“What I learned this morning is that retraining GPT-3 in some new political direction isn’t a half-million dollar proposition.”
Our reading, the sentence that was checked:
Retraining GPT-3 to take on a new political slant costs far less than half a million dollars.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Two frightening things before breakfast, paragraph 6.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-two-frightening-things-before-breakfast-57e353-a2
We recorded 8 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 2023-02-16. 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
- retraining GPT-3
- in some new political direction
- isn’t a half-million dollar proposition
- What I learned this morning
Our interpretation
- The claim concerns the completed RightWingGPT experiment announced on the post date.
- Retraining means fine-tuning an existing GPT-3 model, as the surrounding account specifies.
- The cost comparison concerns computational expenses for the experiment, not labor, deployment, or training GPT-3 from scratch.
- The ordinary reading is that a political fine-tune was achievable for substantially less than $500,000.
- metric_scope is null because this concerns an individual researcher's experiment expense, not a company's financial metric.
Counts as supported if: A documented GPT-3 fine-tuning experiment produced a changed political slant at computational expense substantially below $500,000.
Counts as contradicted if: Reliable records show that the cited experiment did not produce the claimed political change, or that its comparable computational expense approached or exceeded $500,000.
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 February 16, 2023, Rozado reported that fine-tuning GPT-3 Davinci produced right-of-center responses on held-out political tests and open-ended questions [e1]. He reported computational costs below $300 for trialing, training and testing, supporting the claim that this political fine-tune cost substantially less than $500,000 in computation [e1].
Limitations
- The results and costs are self-reported; independent billing verification is unavailable.
- The evidence does not establish total costs including labor, or the cost or effectiveness of a misinformation operation.
Evidence (1 item)
Supports the claim: David Rozado,
300 USD dollars
Why it matters here: Rozado identifies the base model as GPT-3 Davinci and reports right-of-center responses on held-out political tests and open-ended questions. The completed experiment announced that day had reported computational expense far below Marcus's comparison figure.
- Publisher
- David Rozado
- Kind of source
- Demonstration
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Two frightening things before breakfast
What I learned this morning is that retraining GPT-3 in some new political direction isn’t a half-million dollar proposition.
- Where
- paragraph 6
- Published
- 2023-02-16
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-two-frightening-things-before-breakfast-57e353-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-two-frightening-things-before-breakfast-57e353-a2.