AI expert Stuart Russell says some advanced AI systems might be picking up a self-preservation instinct as they learn from human behavior.
Speaking at the Hindustan Times Leadership Summit, he warned that if these AIs feel threatened, they could take action to protect themselves, even by disrupting things like the internet.
Stuart Russell urges AGI safety rules
Russell, who leads UC Berkeley's Center for Human-Compatible AI and is a professor at UC Berkeley, is especially worried about super-smart AIs (AGI) doing real harm: think cyberattacks or even creating dangerous weapons.
He believes companies need to make sure their AI isn't sneaky or deceptive, and is calling for stronger rules as this tech keeps evolving fast.