“Large Language Models are — and always will be — incapable of reasoning or consciousness”
Our reading, the sentence that was checked:
LLMs are and always will be incapable of reasoning or consciousness.
Assessment: Contradicted Settled before its deadline on 2026-10-29 Evidence as of
Where and when Ed Zitron wrote it
Where's Your Ed At, Expectations Versus Reality, paragraph 62.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment wyea-expectations-versus-reality-fe4f4f-a3
We recorded 28 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2024-05-13. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Forecast
- Deadline, as the speaker wrote it
- “always”
- Deadline state when assessed
- Settled before its deadline
- Deadline state on 2026-10-29
- Settled before its deadline (the release’s reference date)
From the speaker
- Large Language Models
- are — and always will be
- incapable of reasoning or consciousness
Our interpretation
- Reasoning means using supplied facts or constraints to reach correct conclusions on new problems, rather than merely retrieving an answer.
- The categorical wording denies both reasoning and consciousness. Demonstrating either capability contradicts that combined denial.
- Successful reasoning need not be infallible or human-like.
- Always supplies no finite deadline. Consciousness is not inferred from benchmark performance.
Counts as supported if: Evidence would have to establish permanent incapacity for both reasoning and consciousness across LLMs. Failures of existing models alone cannot establish this.
Counts as contradicted if: A reproducible evaluation demonstrates that an LLM solves newly generated inference problems substantially above chance without an external solver supplying its answers, with checks on the reasoning or solutions; alternatively, compelling evidence establishes consciousness.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Did the forecast come true?
Outcome: Contradicted
The dated evidence goes against what the claim committed to, under the criteria frozen before grading.
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
- Settled before its deadline
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The May 20, 2025 SATBench evaluation reports 89.4% accuracy for o4-mini against a 50% chance baseline on generated, solver-checked logic problems, with no external tools allowed during model evaluation [e1](https://arxiv.org/html/2505.14615v1). Its results and reasoning-trace checks demonstrate limited inference under the contract's definition, contradicting permanent incapacity for reasoning [e1](https://arxiv.org/html/2505.14615v1). The June 7, 2025 study reports accuracy collapse at higher puzzle complexity, establishing substantial limits rather than the absence of all reasoning capability [e2](https://arxiv.org/abs/2506.06941v1).
Limitations
- The studies do not establish consciousness or settle philosophical definitions of reasoning.
- SATBench is a preprint; its reasoning-trace assessment used an LLM judge with partial human validation, and this assessment did not independently rerun the experiment.
- The evidence does not establish general or consistently reliable reasoning.
Evidence (2 items)
Goes against the claim: Anjiang Wei and colleagues, arXiv,
The o4-mini model stands out with a remarkable accuracy of 89.4%.
Why it matters here: SATBench reports performance well above its 50% chance baseline on generated, solver-checked puzzles, without allowing external tools during model evaluation. Reasoning-trace checks and partial human validation support a limited inference capability, although they also reveal errors.
- Publisher
- Anjiang Wei and colleagues, arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: Parshin Shojaee and colleagues, arXiv,
complete accuracy collapse
Why it matters here: Controlled puzzle experiments show substantial limits and failures at higher complexity. Those findings constrain claims of general or reliable reasoning, but do not establish the absence of all reasoning capability.
- Where
- Abstract
- Publisher
- Parshin Shojaee and colleagues, arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Where's Your Ed At, Expectations Versus Reality
Large Language Models are — and always will be — incapable of reasoning or consciousness
- Where
- paragraph 62
- Published
- 2024-05-13
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment wyea-expectations-versus-reality-fe4f4f-a3: Current. Outcome: Contradicted. 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 wyea-expectations-versus-reality-fe4f4f-a3.
HeldBy this site
The final-text check failed: citations_correct.
About assessment wyea-expectations-versus-reality-fe4f4f-a2, which was not published.