How awful it was that 12 people were killed, apparently on purpose, by someone using a stolen lorry as a weapon. But actually, how awful?
To assess awfulness, we have to make a comparison, and the obvious figure to compare, is how many people were killed on the roads in Germany? In 2013, there were 3540 road deaths.
That's about 10 per day.
I have to say that this is a *much* bigger problem than the recent crime in Berlin. And that is why we need self-driving cars, and we need them as soon as possible.
Our legislators have to move urgently to create law on liability in case of accidents (because there surely will be some). If I'm a passenger, then I have no liability, but if something goes wrong, then it may well be that no-one is at fault (for example, if the car is struck by lightning), or that someone is at fault, in which case is it the car manufacturer, the software designer, or who? With that legislated, insurance becomes possible, and the costs can be built in to the price of the vehicle.
The biggest incentive is the much lower accident and death rates we'll get when cars are driven by software that doesn't get distracted, doesn't fall asleep, doesn't drive while drunk and doesn't text on phones while driving.
And it should also be feasible to make it impossible to hijack a lorry and drive it into a crowd.
Self driving cars may work in a controlled environment with additional environmental inputs such as smart traffic signals and cars that communicate with each other. But we are nowhere near that stage at the moment. Where I live we have single lane country roads, with potholes galore and ill defined edges, where 2 cars can barely pass each other with the additional hazards of horses, cyclists, combine harvesters in summer and tractors most of the year. I often imagine the decision making the code would need to deploy in certain situations I observe. My conclusion is often a stalemate with competing decisions causing the vehicle logic to give up and hand control to the driver. Like you, I have been in IT and software for decades and I think I understand how complex the problem of self-drive cars is (try automating a Raspberry pi on a RC vehicle chassis and see how it copes). The road system and human behaviour would need significant modification to allow SD cars to really work. The other issue is modern 'Agile' software delivery methods. Where software is poorly designed, poorly tested and released too quickly in effect making users the beta testers... Unless the automotive industry starts to work like the aviation industry I think we'll see an initial rise in deaths while all the bugs are ironed out. I guess Elon Musk may be the guy to do this with Space-X and Tesla though. I would challenge any of his vehicles to self drive around the narrow country lanes of the UK. I am not saying never but I think we are 25 to 30 years from universal self drive. Finally self drive could make terrorist activity easier as terrorists no doubt will find the ability to hack into vehicles' systems and control them from anywhere on the globe. Imagine a virus/malware introduced to a vehicle's OS (and yes I think they will be daft enough to make the systems open and connected enough to allow that to happen).
ReplyDelete