“If the juicing of benchmarks and “blending” of test data is the primary scandal, there is a secondary scandal too: a bunch of well-paid pundits, such as Roose, Newton, and Cowen, are failing to reckon with the repeated failures of massive companies to build GPT-5.”
Our reading, the sentence that was checked:
A number of well-paid pundits, such as Roose, Newton and Cowen, are failing to reckon with the repeated failures of massive companies to build GPT-5, which Marcus calls a secondary scandal next to benchmark juicing.
Assessment: Contradicted Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Deep Learning, Deep Scandal, paragraph 9.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-deep-learning-deep-scandal-d85109-a1
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 2025-04-07. 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
- such as Roose, Newton, and Cowen (scope, stated by the speaker)
- framed as a secondary scandal, if benchmark juicing is the primary one (qualifier, stated by the speaker)
From the speaker
- such as Roose, Newton, and Cowen
- are failing to reckon with
- repeated failures of massive companies to build GPT-5
- secondary scandal
Our interpretation
- The ordinary factual reading is that the named commentators have not substantively addressed reported next-generation model setbacks and their implications for expectations about AI progress.
- Reckoning with setbacks means considering them and explaining their implications. It does not require agreeing with Marcus or lengthening an AGI timeline.
- The surrounding reference to GPT-5-level models makes GPT-5 a description of an expected generation of capability, rather than exclusively an OpenAI product name.
- The target is the public commentary record available through April 7, 2025, the latest observable state before publication. No starting date is supplied, so the start remains unspecified.
- The scandal characterization is an opinion. The introductory comparison with benchmark misconduct is rhetorical, not a conditional forecast.
Counts as supported if: A sufficiently complete record of the named commentators' relevant coverage shows that they consistently omitted substantive consideration of reported model setbacks and their implications.
Counts as contradicted if: Direct examples show that the named commentators substantively considered the reported setbacks and explained their implications, contradicting the characterization of those commentators as failing to reckon with them.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Contradicted
The dated evidence goes against 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
On the ordinary reading of substantive engagement, the characterization is contradicted. Newton examined reported setbacks at three major laboratories, possible limits of LLMs and their implications for superintelligence. Cowen acknowledged slowdown reports and explained why his outlook remained optimistic. These are direct counterexamples involving two people Marcus named. Their conclusions need not agree with Marcus to count as reckoning with the evidence. This finding does not establish that all three addressed every contamination concern or revised numerical AGI timelines.
Limitations
- The result depends on reading reckoning as substantive consideration, rather than agreement about how much weight setbacks deserve.
- The assessment concerns the named examples, not every member of an unspecified pundit group.
- The underlying technical truth of the model-setback reports is not independently adjudicated here.
Evidence (2 items)
Goes against the claim: Platformer — Casey Newton,
had underperformed expectations
Why it matters here: Newton explicitly covered setbacks at OpenAI, Google and Anthropic, then examined data limits, benchmark gaming, the possible insufficiency of LLMs, and alternative routes to progress. This is substantive engagement even though he remained optimistic.
- Where
- Section I, paragraph introducing the Bloomberg report; implications discussed in sections II and III
- Publisher
- Platformer — Casey Newton
- Kind of source
- News report
- Describes events of
- Public from
Goes against the claim: Marginal Revolution — Tyler Cowen,
I remain bullish on AI progress,
Why it matters here: Cowen considered the slowdown reports and explained that his outlook rested on the research environment rather than confidence in scaling laws. This demonstrates consideration of their implications, although it does not provide a numerical AGI timeline.
- Publisher
- Marginal Revolution — Tyler Cowen
- 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, Deep Learning, Deep Scandal
If the juicing of benchmarks and “blending” of test data is the primary scandal, there is a secondary scandal too: a bunch of well-paid pundits, such as Roose, Newton, and Cowen, are failing to reckon with the repeated failures of massive companies to build GPT-5.
- Where
- paragraph 9
- Published
- 2025-04-07
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-deep-learning-deep-scandal-d85109-a1: Current. Outcome: Contradicted. 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-deep-learning-deep-scandal-d85109-a1.