AI still fails at true flexible reasoning

A new benchmark reveals that AI models are far weaker at flexible, novel reasoning than standard tests suggest.

Also today

Today's top five AI stories, one line each, at 7am Eastern. Free. One-click unsubscribe.