“In 2019 I pointed out through a series of examples on Twitter with GPT-2 that LLMs were subject to essentially the same critique: greedy with respect to data, uninterpretable, unreliable as reasoners, subject to hallucinations, etc, that they lacked world models (“they don't develop robust representations of *how events unfold over time*), and that they would be inadequate on their own.”
Our reading, the sentence that was checked:
In 2019, using a series of GPT-2 examples on Twitter, Marcus pointed out that LLMs were greedy with respect to data, uninterpretable, unreliable as reasoners, and subject to hallucinations, that they lacked world models, and that they would be inadequate on their own.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Open letter responding to Yann LeCun, paragraph 10.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-open-letter-responding-to-yann-lecun-1d7afe-a2
We recorded 5 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 2024-06-04. 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
- in 2019, through examples with GPT-2 on Twitter (scope, stated by the speaker)
From the speaker
- In 2019
- a series of examples on Twitter with GPT-2
- greedy with respect to data
- uninterpretable
- unreliable as reasoners
- subject to hallucinations
- they lacked world models
- they would be inadequate on their own
Our interpretation
- This is a claim about what Marcus publicly argued in 2019, not whether those technical arguments were correct.
- The series must substantively express all the listed criticisms. It need not use exactly the terminology used in 2024.
- The prediction of inadequacy is assessed as something allegedly stated in 2019; its eventual truth is outside this claim.
Counts as supported if: Dated records establish that Marcus's 2019 GPT-2 Twitter examples expressed the listed concerns about data requirements, interpretability, reasoning, hallucinations, world models and standalone inadequacy.
Counts as contradicted if: A sufficiently complete contemporaneous record establishes that the attributed series occurred outside 2019 or materially misrepresents what Marcus argued. Failure to retrieve a tweet is insufficient.
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
On October 28, 2019, LeCun attributed to Marcus the view that machines trained to predict missing words “don't learn much about quantities” [e1]. This records a narrow attribution about quantities; it does not establish an explicit general criticism of reasoning reliability or the full set of criticisms in the claimed GPT-2 Twitter series [e1]. The reply is insufficient to establish or contradict Marcus's complete historical account [e1].
Limitations
- The reply does not document the boundaries or full contents of the claimed 2019 GPT-2 Twitter series [e1].
- It does not establish the listed criticisms about data requirements, interpretability, hallucinations, world models or standalone inadequacy [e1].
- It does not establish whether the technical criticisms were correct [e1].
Evidence (1 item)
Supports the claim: Yann LeCun, Twitter,
don't learn much about quantities.
Why it matters here: LeCun independently acknowledges a contemporaneous criticism concerning quantities. This supports part of the historical account, but does not establish the other listed criticisms or their expression through a series of GPT-2 examples.
- Publisher
- Yann LeCun, Twitter
- Kind of source
- Secondary summary
- Describes events of
- Public from
What would settle it
Dated contemporaneous records or independently preserved quotations connecting each listed criticism to Marcus's 2019 GPT-2 Twitter series would support the account. A sufficiently complete contemporaneous record showing a material mismatch in content or date would contradict it. Failure to retrieve tweets would not settle it.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Open letter responding to Yann LeCun
In 2019 I pointed out through a series of examples on Twitter with GPT-2 that LLMs were subject to essentially the same critique: greedy with respect to data, uninterpretable, unreliable as reasoners, subject to hallucinations, etc, that they lacked world models (“they don't develop robust representations of *how events unfold over time*), and that they would be inadequate on their own.
- Where
- paragraph 10
- Published
- 2024-06-04
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-open-letter-responding-to-yann-lecun-1d7afe-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-open-letter-responding-to-yann-lecun-1d7afe-a2.