“And even if this specific example is patched up, there will inevitably be others of a similar flavor.”
Our reading, the sentence that was checked:
Even if this specific LLM error is patched, other errors of a similar flavor will inevitably appear.
Assessment: Supported Settled before its deadline on 2026-10-29 Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Doug Lenat, 1950-2023, paragraph 8.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-doug-lenat-1950-2023-8238f4-a1
We recorded 7 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 2023-09-01. 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
- Settled before its deadline
- Deadline state on 2026-10-29
- Settled before its deadline (the release’s reference date)
Conditions
- even if this specific example is patched up (qualifier, stated by the speaker)
From the speaker
- even if this specific example is patched up
- there will inevitably be others
- of a similar flavor
Our interpretation
- The plain reading is that further LLM outputs would contain fluent but obvious commonsense errors after the post.
- I read 'even if' as a concession: recurrence is predicted whether or not the particular Bard example is repaired. It does not require a documented repair before the prediction applies.
- 'Similar flavor' refers to the surrounding description of plausible language containing obvious nonsense, rather than an identical question or error.
- 'Inevitably' expresses confidence that further examples will occur. I do not interpret it as a separately testable assertion that every future model must remain defective forever.
Counts as supported if: Reliable records document subsequent LLM outputs containing obvious commonsense errors presented as plausible answers.
Counts as contradicted if: A decisive record establishes that no further qualifying error can occur. A finite period without reported errors or correction of individual examples is insufficient because there is no deadline.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Did the forecast come true?
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
- Settled before its deadline
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
Further errors occurred. In May 2024, Google acknowledged that its language-model search feature mishandled satire and offered nonsensical advice, including glue on pizza. Separate experiments documented commonsense reasoning failures with plausible explanations. These satisfy the ordinary recurrence reading. 'Even if' is treated as a concession, so an actual repair of the original Bard example is not required. The evidence does not establish that every future LLM must always have such failures.
Limitations
- The result depends on the stated readings of 'even if', 'similar flavor', and 'inevitably'.
- AI Overviews combines a language model with retrieval and ranking; the evidence does not isolate which component caused each error.
- The research evidence is a preprint about a limited task and also reports successful responses.
- The original illustrated example was unavailable in the extracted post text.
Evidence (2 items)
Supports the claim: Google,
AI Overviews that featured sarcastic or troll-y content
Why it matters here: Google identifies AI Overviews as powered by a customized language model and acknowledges advice about putting glue on pizza, alongside mishandling a question about eating rocks. These are subsequent commonsense failures matching the surrounding description of plausible answers containing nonsense.
- Publisher
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Supports the claim: arXiv; Marianna Nezhurina and colleagues,
frequent failures to solve this simple problem
Why it matters here: The experiments test counting sisters from a brother's perspective and report wrong answers accompanied by plausible but faulty explanations. This independently documents further commonsense reasoning errors, although the task is narrower than ordinary conversation.
- Publisher
- arXiv; Marianna Nezhurina and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Doug Lenat, 1950-2023
And even if this specific example is patched up, there will inevitably be others of a similar flavor.
- Where
- paragraph 8
- Published
- 2023-09-01
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-doug-lenat-1950-2023-8238f4-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.