“It will be even better than GPT-4 at creating convincing sounding prose.”
Our reading, the sentence that was checked:
GPT-5 will be better than GPT-4 at creating convincing-sounding prose.
Assessment: Unresolved No deadline stated on 2026-10-29 Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, GPT-5 and irrational exuberance, paragraph 8.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-gpt-5-and-irrational-exuberance-32fe5d-a1
We recorded 17 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 2023-04-08. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Forecast
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No deadline stated
- Deadline state on 2026-10-29
- No deadline stated (the release’s reference date)
From the speaker
- even better than GPT-4
- creating convincing sounding prose
Our interpretation
- The plain reading concerns fluent, coherent, plausible prose that readers find convincing, whether or not its factual content is true.
- Better means a reproducible overall preference across representative prose tasks, not superiority on every prompt.
- GPT-4 means the model known at the post date; comparisons with GPT-4o alone do not settle this claim.
- No deadline is stated.
Counts as supported if: Reliable matched-prompt evaluations show that readers consistently judge released GPT-5 prose more convincing than original GPT-4 prose.
Counts as contradicted if: Reliable matched-prompt evaluations show GPT-5 to be worse, or establish practical equivalence with sufficient precision to exclude a meaningful improvement.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Did the forecast come true?
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 deadline stated
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The evidence is suggestive of improved writing. OpenAI's examples compare GPT-5 with GPT-4o, a reviewer favors GPT-5 on one fiction prompt, and a human-rating study reports better academic prose. None provides a sufficiently clear and reliable comparison with the original GPT-4. The forecast remains unresolved; this is an evidence gap, not a finding that GPT-5 writes worse.
Limitations
- Writing preference varies with genre, prompting, and generation settings.
- Academic usefulness and literary quality only partly overlap with convincing-sounding prose.
- The reviewed study's model naming prevents a confident attribution to original GPT-4.
Evidence (3 items)
Context only: OpenAI,
GPT‑5 and GPT‑4o
Why it matters here: OpenAI presents improved writing and selected examples. The explicit comparator is GPT-4o, so this does not directly resolve the forecast about GPT-4.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: Tom's Guide; Amanda Caswell,
tighter, more original and emotionally sharper.
Why it matters here: The reviewer prefers GPT-5 on a matched fiction-opening prompt. This is suggestive, but one unblinded judgment and an unspecified GPT-4 browser configuration cannot establish a general preference over the original model.
- Publisher
- Tom's Guide; Amanda Caswell
- Kind of source
- Demonstration
- Describes events of
- Public from
Context only: arXiv; Georgios P. Georgiou,
clear advantages
Why it matters here: The study uses human ratings with model identities concealed and reports stronger academic writing. However, its clinical-results discussion calls the baseline GPT-4o, and the methods do not resolve the exact model identity.
- Publisher
- arXiv; Georgios P. Georgiou
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
A blinded, matched-prompt human comparison using identified original GPT-4 and GPT-5 versions, covering several prose genres and reporting preference uncertainty.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, GPT-5 and irrational exuberance
It will be even better than GPT-4 at creating convincing sounding prose.
- Where
- paragraph 8
- Published
- 2023-04-08
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-gpt-5-and-irrational-exuberance-32fe5d-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-gpt-5-and-irrational-exuberance-32fe5d-a1.