Scores by kind of claim
Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.
-
Factual claims
Statements about the present or the past, checked against the record: did they hold up?
76 69 82 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
56 42 70 -
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
Insufficient record: 4 promises resolved, 5 needed for a score.
All types combined: 72 (likely 66 to 78, high confidence), from 93 claims: 76 factual claims, 17 predictions. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Dario Amodei's claims
12 claims
“They were not substantially more resource-constrained than US AI companies, and the export controls were not the main factor causing them to "innovate".”
Dario Amodei, essay author 13 October 2022 On DeepSeek and Export Controls (darioamodei.com)
DeepSeek was not substantially more resource-constrained than US AI companies, and export controls were not the main factor causing its innovation.
Mixed Partly true and partly false in roughly equal measure.
The second judge said mostly false; the verdict shown is the first judge's.
What would count as true or false
The recorded date (2022-10-13) predates DeepSeek's prominent models. Judge the substance against DeepSeek's situation around its V3/R1 releases (late 2024 to early 2025), using what was knowable at that time. The claim has two parts: (a) DeepSeek was not substantially more compute- or resource-constrained than US AI companies; (b) export controls were not the main driver of its innovations. TRUE: credible evidence shows DeepSeek's compute and resources were broadly comparable to those of US frontier labs, and its key innovations (such as MLA, MoE efficiency and RL techniques) were not primarily responses to export controls. MOSTLY TRUE: both parts are largely supported, with a minor caveat (for example DeepSeek had somewhat less compute but not dramatically less). MIXED: one part is supported and the other is not. MOSTLY FALSE: DeepSeek had clearly less compute than leading US labs, though not to the degree popularly claimed, or export controls were a significant but not dominant factor. FALSE: credible evidence shows DeepSeek was much more resource-constrained than US labs, and export controls were the main driver of its innovations.
Evidence
- The Rise of DeepSeek: What the Headlines Miss RAND · 28 January 2025
“The compute gap between the United States and China—further widened by export controls—remains DeepSeek's primary constraint.”
- The Rise of DeepSeek: What the Headlines Miss RAND · 28 January 2025
“We don't have short-term fundraising plans. Our problem has never been funding; it's the embargo on high-end chips.”
- The Rise of DeepSeek: What the Headlines Miss RAND · 28 January 2025
“DeepSeek's efficiency gains may have come from previously having access to substantial compute.”
- The Rise of DeepSeek: What the Headlines Miss RAND · 28 January 2025
“China is still running pre-restriction data centers with tens of thousands of chips, while U.S. companies are constructing data centers with hundreds of thousands.”
Reasoning
RAND's commentary does not support part (a). It calls the US-China compute gap, widened by export controls, DeepSeek's 'primary constraint' (E1). It also contrasts China's data centers of tens of thousands of chips with the hundreds of thousands US companies are building (E4). RAND also quotes a statement, without naming the speaker, that the problem 'has never been funding; it's the embargo on high-end chips' (E2). Part (b) gets tentative support: RAND says DeepSeek's efficiency gains may have come from earlier access to substantial compute, not mainly from responding to export controls (E3). With one part not supported and the other partly supported, the claim resolves as mixed.
Cite this claim
Dario Amodei, “DeepSeek was not substantially more resource-constrained than US AI companies, and export controls were not the main factor causing its innovation.” (said 13 October 2022). Verdict: mixed. Reliability API, claim c_b1b8bc01317a, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_b1b8bc01317a
Claims without a verdict
- Unresolvable 157
- Not yet due 40
- Own records only 43
- Not researched 134
- Not checkable 864
Is a verdict here wrong, or do you speak for Dario Amodei? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.