Mongrel wrote:Most crucially, in terms of fatalities-per-mile-driven, Tesla was farthest along (it's not directly stated in the article, but IIRC, Google technically has more miles, but is even now still heavily relying on data collected from cars run purely on fully or partially closed roads or routes under optimal conditions), and even their run has been cut short less than halfway to the distance it would take an autonomous car to simply to EQUAL a human driver, never mind surpass one. Other companies have hit similar failures with far fewer miles driven.
I'm pretty sure Waymo (Google) has fully-autonomous, no-safety-driver-needed cars on the road as we speak, albeit in Phoenix, AZ suburbs with easy-to-handle grids of wide roads. Legally permitted to even give robotic taxi rides to people. (... And they haven't killed anyone because we actually got the tech working
in closed roads before putting them on the road. We're also doing closed roads in Michigan now so that they'll be able to handle shitty weather.)
I think that if anyone can do it, it would probably be Waymo, simply due to sheer weight of data and expertise that Google can bring to bear on analytical problems like this.
But while Waymo has avoided trouble so far by only letting cars out of controlled areas in very small, incremental stages (which is best practices, don't get me wrong), they still have yet to go get out much on untested roads. Back in February, they stated they had logged five million miles on uncontrolled public roads. Which sounds like a lot, but as the article above stated, the Rand study number quoted to equal human drivers is 275
million fatality-free miles.
Only increased exposure will tell if Waymo's systems will be better than or similar to anyone else's in practise. Even Waymo's CEO seems to be poo-poohing expectations for dealing with the unexpected when he talks about what the cars can and can't deal with.
Again, this is all good policy for Waymo, they're going about development in the safest, most responsible way, while companies like Tesla and Uber are out there stretching the tech past its current limits because they've bet their survival on it. But that doesn't mean Waymo is any further ahead in terms of solving the fundamental problems holding back fully autonomous cars.