exponential greed for data
- will exceed what is available
Our interpretation
- The ordinary reading is that growing training-data requirements will eventually encounter an insufficient supply of usable data.
- Sooner or later supplies no dated deadline.
- Usable public human-generated text is a measurable subset of the claim, not its explicitly stated entire scope.
- Data availability must account for quality, access, useful repetition and synthetic or other data sources. Reusing the same corpus across companies does not consume it like a physical resource.
Counts as supported if: Dated measurements show that relevant model-training requirements exceed the available usable data supply under matched definitions, and document an actual constraint rather than only a projected future intersection.
Counts as contradicted if: A general, validated account establishes that usable supply can keep pace with the relevant training requirements indefinitely. Continued sufficiency over a finite period alone does not contradict an eventual forecast.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Did the forecast come true?
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 deadline stated
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
Epoch's June 6, 2024 analysis projected full utilization of usable public human-generated text between 2026 and 2032 with an 80% confidence interval; it did not document an observed supply-demand crossing [e1]. An October 2025 study found conditional benefits from some mixtures of synthetic and natural data, without establishing indefinitely sufficient supply [e2]. TechCrunch's January 8, 2025 report quoted an exhaustion assertion but supplied no matched demand and supply measurements [e3]. These items establish neither an actual broad data constraint meeting the contract nor that usable supply can keep pace indefinitely [e1, e2, e3].
Limitations
- Epoch's quantitative estimate covers public human-generated text, not all usable training data [e1].
- The synthetic-data experiments concern particular mixtures, budgets and training conditions; they do not establish unlimited substitution for natural data [e2].
- The reported exhaustion assertion lacks a transparent inventory and a comparable demand estimate [e3].
Evidence (3 items)
Context only: Epoch AI,
between 2026 and 2032.
Why it matters here: Epoch projects a future constraint for quality- and repetition-adjusted public human text. This is a conditional forecast, not evidence that a crossing had occurred by the evaluation date, and its scope excludes other data sources.
- Publisher
- Epoch AI
- Kind of source
- Independent measurement
- Describes events of
- Public from
Context only: arXiv,
a potential path toward sidestepping these limitations.
Why it matters here: Controlled experiments find benefits from some mixtures of synthetic and natural data, with results depending on generation method and training conditions. This weakens a fixed-supply assumption without proving that supply will remain sufficient indefinitely.
- Where
- Abstract
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: TechCrunch,
there’s little real-world data left
Why it matters here: The report documents an assertion of exhaustion that could appear to settle the forecast. It supplies no transparent inventory or matched demand estimate, so it does not meet the early-resolution rule.
- Where
- Opening paragraph
- Publisher
- TechCrunch
- Kind of source
- News report
- Describes events of
- Public from
What would settle it
A documented post-publication crossing in which actual training requirements exceed usable supply under matched definitions, accounting for quality, access, repetition, synthetic data and substitution, would support the forecast. A general validated account showing that usable supply can keep pace indefinitely would contradict it; finite periods of sufficiency would not.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Breaking news: Scaling will never get us to AGI
Sooner or later, the exponential greed for data will exceed what is available.
- Where
- paragraph 5
- Published
- 2024-04-08
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-breaking-news-scaling-will-never-c46103-a2: 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-breaking-news-scaling-will-never-c46103-a2.