The Decoder· Maximilian Schreiner·· 4 h ago
METR finds Anthropic may have crossed its own AI acceleration warning threshold back in July
METR finds Anthropic may have crossed its own AI acceleration warning threshold back in July
AI summary
Nonprofit METR spent eleven weeks auditing Anthropic's internal research data and concluded that the pace of model development has roughly doubled, with about 80 percent of the speedup driven by AI-assisted research. Anthropic redacted the most informative evidence, though, which means the public has to rely on METR's judgment "much more than would be ideal."
Selection record
Threshold 76Media and individualsFirst 78Second 78
AdmittedSum of both 156 ≥ twice the threshold 152
- Source tier
- Media and individuals; this tier's threshold is 76
- Pre-filter
- passed：METR审计Anthropic模型加速与AI研发影响
- Why it was chosen
- A third-party audit uses internal data to estimate that Anthropic's model capability has been improving about twice as fast, and suggests public risk reports may understate the acceleration.
A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.
Source: The Decoder · the-decoder.com