“GPT doesn’t talk randomly, because it’s pastiching things actual people said.”
Our reading, the sentence that was checked:
GPT's output is not random, because it recombines things real people have said.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, How come GPT can seem so brilliant one minute and so breathtakingly dumb the next?, paragraph 41.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-how-come-gpt-can-seem-so-brilliant-33bab9-a1
We recorded 9 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-29. The source may have been edited after 2022-12-01. 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
- GPT doesn’t talk randomly
- because it’s pastiching things actual people said
- synonyms and paraphrases
Our interpretation
- GPT refers to the GPT-3 generation and related systems discussed in the December 2022 post.
- Nonrandom means linguistically structured rather than arbitrary characters or words. It does not mean generation involves no randomness.
- Pastiching is read as generating combinations of patterns learned from human text. Literal retrieval of existing passages is a stronger reading.
- The assessment concerns the source of linguistic structure, not whether generated statements are true.
Counts as supported if: Training documentation and controlled generation evaluations establish that learning patterns from human text produces structured, human-like output.
Counts as contradicted if: Controlled evidence establishes that the output's linguistic structure does not depend materially on patterns learned from human text.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Does the evidence support the cause?
Outcome: Supported
The dated evidence matches 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
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The ordinary reading is supported: GPT learns linguistic patterns from human text and uses them to generate structured prose. The GPT-3 study documents both the training material and human-like generated output. This supports Marcus's distinction between learned language generation and arbitrary word production. It does not validate literal cut-and-paste generation or the stronger claim that every fragment has an identifiable human source.
Limitations
- The result depends on reading pastiching as learned recombination rather than literal copying.
- The causal inference rests on documented training and generation results, not a complete mechanism-level intervention.
Evidence (2 items)
Supports the claim: OpenAI researchers / arXiv,
news articles
Why it matters here: The study documents training on web text, books and Wikipedia, alongside controlled assessments of generated articles. Together these support learned language patterns as the basis of structured output, without establishing literal cut-and-paste generation.
- Publisher
- OpenAI researchers / arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: OpenAI,
predict the next word
Why it matters here: This documents the preceding GPT model's text-prediction objective. It explains the relevant training approach but is not a direct test of December 2022 GPT systems.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, How come GPT can seem so brilliant one minute and so breathtakingly dumb the next?
GPT doesn’t talk randomly, because it’s pastiching things actual people said.
- Where
- paragraph 41
- Published
- 2022-12-01
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-how-come-gpt-can-seem-so-brilliant-33bab9-a1: Current. Outcome: Supported. 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-how-come-gpt-can-seem-so-brilliant-33bab9-a1.