Google DeepMind unveils plan to protect itself from its own rogue AI agents (opens in new tab)
For years, AI safety research focused on “alignment.” The Google DeepMind road map assumes some AI agents may go rogue, and focuses on monitoring and access control.
Read the original article