So here's one thing about this incident where humans do have a system for this already.
People are very good at reading body language. It's possible to gather someone's likely intentions (wrt moving) in a near instant. A defensive driver will see body language that could dart out into the street and slow down, just in case, and give more attention to the person. We are really good at this.
I know you can say it was dark and therefore maybe harder to see this, but a self-driving car shouldn't take its eyes off anything anyway, so that seems irrelevant. Unless it's almost pitch black, people can read this sort of information.
So what gives? Uber's self-driving system clearly lacks the full suite of decisions made by human drivers. This makes me skeptical that their cars can be significantly safer than human drivers.
How do you know that this is not possible? It is not useful to make broad generalizations about the field because of a single company, especially with so little information.
The chief of police says: "“It’s very clear it would have been difficult to avoid this collision in any kind of mode [autonomous or human-driven] based on how she came from the shadows right into the roadway,”"
It's cliche in America to absolve drivers of any culpability when killing pedestrians and cyclists let alone a case involving a local government that was actively courting a known unethical company.
Until they release footage I'll be taking anything from Uber and Tempe officials with a grain of salt, they'll spin this as hard they can.
P.S. I'm surprised he used the word shadows considering the car was equipped with a fancy lidar unit.
> "It’s very clear it would have been difficult to avoid this collision in any kind of mode [autonomous or human-driven] based on how she came from the shadows right into the roadway,"
How the hell would the chief of police know how to estimate what is possible for an autonomous vehicle to avoid?
Has he even seen that video of the Tesla predicting an accident two cars ahead?
People are very good at reading body language. It's possible to gather someone's likely intentions (wrt moving) in a near instant. A defensive driver will see body language that could dart out into the street and slow down, just in case, and give more attention to the person. We are really good at this.
I know you can say it was dark and therefore maybe harder to see this, but a self-driving car shouldn't take its eyes off anything anyway, so that seems irrelevant. Unless it's almost pitch black, people can read this sort of information.
So what gives? Uber's self-driving system clearly lacks the full suite of decisions made by human drivers. This makes me skeptical that their cars can be significantly safer than human drivers.