“The hypothesis I have been floating since 2022 that scaling data alone would not be enough for robust AI has become increasingly prevalent.”
Our reading, the sentence that was checked:
Marcus's hypothesis, which he has been floating since 2022, that scaling data alone would not be enough for robust AI has become increasingly prevalent.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, AGI isn’t coming in 2025, and GPT-5 probably isn’t either., paragraph 7.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-agi-isnt-coming-in-2025-and-gpt-5-439e22-a2
We recorded 8 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 2025-01-09. 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
- hypothesis floated since 2022 (qualifier, stated by the speaker)
From the speaker
- hypothesis I have been floating since 2022
- scaling data alone
- would not be enough for robust AI
- has become increasingly prevalent
Our interpretation
- The claim concerns the spread of a view, not whether that view has been scientifically proved.
- The plain reading is that skepticism about achieving reliable, broadly capable AI simply by adding training data had gained visible support among researchers and industry figures.
- The comparison runs from 2022 through the January 9, 2025 post. This is a historical trend, rather than a financial-style observation requiring a latest reporting quarter.
- Prevalence means broader public endorsement, not majority agreement. Marcus supplies no numerical threshold or population.
- Statements about pretraining limits count when they question the sufficiency of that approach. Mere reports of slower improvements do not establish agreement with every part of Marcus's robustness hypothesis.
- The reference to advocacy since 2022 is a historical component to verify, not a conditional forecast.
Counts as supported if: Independent records confirm Marcus advanced the position in 2022, and contemporaneous reporting or public statements document a subsequent broadening of similar skepticism by January 9, 2025.
Counts as contradicted if: Reliable historical records refute the advocacy timeline, or comparable evidence shows that endorsement of the view remained unchanged or declined. Continued disagreement alone would not contradict an increase in prevalence.
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
Big Think's December 22, 2024 report confirms that Marcus argued in 2022 that deep learning faced limits in reasoning and reliability [e1]. In November 2024, Reuters reported prominent scientists questioning pretraining scaling, including Sutskever, and TIME described related skepticism as gaining ground [e2, e3]. These reports support a qualitative increase in public support for doubts about scaling alone before January 9, 2025 [e1, e2, e3]. Amodei continued to express confidence in scaling in November 2024, showing that the view remained disputed [e4].
Limitations
- The evidence does not measure prevalence in a defined population or establish majority agreement.
- Doubts about pretraining limits do not establish endorsement of Marcus's precise robustness hypothesis or prove that hypothesis correct.
- The Reuters article was accessed through an attributed reproduction.
Evidence (4 items)
Supports the claim: Big Think,
Marcus wrote an article titled “Deep Learning Is Hitting a Wall.”
Why it matters here: Independent reporting confirms the 2022 advocacy timeline. The surrounding discussion identifies the argument as concerning reliability, reasoning and understanding, rather than a claim that all technical improvement would stop.
- Publisher
- Big Think
- Kind of source
- News report
- Describes events of
- Public from
Supports the claim: Reuters, reproduced by China Intellectual Property Lawyers Network,
“bigger is better” philosophy.
Why it matters here: Reuters explicitly reports a change in public positions and identifies Sutskever, previously a scaling advocate, as saying pretraining gains had plateaued. This supports broader skepticism about pretraining alone, although it does not establish agreement with Marcus's complete theory of robustness.
- Publisher
- Reuters, reproduced by China Intellectual Property Lawyers Network
- Kind of source
- News report
- Describes events of
- Public from
Supports the claim: TIME,
has gained ground among some
Why it matters here: The passage concerns proposed limits from current techniques or scarce fresh data. Nearby interviews with Sasha Luccioni and Ege Erdil document related doubts about information limits and scaling alone.
- Publisher
- TIME
- Kind of source
- News report
- Describes events of
- Public from
Context only: Lex Fridman Podcast; statement by Anthropic CEO Dario Amodei,
no ceiling below the level of humans.
Why it matters here: Amodei continued to express confidence in scaling, while also allowing for new training methods. This limits any inference of consensus but does not contradict an increase in skeptical views.
- Publisher
- Lex Fridman Podcast; statement by Anthropic CEO Dario Amodei
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, AGI isn’t coming in 2025, and GPT-5 probably isn’t either.
The hypothesis I have been floating since 2022 that scaling data alone would not be enough for robust AI has become increasingly prevalent.
- Where
- paragraph 7
- Published
- 2025-01-09
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-agi-isnt-coming-in-2025-and-gpt-5-439e22-a2: 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-agi-isnt-coming-in-2025-and-gpt-5-439e22-a2.