What happens when AI passes through the ‘uncanny valley’?

Robots are close to being so convincing that we can’t tell them apart from humans — and that could be a problem