A robotaxi doesn’t care where it can or can’t drive. It just follows graph search and speed limits.
That means we can design cities around how we want them to look, instead of bending everything around today’s messy car infrastructure.
Think about SF, its size is (famously) around 7 by 7 miles. So it'd take 12 minutes to cross (as the bird flies) from one side to another at 35 mph and 17 minutes at 25mph. Which is completely unrealistic, because real travel times are dominated by traffic lights and congestion.
This calculation changes only when we're talking about long-distance travel on freeways. But honestly, I expect that fast long-distance trains with seamless transfer to self-driving taxis would be a better idea.
...dramatically reduces likelihood and consequences of a crash
(Musk 6 years ago saying it's happening https://www.youtube.com/watch?v=d8X8NdcV7Wc It hasn't yet of course)
I've seen plenty of robotaxi huckers advocate for speed limits 'appropriate for robot response times'
It's part of discussions around hypothetical futures where everything is self-driving and the vehicles communicate with each other to form dense convoys on places like freeways where there aren't pedestrians.
I certainly haven't heard any mainstream suggestions that self-driving taxis ought to drive faster than humans in spaces they share with human drivers and human pedestrians.
Where?
So yeah, they'll do the same thing as humans eventually.
Speeding can usually be brushed off as carelessness. Where it can’t, we charge it more harshly.
A robot programmed to speed serves a jury mens rea on a plate.
What's considered normal for humans, driving higher than the speed limits, will not for automatic cars.