“They are creating interactive fiction that is trained to predict the language of actual beings.”
Our reading, the sentence that was checked:
LLMs are interactive fiction trained to predict the language of actual beings.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, The Pope appears to understand AI better than Geoffrey Hinton does., paragraph 19.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-the-pope-appears-to-understand-ai-b85b53-a1
We recorded 9 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 2026-05-31. 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
- creating interactive fiction
- trained to predict the language of actual beings
Our interpretation
- In context, 'interactive fiction' means generated conversational characters rather than genuine experiencing interlocutors; it does not mean every factual answer is false.
- Language-prediction training and conversational simulation are checkable components. They do not alone settle the stronger claim that the interlocutor is merely fictional.
- The statement concerns systems existing by 2026-05-31. There is no reporting cadence or specified historical starting date, so assess the pre-post record.
Counts as supported if: Evidence establishes language-prediction training and conversational simulation, and independently supports the stronger distinction between a simulated character and an experiencing entity.
Counts as contradicted if: Reliable evidence establishes genuine experience in a relevant LLM interlocutor, or establishes that the asserted training-and-simulation characterization materially misdescribes the systems covered.
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
Prediction training and conversational role play have primary-source support. The surrounding contrast with 'beings' makes the full claim stronger: generated characters are said to lack genuine experience. The cited role-play account and limited consciousness assessment do not settle that broader assertion for the systems covered. Scoring the entire sentence supported would treat evidence for its training clause as proof of its claim about experience.
Limitations
- On a purely behavioral reading, the role-play characterization has support.
- The evidence search does not establish either consciousness or its absence in all relevant systems.
- The cited conceptual accounts are not direct measurements of subjective experience.
Evidence (2 items)
Supports the claim: arXiv, rendered by ar5iv,
role-play all the way down
Why it matters here: The paper supports a role-play account and describes next-token training on human text. It focuses on base models and explicitly says reinforcement learning may weaken the distinction between simulator and simulated character, so it does not settle the full claim's scope.
- Publisher
- arXiv, rendered by ar5iv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv,
indicator properties
Why it matters here: The report assesses consciousness using properties drawn from scientific theories and tentatively finds no conscious systems among those examined. That limited assessment does not establish that every LLM covered by the May 2026 statement is merely fictional.
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
A clear operational meaning for 'interactive fiction' and 'beings', followed by evidence assessing the relevant systems against those criteria. If the intended claim is only about prediction training and simulated personas, that narrower claim can be assessed separately.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, The Pope appears to understand AI better than Geoffrey Hinton does.
They are creating interactive fiction that is trained to predict the language of actual beings.
- Where
- paragraph 19
- Published
- 2026-05-31
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-the-pope-appears-to-understand-ai-b85b53-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-the-pope-appears-to-understand-ai-b85b53-a1.