> Long as the experiments aren't life-threatening, are you really proposing that they only test new ideas in fake limited artificial environments, rather than i

> Long as the experiments aren't life-threatening, are you really proposing that they only test new ideas in fake limited artificial environments, rather than in real-world environments where they can encounter the full scope of possible failure modes? Are people good at reliably determining which experiments are life-threatening? Especially when those people are under pressure to move fast?