“they cost almost nothing to operate, and so they are on a path to reducing the cost of generating disinformation to zero.”
Our reading, the sentence that was checked:
Because generative AI systems like ChatGPT, Dall-E 2, and Lensa cost almost nothing to operate, they are on a path to reducing the cost of generating disinformation to zero.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, AI's Jurassic Park moment, paragraph 10.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-ais-jurassic-park-moment-b9a4db-a2
We recorded 15 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 2022-12-10. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Causal claim
- 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)
Conditions
- systems like ChatGPT, Dall-E 2, and Lensa (scope, stated by the speaker)
From the speaker
- cost almost nothing to operate
- and so
- on a path to reducing the cost of generating disinformation to zero
Our interpretation
- The ordinary economic reading concerns a negligible marginal cost per additional generated item, rather than literally costless computation or negligible total company spending.
- The causal proposition is that cheap model operation drives down disinformation-production costs. It gives no deadline for reaching zero.
- I assess the asserted mechanism and trajectory as understood in the contemporary November–December 2022 setting. Later prices can provide context but cannot establish the original operating-cost measurement.
- No company, currency, output unit or accounting basis is specified. The adjacent custom-training price is a different cost category.
Counts as supported if: Comparable operating-cost and production measurements show negligible marginal costs and establish that those costs cause disinformation-production costs to approach an economically negligible level.
Counts as contradicted if: Comparable measurements establish substantial unavoidable production costs, or show that operating costs do not explain the asserted decline.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Does the evidence support the cause?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
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
The evidence supports cheap access and falling inference prices. It does not settle the stronger causal claim. Free preview access measures customer price, while OpenAI's contemporary per-chat estimate was positive and imprecise. Stanford's later findings concern prices across models, not total disinformation-production costs. The unspecified accounting basis also prevents a supported or contradicted financial assessment.
Limitations
- The economically negligible reading of 'zero' is an interpretation.
- The contemporaneous operating-cost estimate is tentative and lacks a defined output unit.
- No cited evidence isolates the causal contribution of inference costs.
Evidence (3 items)
Context only: Sam Altman / X,
average is probably single-digits cents per chat; trying to figure out more precisely and also how we can optimize it
Why it matters here: This tentative company estimate is consistent with small but positive per-chat costs. It does not define chat length or cost components, and does not measure the broader disinformation-production metric.
- Publisher
- Sam Altman / X
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Measures
- OpenAI, Tentative average ChatGPT cost per chat, USD cents per chat, company reported, 2022-12-05 to 2022-12-05
- Describes events of
- Public from
Context only: OpenAI,
During the research preview, usage of ChatGPT is free.
Why it matters here: The announcement establishes a zero customer access price during the preview. That is a different measurement from the cost of operating the model.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Measures
- OpenAI, Customer price for ChatGPT research-preview access Free access, company reported, 2022-11-30 to 2022-11-30
- Describes events of
- Public from
Context only: Stanford Institute for Human-Centered Artificial Intelligence,
Depending on the task, LLM inference prices have fallen anywhere from 9 to 900 times per year.
Why it matters here: Stanford reports a large subsequent decline in inference prices at comparable benchmark performance. These later customer prices do not measure provider operating expense or the full cost of producing disinformation.
- Publisher
- Stanford Institute for Human-Centered Artificial Intelligence
- Kind of source
- Independent measurement
- Measures
- Multiple model providers, Inference prices at comparable task performance, USD Price per million tokens, estimate, 2022-11 to 2024-10
- Describes events of
- Public from
What would settle it
Comparable measurements of operating expense and marginal disinformation-production cost, including human effort, with a defined output unit and evidence identifying what drove any decline.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, AI's Jurassic Park moment
they cost almost nothing to operate, and so they are on a path to reducing the cost of generating disinformation to zero.
- Where
- paragraph 10
- Published
- 2022-12-10
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-ais-jurassic-park-moment-b9a4db-a2: Current. Outcome: Unresolved. 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.