Research · The Decoder ·

METR introduces a new metric to calculate exactly when AI agents become more expensive than humans

METR introduces a new metric to calculate exactly when AI agents become more expensive than humans

METR introduced “expenditure horizon,” a metric that estimates when AI agents become more costly than humans for comparable tasks. Early NanoGPT speedrun results were underwhelming, though the metric has limitations and newer models may produce different results.

Read the full story at The Decoder →