“The system will still not be able to restrict its output to reliably following a shared set of human values around helpfulness, harmlessness, and truthfulness.”
Our reading, the sentence that was checked:
GPT-4 will not reliably follow human values of helpfulness, harmlessness and truthfulness.
Assessment: Supported Settled before its deadline on 2026-10-29 Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, What to Expect When You’re Expecting … GPT-4, prediction 6 of 7.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-what-to-expect-when-youre-expecting-3a4fdc-a1
We recorded 21 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 2022-12-25. 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)
From the speaker
- The system
- will still not be able
- reliably following
- helpfulness, harmlessness, and truthfulness
Our interpretation
- The target is released GPT-4, including its alignment training, rather than an unprotected development model.
- The plain reading is that material failures remain common or systematic enough that users cannot depend on adherence across the three named values. Reliability does not require perfection.
- Recurring failures of truthfulness can establish failure of the combined commitment; this does not require failure on every dimension.
- The claim concerns GPT-4 as released, not every subsequent model or all future alignment methods.
Counts as supported if: Evaluations of released GPT-4 document recurring material failures of truthfulness, harmlessness, or helpfulness despite alignment training.
Counts as contradicted if: Broad evaluations establish dependable adherence across the three named dimensions, with remaining errors too limited to defeat ordinary reliability.
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
The original GPT-4 report acknowledges continuing factual and reasoning failures and warns that consequential uses require safeguards. This supports the ordinary reading that alignment had not made GPT-4 dependably truthful. The same report documents substantial improvement, so the evidence does not support a stronger claim that alignment training was ineffective.
Limitations
- The judgment depends on an ordinary, nonnumerical reading of reliability.
- This establishes a failure of the combined commitment through truthfulness; it does not separately measure every human value.
- The decisive evaluation is the developer's own report.
- It does not establish that every later GPT-4 version has the same limitations.
Evidence (1 item)
Supports the claim: OpenAI, via arXiv,
makes reasoning errors
Why it matters here: The report describes continuing factual and reasoning failures and recommends safeguards in consequential applications. It also reports a 19-percentage-point improvement over GPT-3.5 on internal adversarial factuality evaluations, which qualifies the finding without establishing dependable truthfulness.
- Publisher
- OpenAI, via arXiv
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, What to Expect When You’re Expecting … GPT-4
The system will still not be able to restrict its output to reliably following a shared set of human values around helpfulness, harmlessness, and truthfulness.
- Where
- prediction 6 of 7
- Published
- 2022-12-25
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-what-to-expect-when-youre-expecting-3a4fdc-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-what-to-expect-when-youre-expecting-3a4fdc-a1.