Uber’s self-driving car unit was burning $20M a month
techcrunch.com
techcrunch.com
That's all that ever needs to be said about Uber's self driving system, the company should be disbanded and whomever the highest-level person in the company knew vehicles were operating in such a way should be imprisoned.
1: https://techcrunch.com/2018/05/24/uber-in-fatal-crash-detect...
I have knowledge of mining in Chile, where things are thus: One dead "viejo", and there's an investigation by the government, the mine will need to take more protocols and invest more in safety, make some pretty good explanations, and even people in management could go to jail for some time. It shakes the mining operation to its foundations, which should ideally be safe but is still intrinsically risky to miners. So in a way it is understood to be a risk and the workers know this to some degree, so the mine can continue operating after that. Two dead "viejos?" That's game over. The mine (as in the "yacimiento," the underlying mineral formation) is thereafter not to be exploited.
(Miners in Chile are universally known as "viejo's", regardless of whether they're actually old or not.)
But ... for a project like this, it's the kind of thing they should have planned for and could have easily worked around.
First of all, Uber's setup was to have one person do both the analytics monitoring and safety roles. That's ridiculous. The safety person needs to focus on that alone.
Second, for any boring job, it's expected that people will tend to mentally check out, and you should design it so that a single checkout doesn't cascade to death. They could have had a periodic "press this button to prove you're paying attention", or limited the employee from bringing any device except a limited one with no distractions.
So yeah, she shouldn't have been watching a movie, but Uber shouldn't have had better safeguards for the inevitable attention drift.
I agree that victim blaming is _not_ useful when trying to figure out why a safety critical application failed and should not even be in the equation when developing, operating, or maintaining such safety-critical systems.
But to simply answer the question "Who is to blame for her death?" the blame does _not_ rest 100% solely on Uber, nor the driver, nor the victim. All 3 failed to prevent the accident (and all 3 had the power and indeed responsibility to prevent it).
People are responsible for their own risk management. If I take an unnecessary risk (like jaywalking at night on a freeway) and I get hit by a car (autonomous or not), that's partially on me.
So you think these companies should be shutdown immediately ? Mark Z in jail ?
For example, from [0]:
'Across India, social media rumors have caused rural villagers to patrol in anxious groups on the look-out for anyone they don’t recognize. These mobs have already killed numerous people. In May and June alone, at least six people died in WhatsApp-related mob attacks in eastern Assam, western Maharashtra and southern Tamil Nadu. There’s also simmering tensions over Hindu vigilante groups who’ve targeted and killed Muslims.'
[0] https://www.bloomberg.com/news/articles/2018-06-20/one-cop-s...
For instance, everybody seems to agree that lidar is the most important sensor for a truly driverless vehicle. And everybody agrees it's too expensive to sell to consumers. They're orthogonal approaches to similar problems with surprisingly little overlap. It's like comparing game development for a 2D sidescroller on the Nintendo Switch with a 3D first person shooter for a gaming PC. They're completely different technologically, even though at the end they're both making games.
When you come to trust these systems, you cease to verify if they're working properly. Unfortunately, the system will eventually fail, you'll have a false negative, and you will have been since trained into not looking into your blind spot the tradition way because 'the machine did it for you.'
You will potentially cause an accident. Same thing with lane assist, same thing with adaptive cruise control. Perhaps not you, specifically, but some less vigilant driver.
http://pages.cs.wisc.edu/~gdguo/driving/BlindSpot.htm
People don't like to do this because then they can't see the side of their own car in the mirror.
Perhaps that should be added to driving tests. You shouldn't get a license if you don't know how to make the car safe to drive in.
Am I wrong in stating it's because people want to see the side of their car? It was for me when I first started driving and for everyone else I've asked.
Is it because of my belief that people should only be licensed to drive if they can do so safely?
If I'm wrong, please give me the opportunity to learn what you believe is right. I'm open to improving myself, and recognize that I am not, nor ever will be 100% right in everything.
Many people might reasonably believe that the traditional alignment we were taught, with the look-over-the-shoulder to check the blind spot that we were also traditionally taught, is safe. It comes across as aggressive when you suggest no-one should be allowed a license without converting to your new method.
In 20 years we will have the Wolf of SV movie about Uber.
Is this not Sillicon Valley already, the TV Show?
The reason is (the potential for) self driving cars. Once/if there is mature self driving tech, they can eliminate the cost of having to pay human drivers. When you consider Uber in that light, as long as they don't run out of money before self driving maturity, then their high valuation makes sense.
This may seem obvious to some but this fact has been entirely absent from every conversation I have heard about the valuation of uber's IPO ( looking at you Marketplace's "Make Me Smart" ;). It seems like it is taboo to point out that Uber needs to eventually eliminate all their human drivers from the equation.
With that said, I think it's wrong. There is significant competition on the way to making a working SDC. And Uber will not be alone in producing one. If and when that happens, yes, they can pay a lot less for their inputs. But the output -- the value of the ride -- will also fall in value, since their competitors don't deal with that cost either! It doesn't translate into superprofits.
At the most optimistic, Uber might arrive at a street-ready SDC two years before anyone else. Two years of above-normal profits, equal to keeping the revenue they'd pay human drivers, it just not enough to pay back the costs of the program and extreme returns investors demand.
IMHO, it would make much more sense for Uber to just license the SDC tech as it becomes available, playing the vendors off each other, and focus on making their piece of the product -- the app/customer experience -- better.
Without a moat, there will be little to no long-term profits.
Users are not loyal to a particular ride-sharing app. It's incredibly trivial to switch, and if the new hot autonomous company can offer rides for half the price of Uber, users will switch.
I mean, isn't it likely that a new company/startup will show up when self-driving systems are robust in 5-10-15 years?