“The system stubbornly insists on tree growing out of (rather than on top of the) monkey, which is not what I asked for. It recognizes its own error but still can’t get the image right:”
Our reading:
ChatGPT could not generate a tree on top of a monkey, even after acknowledging its own error.
Assessment: Unresolved
What happened
Was it true for the period described?
Outcome: Unresolved
We checked, and the evidence so far doesn’t settle it.
Judged on a plain reading. They named no measure, so we used the reading a typical informed reader would take. It is spelled out under “How we judged it”.
Evidence up to . Checked .
Why
The retrieved records do not establish the sequence within Marcus’s conversation. A separate user’s similar experiment cannot verify that ChatGPT acknowledged an error and then repeated it in this exchange. The available evidence therefore neither supports nor contradicts the reported incident.
Caveats
- The essential dialogue and images were not available in the page reader’s text.
- The visual distinction should be judged using the original clarification, not an invented preferred composition.
Evidence (0 items)
Nothing we found settles this yet.
How we judged it
- Kind of claim
- Observation
- Deadline
- None given. We don’t add one.
Conditions
- even after recognizing its error (their words)
Where we had to interpret
- This describes the particular ChatGPT conversation reported in the post.
- Recognition means an explicit acknowledgment in the dialogue, without implying an internal mental state.
- The claimed sequence is an incorrect image, acknowledgment of the error, and a subsequent image that still depicts growth out of the monkey.
- The target is the conversation reported by June 25, 2025. Its exact date is unspecified, so the start is left unknown.
Right if: The actual conversation and its images show the requested distinction, an explicit acknowledgment of the error, and continued depiction of a tree growing out of the monkey after that acknowledgment.
Wrong if: An authenticated record of the same conversation shows that the alleged acknowledgment did not occur, or that the allegedly still-incorrect subsequent image actually satisfied the clarified request. Results from another conversation cannot settle this report.
We wrote this test after the outcome was known.
What would settle it
An independently authenticated copy of the original dialogue and images showing the acknowledgment and the next relevant generation.
We found 8 claims in this post. See them all.
Full record
Source and dates
Marcus on AI, Image generation: Still crazy after all these years, paragraph 32.
- Said
- We read the source
- live page, full text, public
- Evidence up to
- Checked
We read this wording on 2026-10-28. The post may have been edited after 2025-06-25.
Where it was said (1)
Said once in the posts we read.
: Marcus on AI, Image generation: Still crazy after all these years
The system stubbornly insists on tree growing out of (rather than on top of the) monkey, which is not what I asked for. It recognizes its own error but still can’t get the image right:
- Where
- paragraph 32
- Published
- 2025-06-25
- We read it
- , live page
- Access
- full text, public
History
Versions
- Current: Unresolved. Evidence up to 2026-10-28.
What happened to this record
PublishedBy us
Published after both AI reviews and the final check passed.
Held backBy us
A reviewer said our test wasn’t fair to what was actually said.