Measuring Is Not Enough Anymore

Per a LessWrong essay, a researcher questions whether measuring AI capabilities is the right strategy for reducing AI existential risk. METR's time horizon graph—tracking how rapidly AI completes complex, long-term tasks—has become influential in AI safety research and government policy. However, the most prominent public citations have come from venture capital firms, particularly Sequoia, using this data to justify accelerating AI investment rather than promoting caution. The author argues this creates a "tragedy of the commons," where safety research inadvertently amplifies the case for faster AI development, and suggests rebalancing such work toward contributions that more directly reduce existential risk.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton