“To get there, though, what we need is a much smarter AI, a kind that can reason, and not hallucinate, grounded in reality and not just corpus statistics— which require fundamental research of a sort that is currently getting starved out in the LLM euphoria.”
Our reading, the sentence that was checked:
Getting to reliable self-driving cars that are safer than humans requires a much smarter AI that can reason, does not hallucinate, and is grounded in reality rather than just corpus statistics, and building that AI requires fundamental research.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Why driverless cars might—or might not—be at the end of the road, paragraph 23.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-why-driverless-cars-mightor-might-ab4c33-a1
We recorded 8 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 2023-10-27. 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)
Conditions
- to get to reliable self-driving cars that are eventually safer than humans (qualifier, stated by the speaker)
From the speaker
- To get there
- what we need is a much smarter AI
- can reason, and not hallucinate
- grounded in reality and not just corpus statistics
- require fundamental research
Our interpretation
- Following the supplied normalized proposition, this assessment concerns necessary technical capabilities and research, rather than separately adjudicating the accompanying research-funding assertion.
- The plain reading is that reliable, safer-than-human driving requires improved reasoning grounded in the driving environment and control of fabricated representations, with fundamental research needed to obtain those capabilities.
- The claim concerns necessity, not merely whether reasoning, sensing or research can help.
- It does not explicitly require symbolic AI, a particular architecture or a ban on statistical learning.
Counts as supported if: Technical and safety evidence identifies the stated capabilities and fundamental research advances as necessary for reliable, safer-than-human driving, rather than merely present in successful systems.
Counts as contradicted if: A documented system achieves reliable, safer-than-human driving while demonstrably lacking an asserted necessary capability, or a well-supported technical account establishes that the claimed fundamental advance was unnecessary.
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
The records establish technical approaches and safety results, but not the claimed causal necessity. Waymo described contextual reasoning before the post, and later research combines language models with physical sensor data. Neither fact proves that a new kind of AI was required or unnecessary. Lower crash rates also do not show whether a deployed system lacks any capability Marcus says is essential.
Limitations
- The key capabilities and required research advances lack precise definitions.
- Company descriptions of reasoning are not independent capability tests.
- A research model cannot be assumed to be the system responsible for fleet safety results.
Evidence (3 items)
Context only: Waymo,
reason about the context
Why it matters here: Waymo described contextual reasoning combined with camera and lidar perception before Marcus's post. This complicates the meaning of a new, smarter AI but does not independently establish necessity or freedom from hallucinations.
- Publisher
- Waymo
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: arXiv,
Multimodal Model
Why it matters here: The research combines a multimodal language model with camera data and driving outputs, showing that statistical learning and physical grounding are compatible. Its benchmark results do not establish deployed safety or prove that this approach is necessary.
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv,
no statistically significant disbenefit
Why it matters here: The study reports lower injury-related crash rates over 56.7 million driverless miles through January 2025. Safety outcomes alone do not establish whether the system possesses Marcus's proposed capabilities or whether fundamental research was indispensable.
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
A documented successful driving system with capability tests and development history could supply a counterexample. Supporting necessity would require evidence that removing or lacking a specified capability prevents the required safety performance, with evidence connecting that capability to the claimed research advance.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Why driverless cars might—or might not—be at the end of the road
To get there, though, what we need is a much smarter AI, a kind that can reason, and not hallucinate, grounded in reality and not just corpus statistics— which require fundamental research of a sort that is currently getting starved out in the LLM euphoria.
- Where
- paragraph 23
- Published
- 2023-10-27
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-why-driverless-cars-mightor-might-ab4c33-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-why-driverless-cars-mightor-might-ab4c33-a1.