“Knockoffs of GPT-3 are getting cheaper and cheaper, which means that the cost of generating misinformation is going to zero, and the quantity of misinformation is going to rise—probably exponentially.”
Our reading, the sentence that was checked:
Knockoffs of GPT-3 are getting cheaper and cheaper.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Bravo @ElonMusk for wanting Twitter to become the world’s “most accurate source of information”, paragraph 14.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-bravo-elonmusk-for-wanting-twitter-7050d7-a1
We recorded 9 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 2022-11-09. 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
- Knockoffs of GPT-3
- are getting cheaper and cheaper
Our interpretation
- Assess only the historical cost clause, not the forecasts elsewhere in the sentence.
- Knockoffs means independently developed GPT-like text generators, including open models; it does not require identical performance to GPT-3.
- The ordinary contextual reading concerns the cost of using these models to generate text, including self-hosted inference costs.
- For this ongoing trend, use 2022 year-to-date through the last complete day before publication. This lookback is an evaluator choice because the speaker supplied no reporting period.
- This concerns product-use costs, not any company's financial accounts, so metric_scope is null.
Counts as supported if: Dated evidence available for the target period shows that users could generate comparable text with GPT-like alternatives at lower operating costs through new services or improved inference methods.
Counts as contradicted if: Comparable historical measurements show that the relevant attainable generation costs remained unchanged or rose, with purported reductions explained by materially reduced capability or excluded costs.
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
Under the contextual reading of cheaper text generation, dated 2022 releases support the claim. DeepSpeed demonstrated lower generation costs for open GPT-like models relative to their original implementations before Marcus wrote the post. This supports falling attainable operating costs; it does not establish that every alternative provider repeatedly lowered its API prices.
Limitations
- The result depends on including inference operating costs in cheaper.
- Reported savings depend on hardware, batching, and workload.
- The evidence does not establish equal capability between every named alternative and GPT-3.
Evidence (3 items)
Supports the claim: Microsoft DeepSpeed,
lower latency and cost
Why it matters here: The release compares optimized and baseline implementations of open text generators, including BLOOM and OPT, using dollars per million generated tokens. Its cost-sensitive benchmarks report substantial reductions, directly supporting cheaper attainable inference before the post.
- Publisher
- Microsoft DeepSpeed
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: Microsoft DeepSpeed,
ZeRO-Inference offers scaling benefits in two ways.
Why it matters here: An earlier release documents moving model weights into cheaper CPU or storage memory and running large models on fewer GPUs. This corroborates improving affordability, although its principal measurements concern hardware requirements and throughput.
- Publisher
- Microsoft DeepSpeed
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: CoreWeave,
launched GooseAI
Why it matters here: CoreWeave advertised lower-cost access to open language models early in 2022. This is a provider comparison rather than proof of successive price reductions or equal model quality.
- Publisher
- CoreWeave
- 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, Bravo @ElonMusk for wanting Twitter to become the world’s “most accurate source of information”
Knockoffs of GPT-3 are getting cheaper and cheaper, which means that the cost of generating misinformation is going to zero, and the quantity of misinformation is going to rise—probably exponentially.
- Where
- paragraph 14
- Published
- 2022-11-09
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-bravo-elonmusk-for-wanting-twitter-7050d7-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-bravo-elonmusk-for-wanting-twitter-7050d7-a1.