“There’s no reason to think that any such thing yet exists, nor that building such a comparison tool is any easier than solving AGI itself.”
Our reading, the sentence that was checked:
There is no reason to think that a technology yet exists for comparing an LLM's output with the text of a reliable source and determining with very high accuracy whether they say the same thing, nor that building such a comparison tool is any easier than solving AGI.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, No, RAG is probably not going to rescue the current situation, paragraph 26.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-no-rag-is-probably-not-going-to-rescue-b72e11-a1
We recorded 11 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 2024-02-23. 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
- there is no reason to think (qualifier, stated by the speaker)
From the speaker
- no reason to think
- any such thing yet exists
- very high accuracy
- any easier than solving AGI itself
Our interpretation
- The checkable component concerns public evidence existing by February 23, 2024 for highly accurate comparison of generated text with a reliable source.
- Yet exists refers to accumulated evidence by publication, so the historical interval has no specified starting date.
- Factual-consistency and entailment evaluators are relevant evidence, but their existence alone does not establish very high accuracy at broad semantic comparison.
- No reason to think is an assertion about the evidence, not a conditional forecast antecedent.
- The comparison with solving AGI supplies neither an operational definition of AGI nor a measurable standard of difficulty.
Counts as supported if: Contemporaneous comprehensive evaluations establish that available comparison methods remain materially unreliable for the intended broad task. Supporting the additional AGI comparison would require evidence connecting that task's difficulty to a defined general-intelligence problem.
Counts as contradicted if: Public prepublication research validates a broadly applicable comparator at very high accuracy, or establishes that the specified comparison capability is attainable without the claimed AGI-level difficulty.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
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 horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
Source-comparison technology and substantial evaluation evidence existed before the post. That makes a reading that no relevant technology existed untenable. But Marcus qualified the technology by very high accuracy, without defining its threshold or scope. The studies show progress and remaining errors; neither establishes the full qualified claim. They also do not settle his comparison with solving AGI.
Limitations
- The compound claim combines an empirical existence question with an undefined difficulty comparison.
- Factual consistency is related to, but not always identical to, semantic equivalence.
- Neither an absent qualifying result nor existing imperfect tools establishes AGI-level difficulty.
Evidence (2 items)
Context only: arXiv; Zha and colleagues, UC San Diego,
information alignment
Why it matters here: A functioning comparator and extensive evaluations existed before the post, providing material evidence of progress. The authors nevertheless report English-only coverage, false positives and false negatives, and limits on appropriate use; this does not establish Marcus's unspecified broad accuracy standard.
- Publisher
- arXiv; Zha and colleagues, UC San Diego
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv; Honovich and colleagues,
domain-specific challenges
Why it matters here: TRUE documents useful results from entailment and question-answering methods alongside residual errors and evaluation-label problems. It supports caution about reliability but supplies no equivalence between comparison technology and solving AGI.
- Publisher
- arXiv; Honovich and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
A prepublication evaluation demonstrating very high accuracy across the intended comparison tasks would resolve the empirical component. The AGI comparison additionally requires defined capabilities and a testable meaning of relative difficulty.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, No, RAG is probably not going to rescue the current situation
There’s no reason to think that any such thing yet exists, nor that building such a comparison tool is any easier than solving AGI itself.
- Where
- paragraph 26
- Published
- 2024-02-23
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-no-rag-is-probably-not-going-to-rescue-b72e11-a1: 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-no-rag-is-probably-not-going-to-rescue-b72e11-a1.