India, Nov. 14 -- If keeping ever more powerful AI systems under human control is the defining technological question of our time, Berkeley professor Stuart Russell has spent a decade pressing it on the industry he helped build. Can a machine smarter than us be switched off? Whose objectives will it serve? Who should govern the race to build it? Some of the most rigorous answers have come from the Center for Human-Compatible AI, which he founded at Berkeley in 2016. His 2019 book Human Compatible argues that the dangers of superhuman machines can be avoided only if AI is rethought from the ground up. Prior to becoming one of the field's most prominent voices of caution, Russell had helped shape the field itself. His research on probabilisti...