“We don’t know how kids do it; we do know that what we are doing with machines currently isn’t really working.”
Our reading, the sentence that was checked:
Current machine approaches to language acquisition are not really working.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Learning Language is Harder Than You Think, paragraph 40.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-learning-language-is-harder-than-b73ac2-a1
We recorded 5 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-29. The source may have been edited after 2022-07-20. 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)
From the speaker
- what we are doing with machines currently
- isn’t really working
Our interpretation
- Working refers to the immediately preceding goal of interpreting sentences and expressing intended meanings with general language competence.
- The claim is not read as denying every useful application or every improvement on a language benchmark.
- The separate statement about how children learn is outside the normalized proposition being assessed.
- Currently refers to approaches existing on July 20, 2022. With no regular reporting cycle, I use current-year evaluations through that date.
Counts as supported if: Convergent evaluations establish that contemporary approaches failed to achieve reliable general interpretation and expression of meanings.
Counts as contradicted if: A contemporary approach demonstrably achieves reliable general interpretation and expression of meanings across varied unfamiliar situations.
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
Contemporary approaches had important failures, but also measurable successes. Structural generalization tests sharply separated weaker sequence models from successful compositional parsers, and LaMDA improved dialogue performance. These findings defeat a reading that nothing useful was working. They do not settle the contextual claim that general interpretation and expression of meaning had not been achieved.
Limitations
- The selected tasks cover components of language competence.
- Progress does not by itself establish completion of the stated goal.
- Observed failures do not establish that every contemporary approach failed.
Evidence (2 items)
Context only: arXiv; Weißenhorn et al.,
low accuracies
Why it matters here: The comparison identifies a substantial weakness in sequence-to-sequence models while showing that other machine approaches succeed on the same structural tests. Neither result establishes general bidirectional language competence.
- Publisher
- arXiv; Weißenhorn et al.
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv; Thoppilan et al.,
significantly improves performance
Why it matters here: LaMDA's evaluations report improvements in dialogue quality, safety, and factual grounding, while retaining gaps from humans on some measures. This establishes progress without demonstrating the full language-acquisition goal.
- Publisher
- arXiv; Thoppilan et al.
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
A representative evaluation of July 2022 approaches against an explicit standard for reliable general comprehension and meaning-preserving generation.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Learning Language is Harder Than You Think
We don’t know how kids do it; we do know that what we are doing with machines currently isn’t really working.
- Where
- paragraph 40
- Published
- 2022-07-20
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-learning-language-is-harder-than-b73ac2-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-learning-language-is-harder-than-b73ac2-a1.