“We can’t really build humanoid domestic robots without general intelligence, and we don’t yet have any serious candidates.”
Our reading, the sentence that was checked:
Humanoid domestic robots require general intelligence, and there are no serious candidates for such intelligence yet.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Three baffling claims about AI and machine learning in four days, statistical errors in top journals, and claims from Yann LeCun that you should not believe., paragraph 86.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-too-much-benefit-of-the-doubt-114c2e-a1
We recorded 13 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 2022-10-16. 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
- humanoid domestic robots
- without general intelligence
- we don’t yet have any serious candidates
- safe and useful domestic humanoid robot
- at present we have no viable option
Our interpretation
- Read in context, the claim concerns an autonomous humanoid that can safely perform varied, useful household work in ordinary homes.
- A serious candidate means an existing intelligence system credibly capable of serving as that robot’s controller. It need not be a finished consumer product, but a research ambition alone does not qualify.
- General intelligence means broad learning, reasoning and adaptation beyond a fixed collection of trained tasks; Marcus supplies no precise test.
- The ongoing observation is assessed as of 2022-10-16. Robotics research has no standard reporting quarter, so the target is the post-date snapshot. Earlier demonstrations are background, not direct measurements of that snapshot.
- The necessity clause requires separate evidence. Failure to find an adequate robot does not establish that general intelligence is necessary.
Counts as supported if: Evidence establishes that safe, broadly useful autonomous domestic humanoids require general intelligence, and a sufficiently comprehensive assessment of systems available on the post date finds no credible candidate controller.
Counts as contradicted if: A documented safe, broadly useful autonomous domestic humanoid operating without general intelligence disproves the necessity clause; credible evaluation of an existing general-intelligence candidate suitable for that role disproves the no-candidates clause.
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 records show relevant progress: Gato combined several kinds of tasks, PaLM-SayCan performed kitchen instructions, and ASIMO demonstrated autonomous humanoid capabilities. None establishes the safe, broadly useful domestic humanoid described in context. But limited demonstrations and the absence of a qualifying counterexample do not prove that general intelligence is necessary or that no serious candidate existed. Calling any generalist research system a serious candidate would make the claim easier to contradict; requiring a finished household product would make it easier to support. Neither threshold is supplied by Marcus. The evidence does not settle the combined claim.
Limitations
- The sources report their developers’ results, rather than independent household safety evaluations.
- The claim combines a necessity assertion with an observation about available candidates.
- The search cannot establish a universal absence.
Evidence (3 items)
Context only: arXiv; DeepMind authors,
multi-modal, multi-task, multi-embodiment generalist policy.
Why it matters here: Gato is material evidence against interpreting Marcus as saying no generalist research systems existed. The abstract describes multiple capabilities, including robot-arm control, but does not establish a viable controller for safe domestic humanoids.
- Publisher
- arXiv; DeepMind authors
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: Google Research,
successful planning for 84%
Why it matters here: Google reported kitchen-task planning and execution, with the table reporting 74% execution success. This is relevant progress toward useful domestic control, but the evaluation does not establish broad humanoid competence or household safety.
- Publisher
- Google Research
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: Honda,
without being controlled by an operator.
Why it matters here: Honda described autonomous humanoid movement and, elsewhere on the page, object handling and pouring. These capabilities challenge an overly broad reading that humanoids cannot perform useful actions, but do not establish a safe, broadly useful domestic robot.
- Publisher
- Honda
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
What would settle it
A documented pre-post humanoid meeting the contextual household requirements without general intelligence would refute the necessity clause. A credible contemporaneous candidate evaluation could refute the second clause. Supporting both requires evidence for the necessity claim and a comprehensive assessment of the candidates then available.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Three baffling claims about AI and machine learning in four days, statistical errors in top journals, and claims from Yann LeCun that you should not believe.
We can’t really build humanoid domestic robots without general intelligence, and we don’t yet have any serious candidates.
- Where
- paragraph 86
- Published
- 2022-10-16
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-too-much-benefit-of-the-doubt-114c2e-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-too-much-benefit-of-the-doubt-114c2e-a1.