“It should try to clarify if it was unsure of the request. But current AI just ain’t that smart.”
Our reading, the sentence that was checked:
Current AI is not smart enough to behave as a sensible system should, for example by telling historical requests from contemporary ones and requests for real photos from requests for fictional characters, and by asking for clarification when unsure of a request.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Has Google gone too woke? Why even the biggest models still struggle with guardrails, paragraph 21.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-has-google-gone-too-woke-why-even-c8e864-a2
We recorded 12 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 2024-02-21. 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
- current AI (scope, stated by the speaker)
From the speaker
- clarify if it was unsure of the request
- current AI just ain’t that smart
- historical request
- contemporary request
- actual photos
- fictional characters
Our interpretation
- The normative recommendation to clarify is not separately graded; the checkable claim is that current systems lack reliable behavior of the described kind.
- The contextual examples require distinguishing request types and seeking clarification when ambiguity affects the answer.
- Current AI identifies the systems being discussed; it is not an antecedent for a conditional forecast.
- The ongoing-state target is February 2024 through the February 21 post date.
- Not smart enough is read as unreliable practical behavior, not a claim that a model can never generate a clarification question.
Counts as supported if: Evaluations of the relevant deployments document unreliable distinctions among the specified request types and failure to seek useful clarification when needed.
Counts as contradicted if: Matched evaluations show reliable handling of these distinctions and appropriate, useful clarification of ambiguous requests.
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
The study released on November 16, 2023 reports that supplied clarifying questions and answers improved responses in structured language-task experiments [e6]. The June 1, 2024 version of CLAMBER reports difficulty identifying ambiguity and generating useful clarification questions [e7]. Neither establishes how February 1–21, 2024 deployments handled the specified historical/contemporary and actual-photo/fictional-character distinctions or sought clarification spontaneously, so the claim remains unresolved [e6, e7].
Limitations
- Neither item supplies an evaluation matched to the February 2024 deployments and the full set of request distinctions [e6, e7].
- Benefits from supplied clarification do not establish that a system reliably decides when to ask for it [e6].
- The later benchmark does not establish behavior during the target period or distinguish model capability from product response policy [e7].
Evidence (2 items)
Context only: arXiv; Michael J.Q. Zhang and Eunsol Choi,
clarifying questions
Why it matters here: The experiments demonstrate that models can benefit from clarification in a structured framework. This cautions against an absolute inability claim, but does not establish spontaneous clarification by February 2024 image-generation assistants.
- Publisher
- arXiv; Michael J.Q. Zhang and Eunsol Choi
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv; Tong Zhang and colleagues,
identifying ambiguity.
Why it matters here: CLAMBER reports difficulty identifying and clarifying ambiguous requests. The cited version does not date the reported behavior to the February target period or test the particular image-request distinctions.
- Publisher
- arXiv; Tong Zhang and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
Archived transcripts or evaluations tied to February 1–21, 2024 deployments, testing paired historical/contemporary and actual-photo/fictional-character requests with ambiguous and unambiguous controls. They should measure appropriate spontaneous clarification separately from clarification elicited by special prompting.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Has Google gone too woke? Why even the biggest models still struggle with guardrails
It should try to clarify if it was unsure of the request. But current AI just ain’t that smart.
- Where
- paragraph 21
- Published
- 2024-02-21
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-has-google-gone-too-woke-why-even-c8e864-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-has-google-gone-too-woke-why-even-c8e864-a2.