> To be blunt: if the algo kills, why wouldn't you ground the algo (all instances of it)? It depends what the alternative is. If it crashes less than humans, t

> To be blunt: if the algo kills, why wouldn't you ground the algo (all instances of it)? It depends what the alternative is. If it crashes less than humans, then you keep running it despite the flaws.