“one of the things that popped out to me is that the system still quickly lands in the land of hallucination that has so haunted ChatGPT.”
Our reading, the sentence that was checked:
Anthropic's Claude, despite its alignment focus, still quickly hallucinates like ChatGPT.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, 24 Seriously Embarrassing Hours for AI, paragraph 18.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-24-seriously-embarrassing-hours-for-3a1915-a1
We recorded 6 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 2023-01-18. 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
- Riley Goodside... put Claude... to the test
- his detailed comparison
- the system still quickly lands in the land of hallucination
- that has so haunted ChatGPT
Our interpretation
- Plain reading: in Goodside's January 2023 comparison, the early Claude model readily produced made-up facts, as ChatGPT did, despite its alignment focus.
- Period: Goodside's testing of pre-release Claude, published 2023-01-17.
- 'Quickly' is read as hallucinating on ordinary factual prompts, not only after adversarial effort.
Counts as supported if: Goodside's comparison documents Claude producing confabulated or false facts on ordinary prompts.
Counts as contradicted if: Goodside's comparison shows Claude largely avoiding hallucinations where ChatGPT made them.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
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
Goodside and Papay's comparison of the early Claude, published January 17, 2023, records several hallucinations. Claude named the wrong Super Bowl winner. It invented terms it attributed to a Stanislaw Lem story. Its summaries of Lost became 'almost entirely confabulated' by Season 4. These were ordinary factual prompts, so Claude did land in hallucination readily, much as ChatGPT did. The same post notes that Claude sometimes declined questions it could not answer and avoided one hallucination ChatGPT made. So the claim holds, though Claude was not uniformly worse.
Limitations
- The comparison is informal and based on few prompts.
- 'Quickly' is a loose word; the post shows hallucinations on everyday prompts but does not measure speed or frequency.
Evidence (3 items)
Supports the claim: Scale AI,
By the time we reach Season 4, Claude’s memory of the show becomes almost entirely confabulated.
Why it matters here: Goodside's test shows Claude confabulating plot details on an ordinary factual request, much as ChatGPT did.
- Publisher
- Scale AI
- Kind of source
- Independent measurement
- Describes events of
- Public from
Supports the claim: Scale AI,
None of the fictional words given by Claude (hyperconcentration, hypermotorics, or omnivorous transformers) seem to actually appear in Lem’s work
Why it matters here: A second example of Claude inventing details.
- Publisher
- Scale AI
- Kind of source
- Independent measurement
- Describes events of
- Public from
Context only: Scale AI,
generally seems more cognizant of what it cannot do than ChatGPT is.
Why it matters here: Counterpoint: Claude sometimes declined where ChatGPT guessed, and it also avoided ChatGPT's hallucination on the Hofstadter question.
- Publisher
- Scale AI
- Kind of source
- Independent measurement
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, 24 Seriously Embarrassing Hours for AI
one of the things that popped out to me is that the system still quickly lands in the land of hallucination that has so haunted ChatGPT.
- Where
- paragraph 18
- Published
- 2023-01-18
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-24-seriously-embarrassing-hours-for-3a1915-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-24-seriously-embarrassing-hours-for-3a1915-a1.