← All news

METR updates its AI capability graph: fast progress, but limits in judgment

The AI evaluation nonprofit METR has updated its time horizon plot, which tracks the length of tasks AI can handle. The report finds that some capabilities keep improving exponentially. It also indicates that models are less strong at tasks requiring judgment than at coding.

Read the full story at · MIT Technology Review ↗

Get the news in your inbox

A weekly AI digest for businesses. Unsubscribe with one click.