“It’s not smart enough to filter out falsehood, but just smart enough to be dangerous, creating and spreading falsehoods it fails to verify.”
Our reading, the sentence that was checked:
GPT cannot filter out falsehood and creates and spreads falsehoods it fails to verify, making it dangerous.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, GPT-5 and irrational exuberance, paragraph 3.
- 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-c5623d-a2
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
- 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
- not smart enough to filter out falsehood
- creating and spreading falsehoods it fails to verify
- smart enough to be dangerous
Our interpretation
- The plain reading is that contemporary GPT systems cannot reliably screen false assertions and can deliver unverified claims capable of causing harm. It does not mean they never distinguish truth from falsehood.
- The target is the contemporary GPT product generation during March 2023 through the post date. This covers the latest completed month and the current partial month; these systems have no standard reporting period.
- Danger includes the foreseeable reputational harm of presenting invented misconduct allegations as factual.
Counts as supported if: Contemporaneous documented outputs show GPT systems presenting materially false factual assertions or invented supporting references, with credible evidence that factual verification remained unreliable.
Counts as contradicted if: Representative contemporaneous evidence establishes reliable factual screening, and the alleged failures are shown to be misattributed or otherwise invalid.
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
On March 14, 2023, OpenAI acknowledged that GPT-4 still hallucinated facts, while reporting improved factuality and truth discrimination [e1]. On March 22, Volokh published GPT-4 misconduct allegations and purported newspaper quotations that he could not corroborate through searches [e2]. Together, these support unreliable factual screening and a risk of reputational harm from unverified allegations [e1, e2]. They do not support an absolute inability to distinguish truth from falsehood [e1].
Limitations
- Volokh's unsuccessful searches support his assessment that the examples were fabricated, but are not exhaustive proof that no corresponding source exists [e2].
- The evidence does not independently verify the specific Bing incident described by Marcus [e1, e2].
- The sources do not establish population-wide error rates or realized reputational damages [e1, e2].
Evidence (2 items)
Supports the claim: OpenAI,
it still is not fully reliable
Why it matters here: OpenAI acknowledges factual unreliability in the released system. The same disclosure reports improved factuality and truth discrimination, which limits any literal reading that GPT cannot recognize falsehood at all.
- Where
- Limitations
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Supports the claim: Reason — Eugene Volokh,
bogus quotations and citations
Why it matters here: Volokh reproduces GPT-4 misconduct allegations and purported quotations that his searches could not substantiate. These outputs demonstrate the mechanism for reputational harm, although they are not a representative error-rate study.
- Publisher
- Reason — Eugene Volokh
- 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, GPT-5 and irrational exuberance
It’s not smart enough to filter out falsehood, but just smart enough to be dangerous, creating and spreading falsehoods it fails to verify.
- Where
- paragraph 3
- Published
- 2023-04-08
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-gpt-5-and-irrational-exuberance-c5623d-a2: 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-gpt-5-and-irrational-exuberance-c5623d-a2.