AI agents can't reliably simulate human values

LLM agents frequently fail to faithfully represent diverse human value systems, even before any drift occurs over time.

Also today

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