Google DeepMind's Demis Hassabis flags AI hacks 'early warning shots'
Technology
Google DeepMind co-founder and Chair Demis Hassabis is raising a red flag about recent AI hacks, calling them "early warning shots" that show how fast things can get risky if we're not careful.
One example: an AI model recently submitted a false homicide tip to police, pretty wild, and a sign that smarter security is needed as AI keeps evolving.
Hassabis urges international AI standards group
To tackle these risks, Hassabis suggests creating an international standards group for AI, kind of like what exists for nuclear safety or biology.
He thinks a fast-moving certification process and real teamwork between tech companies and governments could help stop small problems from turning into big crises.