Times of IndiaLean R·
Google DeepMind plans to protect itself from AI Agents going rogue but there's a problem
Google DeepMind is bolstering AI safety by treating advanced agents like potential insider threats, not just tools. Their new 'AI Control Roadmap' outlines a tiered defense, from evaluation to real-time shutdown. This includes using AI to monitor other AI, a method facing scrutiny for potential peer-bias issues. The goal is to ensure these powerful autonomous systems remain under control.
Read at Times of India →
