4 ms·

Humans have 1 fatality per 100 million miles, self driving is nowhere close to this.

• Piskvorrr 7y ago
  The statistics for SDVs have an issue which might invalidate them: "fatality per miles driven" doesn't take disengagements into account: how do you even do that, meaningfully? "Fatality avoided because human driver stepped in - another triumph for autonomous driving"? That doesn't make much sense...
  • emiliobumachar 7y ago
    It makes sense in that, even in the real world, the real humans can be counted on to pay attention and intervene to some extent. That's an important counterpoint to the vivid thought experiment of "how could you possibly expect someone to pay attention after hundreds of hours of flawless machine operation?" Unless you're talking about getting rid of the steering wheel and deploying the current system as Level 5. In that case, yes, interventions should count against it.