Vitalik Buterin believes there's line between safe and useful AI, and potentially dangerous initiatives