“On its own, it still requires frequent, massive retraining to be current with the news; GPT 3.5 knew nothing of 2022; GPT-4 seems to know little of 2023.”
Our reading, the sentence that was checked:
On its own, GPT-4 still requires frequent massive retraining to stay current with the news.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, GPT-4’s successes, and GPT-4’s failures, paragraph 7.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-gpt-4s-successes-and-gpt-4s-failures-a647e7-a1
We recorded 12 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 2023-03-15. 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)
Conditions
- on its own (qualifier, stated by the speaker)
From the speaker
- On its own
- requires frequent, massive retraining
- to be current with the news
- GPT 3.5 knew nothing of 2022
- GPT-4 seems to know little of 2023
Our interpretation
- The principal proposition concerns the necessity and scale of training needed to keep standalone release-era GPT-4 current.
- On its own excludes retrieval and supplying current news in the prompt.
- Frequent means recurring updates as news changes; massive means large-scale training, rather than merely any parameter update. Marcus specifies neither a schedule nor a compute threshold.
- The target is the March 2023 release state through the post date, using March month-to-date as the latest public reporting window.
- The knowledge-date clauses are supporting assertions. A training cutoff does not establish literal ignorance of every later fact, or prove that massive retraining is necessary.
Counts as supported if: Technical evidence establishes that keeping standalone release-era GPT-4 current requires repeated large-scale training, rather than substantially smaller updates.
Counts as contradicted if: A demonstrated method keeps that standalone model broadly current through substantially smaller updates without retrieval or current-news context.
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
The original report establishes bounded training knowledge. That supports the concern about freshness, but does not prove that updating standalone GPT-4 required frequent, massive retraining. The searched records do not establish the necessary update scale or rule out smaller updates. Retrieval would not contradict the explicit standalone condition.
Limitations
- A knowledge cutoff is not evidence of complete ignorance of every later event.
- Update methods demonstrated on other models would not automatically settle this claim.
- This is a historical technical claim, not a forecast about later GPT versions.
Evidence (1 item)
Context only: OpenAI,
cutoff point
Why it matters here: The disclosure describes the launch model's bounded training knowledge. It does not measure the cost, frequency, or necessity of large-scale training to update that knowledge.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
What would settle it
Technical records or controlled experiments on the original GPT-4 showing the training resources and update frequency necessary for broad, reliable current-news knowledge without retrieval.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, GPT-4’s successes, and GPT-4’s failures
On its own, it still requires frequent, massive retraining to be current with the news; GPT 3.5 knew nothing of 2022; GPT-4 seems to know little of 2023.
- Where
- paragraph 7
- Published
- 2023-03-15
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-gpt-4s-successes-and-gpt-4s-failures-a647e7-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-4s-successes-and-gpt-4s-failures-a647e7-a1.