Slashdot Mirror
BMW Showcases Self-Driving Concept Car
SmartAboutThings writes: We've just been given a glimpse of what the future of motoring could look like, with BMW showing off its latest concept car, and it's self-driving. The Vision Next 100 was unveiled on Monday, at a ceremony celebrating BMW's 100th birthday, at Munich's Olympic Hall. This comes just a few days after BMW made official its intentions of competing with Google to build software for Self-Driving cars.
The Vision Next 100 has two driving modes, a driver mode and an autonomous mode, or 'ease' as its known. In driver mode the car operates mostly like cars do now, except the BMW indicates the ideal driving line and speed, but when the car is set to autonomous mode the steering wheel retracts and the two front seats turn to face each other. Perfect for two people to have a chat, and if it's only you in the car, put your feet up and relax.
The Vision Next 100 has two driving modes, a driver mode and an autonomous mode, or 'ease' as its known. In driver mode the car operates mostly like cars do now, except the BMW indicates the ideal driving line and speed, but when the car is set to autonomous mode the steering wheel retracts and the two front seats turn to face each other. Perfect for two people to have a chat, and if it's only you in the car, put your feet up and relax.
If only making a few renderings and CG animations was all that was required to make those driverless car things.
BMW showing off its latest concept car, and it would be self-driving if it actually existed, which it doesn't
FTW.
In other news I will be unveiling my concept airplane/car/submarine/helicopter/spaceship.
systemd is Roko's Basilisk.
My first reaction to the seats turning to face each other was "so if the driver has to take control.... huge delay."
This is one of those things that everyone thinks, but seems completely unreasonable if you think about it a little more deeply. I know that self-driving cars right now require the operators to be attentive and take over when necessary -- because they are still being "trained" in their early phase.
However, this is simply not a reasonable thing to expect the general public to do with a self-driving car. The first mass-market car that truly claims to be "self-driving" should NEVER require human intervention to "take control" suddenly. The car needs to be able to handle all situations adequately, and if it can't (e.g., case of impending major mechanical failure), it should have a backup system that will at least exit the current traffic pattern to the side of the road and come to a halt safely.
Anything less is a huge safety problem.
The idea that a human driver will just "take over" control in an emergency is incredibly stupid and naive. Think about how many accidents are caused by drunk drivers. And most of them are caused because drunk drivers have impaired reaction ability -- and a few tenths of a second is an issue.
Now imagine some rich guy who buys a BMW to drive him to work every day. He'll bit sitting there drinking coffee and reading a newspaper, and you somehow expect him to suddenly "take the wheel" when the AI can't cope? Forget about seats turned sideways or whatever -- you're talking a couple seconds minimum before (1) he realizes there is a problem, (2) manages to get rid of his newspaper, coffee, and breakfast sandwich before he can, (3) hit the button to take control, and (4) actually adjust his body to the controls. And with that amount of delay and breakfast bits flying around the car, there's no guarantee that a human driver would even be able to respond well and make safe maneuvers for a few seconds even after he takes control.
Three seconds at 65 mph is almost the length of a football field. Once we get to the point that drivers expect a car to drive itself MOST of the time, we simply cannot rely on the idea that drivers will be alert and ready to take over at a moment's notice. Moreover, I think it will be actually MORE dangerous to have a human driver take over in most such scenarios when unprepared and emerging from distraction.
So, actually, some ideas about the BMW design are better -- assuming that the AI actually functions flawlessly (which I think we're still a long way from). But once we get that level of AI, it's probably better to encourage people to act like they're not really a "driver" anymore.
My second reaction was "...and if you hit something, you'll be oriented sideways, which seems like a really bad idea."
We already have plenty of buses, shuttles, long limos, etc. which have some sideways oriented seating. I realize those seats aren't generally at the front of the vehicle, but still -- if AI is not demonstrably better than most bus or limo drivers in completely autonomous driving, then it shouldn't be allowed on the roads.
Right it's entirely unreasonable to believe that unforeseen failures can occur those things never happen.
Of course failures can happen. What I'm saying is putting a so-called "self-driving" car on the road which can fail in such a way as to require a person to take over SUDDENLY is a major safety hazard, both for the "driver"/passengers and for those cars around it.
It's not like people, birds, and other wild life have ever ran out in front of cars without warning
If an AI can't respond to such things, it should NOT be allowed on the road. Period.
And if the "failure mode" in these scenarios is to suggest that the distracted businessman reading the newspaper while eating his breakfast is supposed to suddenly take the wheel... well, that's just not reasonable. And you can say, "Oh, but these 'drivers' should be alert," but that's just not how people are going to use a "self-driving" car in the real world.
If the distracted guy on his commute who takes 3 seconds to take the wheel is going to do better than the AI in avoiding "people, birds, and other wildlife," then I definitely would not buy such a car, and I would propose that such an AI is a TERRIBLE driver and should be banned from the roads.
and I have never seen a new car broken down on the side of the road.
Different scenario. Re-read my post. If a major failure happens, there needs to be a backup system that can get the car to the side of the road and to a stop safely. Clearly, in your scenario, the human driver was able to do that. Why precisely should we demand less of an AI "driver"?
I didn't say that mechanical failures would be impossible on future cars -- but the AI should be designed in such a way (with various backup systems) that it's basically the LAST thing that will fail... and if it does so, such a failure should never be sudden and in the middle of a highway.
In all but the most catastrophic of failures, the human should NEVER have to suddenly take control. It's dangerous, and it's a stupid idea.