“Building a robot on top of language system that has such little comprehension of the world can’t be a recipe for success.”
Our reading, the sentence that was checked:
Building a robot on top of a language system with so little comprehension of the world cannot be a recipe for success.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, What’s wrong with Google’s new robot project, paragraph 12.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-whats-wrong-with-googles-new-robot-1cbb4b-a1
We recorded 20 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 2022-08-18. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Causal claim
- 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
- Building a robot on top of language system
- such little comprehension of the world
- can’t be a recipe for success
Our interpretation
- Read success in context as dependable, safe performance of varied household requests, rather than completion of a selected demonstration.
- Read the claim as a causal objection: inadequate world understanding prevents dependable robot performance.
- The speaker supplies neither a measurement of world comprehension nor a success threshold. Those omissions limit how strongly experiments can settle the claim.
- Later research may inform the causal assessment, but success with a materially different architecture does not automatically refute a criticism of the 2022 approach.
Counts as supported if: Controlled, representative evaluations establish that limited world understanding in the relevant language system prevents dependable household performance, after separating that limitation from perception, control, and hardware failures.
Counts as contradicted if: Robots using the criticized kind of language system demonstrate sustained, independently verified, safe performance across varied household requests without overcoming the limitation alleged to prevent success.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Does the evidence support the cause?
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
Research shows that language-model-based robotic planning can improve with environmental feedback and additional planning components. That weighs against dismissing the approach outright. But selected successes do not establish dependable household operation, and remaining failures do not prove that limited world comprehension makes success impossible. The evidence does not settle the causal claim under the contextual reading of success.
Limitations
- The result depends on reading success as dependable household performance rather than any successful demonstration.
- The cited studies do not directly measure the speaker's concept of world comprehension.
- Later systems introduce components and capabilities that complicate comparisons with PaLM-SayCan.
Evidence (2 items)
Context only: arXiv; Huang and colleagues,
environment feedback
Why it matters here: The experiments show that language-model-based robotic planning can improve through environmental feedback. This challenges a blanket dismissal of the approach, but does not demonstrate dependable, safe household operation or isolate world comprehension as the decisive cause.
- Publisher
- arXiv; Huang and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv; Zhu and colleagues,
improves long-horizon performance
Why it matters here: Later work reports improved household planning with additional memory and planning components, while substantial task failures remain. Its benchmark abstracts away physical control and uses a different system, so it cannot settle the claim about the criticized architecture.
- Publisher
- arXiv; Zhu and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
Representative, repeated household evaluations with safety checks and controlled comparisons separating language understanding from perception, control, and added feedback, using explicit criteria for success and the system class being tested.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, What’s wrong with Google’s new robot project
Building a robot on top of language system that has such little comprehension of the world can’t be a recipe for success.
- Where
- paragraph 12
- Published
- 2022-08-18
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-whats-wrong-with-googles-new-robot-1cbb4b-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-whats-wrong-with-googles-new-robot-1cbb4b-a1.