Uber ends self-driving operation in Arizona
Uber has shuttered its self-driving testing program in Arizona and laid off close to 300 workers there — most of them test drivers, or "vehicle operators" — two months after one of its autonomous cars killed a pedestrian, the company said on Wednesday. The company had been testing its self-driving technology in the state since 2016, but halted operations in the wake of the March crash. The company's testing was also indefinitely suspended by the Arizona governor's office.
[...] Uber says it still plans to restart its self-driving operations in other locations (like Pittsburgh or San Francisco) once the investigations into the Arizona crash are complete. But in those locations, Uber will "drive in a much more limited way," according to an internal email obtained by ArsTechnica.
(Score: 1) by khallow on Friday May 25 2018, @04:23AM (4 children)
And how well does a test environment match the real world for self-driving vehicles? Sounds like it does a rather poor job.
I think a big part of the problem here is that past some simple level of testing, the real world is the sole adequate testing environment.
(Score: 1) by anubi on Friday May 25 2018, @11:22AM (1 child)
For a good test environment, let the website programmers know the CEO is going to be given a random system and plain old consumer grade ISP line and he's going to evaluate the site. It may or may not have ad blockers, or any combination of same a consumer is apt to have on his system.
If it does not load, the programmer does not get paid. If it loads slow, the programmer agrees to a cut of 1% of pay per second.
Won't happen though. I do not know a single CEO that will give up his high-speed direct connection to his corporate system, and I know very few programmers that can still write really clean minimal code without involving a lot of "agile" programming that saves programmer time, but burns bandwidth, memory, and resources like crazy to do it.
Very few CEO's seem to see their site the way their customers see their site. Maybe its a good thing to keep the CEO ignorant of the customer experience, as it keeps his company from expanding too fast and keeps the goal of getting X percent of market share always below what he wants, so he will spend more, in much the same vein that coupon printers love to use the word "Expires" so that the business has to constantly print new coupons, while the recipients are trained to simply toss the coupon instead of saving it.
Businessmen just love to spend money to do the same thing over and over and over. Makes 'em feel important, I suppose.
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 1) by khallow on Friday May 25 2018, @12:06PM
(Score: 2) by Justin Case on Friday May 25 2018, @01:29PM (1 child)
If something simply cannot be done, then you can't do it. So stop already. Don't release code you can't test. Just don't do it.
But I doubt that's the case. If the guys with hardons for SDCs realized the only way we're going to be allowed to proceed is by setting up a thoroughly realistic test environment, they'd do it. If they don't have the skillz to do that, they are unqualified for this project and should be replaced.
The excuse of "well damn this is just too hard so we'll have to start bouncing high speed steel objects off of human skulls" ... can't you see that anyone seriously proposing such an approach needs to be dismissed from the human race?
(Score: 1) by khallow on Saturday May 26 2018, @04:38AM
You're preaching to the choir. They can't build a good enough testing environment to forego real world testing, so they're not doing that. Instead, they're simply doing what can be done.
The real world is indeed a thoroughly realistic test environment. And it's already there.
No, I don't. Instead, I see that this "excuse" is good enough. We're always trying new things and those new things occasionally hurt people. Let us keep in mind that this approach is extremely successful in the real world where it is most needed.