Retread tire strips in the lane, deer carcass, live deer running across, sensor glass fogs up in foggy conditions, snow on the sensors, heavy rain on the road, icy/snowy roads, etc.
I agree, in uneventful day-to-day driving at least, they will be more perfect than us.
Let’s examine this. How does your driving differ between an adult and a toddler walking along the side of the road? Do you slow way down for the toddler? Slam on the brekes? Or are you just way more observant, in case the toddler decides to dart out in front of you? Computers are very good at being observant, and can afford to treat every pedestrian like a toddler.
If someone is walking along a narrow 2-lane road your self-driving car may very well refuse to proceed if there isn’t enough space. How much space will it need? 6 inches? 6 feet? All that will require fine tuning.
As a motorcyclist, I’m very familiar with the human brain’s capacity to completely ignore motorcycles because it isn’t looking for motorcycles. The eye detects the photons, but the brain sees nothing. Driverless cars that actually see me would be wonderful.
OTOH, will motorcycles still be allowed?
Why would there be a driver in a driverless car?
If you’re going to be that rudely simplistic, I can equally well summarise your view as “it’s a computer therefore it is perfect.”
Of course it’s not. If you noticed, I said I could see self-driving cars working very well on motorways and areas with very few pedestrians. If anyone is claiming that their self-driving car will never have problems with pedestrians, then they’re lying.
Do you really not see the logic in my post, the problems that must come up? I mean, the post wasn’t actually that long; I’m sure you can manage to slog your way through it. If you refuse to read a post then don’t bother responding to it either.
I think what friedo is getting at is that, while computers certainly won’t be perfect, they will certainly be better than human drivers. The skeptics keep pointing out where computers might have problems in one situation or another but ignore that human drivers have the same problems, usually worse.
Take levdrakon’s excellent point about motorcycles. A driverless car will almost never miss a motorcycle yet human drivers do it (relatively) frequently. I’ll point out again that over 30k people die every year in traffic accidents. Computers will reduce that number.
Automated cars definitely will be better for motorcyclists and probably for cyclists too. It’s not like I’ve said they’ll never be better in any way at all; I have actually said they will, then been accused of saying “computers bad!”
What I can actually see happening, because really it is going to be difficult for cars to efficiently deal with pedestrians in the (not terribly unlikely) situations I mentioned, is for pedestrians to be futher sidelined and jaywalking laws to be made much more severe, because that’s the one thing that would really make a difference. I don’t like that idea much at all.
And of course the problem is that when automatic cars first start getting used on ordinary city streets rather than just on test drives, they’ll have more accidents than later generations of automatic car would do, because the designers will be able to iron out the bugs better, and that will scare people off, which will make it less likely that there’ll be later generations of automatic car, etc.
It’d be cool to be in a world where automatic cars were all accepted and worked well, but not so cool to be in the beta-testing generation.
Didn’t realize you were talking about driverless Uber cars. I’d say that this won’t happen for a long time, but knowing Uber …
Still, will there be enough spare autonomous cars at rush hour to meet the demand? There would be some from work-at-homers, or those on the late shift, but not nearly enough to make private ownership a thing of the past.
Within cities where the commute is handled by mass transit certainly.
Let me tell you about my office mate in grad school. She and her husband were driving in Texas, when a car stopped in front of them, apparently for no reason. I think it was dark. He stepped on the gas to pass and wham, hit the cow lying in the road. Dead cow, totaled car, and pissed off rancher. So much for our ability to handle exceptional conditions. I suspect that an autonomous car wouldn’t assume the coast was clear.
Retread tire strips are hardly exceptional around here.
Bad weather is one reason why more sensor development is needed, not that our human success rate in these conditions is so great. I saw an article where someone claims there will be cars on the road in 2017. No flippin’ way.
I wasn’t talking about Uber cars at all. I was referencing Uber to point out that the technology to summon and route cars quickly already exists, and can be leveraged in the future. I see no reason why Uber would survive in an era where drivers aren’t needed.
I’m saying that I see driverless cars becoming the next version of mass transit. A small vehicle with 4-8 (or more) cells in it that closes you off from other people in the car, picking up travelers on the way to a common or similar destination could make rush hour as we know it disappear. It works with our existing infrastructure and is scalable.
I really don’t know where this technology will take us, specifically. I just think the potential for a dramatic change in the way we travel is very possible.
I, too, suspect the first AutomatAutos to be deployed en masse are going to be getting rear-ended a lot by higher-primate-controlled vehicles, and otherwise be the object of great road rage, just for strictly observing the law of the road to the last jot and tittle.
(Hmmm… if you are in an autonomous vehicle and are not driving… Can you now use your cellphone handset? Can you now be at >0.08 BAC? Or only if the vehicle has no manual controls? Or does this still come under “operating”?)
The control systems will be quite reliable quite soon IMO. Though if ever it gets to be widely deployed, for my own vehicle (that is, the one that’s my property and I keep at home) I’d still require a manual override that the system/computer cannot veto – oh, it can stay a drive-by-wire car and be governored to not allow the driver to exceed certain performance parameters, but it must be possible for me to deliberately command the vehicle through driver inputs. And of course in the unlikely event something interferes with the driving robotics I’d prefer the system to failsafe to manual control, NOT to “immobilized at the side of the road” if other systems are still good.
Here is a list of already available collision avoidance systems, including pedestrian avoidance systems. This stuff is going to be widely prevalent and proved in long before cars become self-driving.
This sort of thing is often mentioned in discussions of automated cars. People also often say that they’d be fine with the car doing the driving for the boring, predictable parts, but they’d want to be able to take control in an emergency.
But it’s completely wrong. We don’t need computers for boring driving. We need them for exceptional events!
People are already pretty good at driving in normal predictable places. Just staying in your lane and going the same speed isn’t challenging if you’re sober and awake. It’s the emergencies that we suck at, and that computers are great at.
A computer is going to be able to identify a child running after a ball, or a car shifting into your blind spot as you move, or any other unexpected event way faster than you are because it’s got sensors pointing in every direction and it has a reaction time that puts ours to shame. And it will calculate whether swerving or stopping is the better plan, rather than just slamming the brakes and jerking the wheel like humans do when they’re in a high-stress situation.
The tech sector has its nose rammed firmly up its own asscrack. They truly believe they exist on another plane, and can do absolutely anything better than anyone else. All the way up to the point where people like Tim Draper proposed with a straight face the breakup of California into six states, Silicon Valley of course being one of them. I can’t wait until the bursting of Tech Bubble 2.0.
Go read the GD thread I linked to earlier, at least my last post in it. I’m far from uniformed on the subject, though I’m also not “against” computer controlled cars. I just know our ability to make them a reality is rather limited at the moment.
How many of these threads have we had by now?
Short version - the technology is already incrementally creeping in and will continue to. As it becomes more common it will become less expensive.
Enhanced cruise control in which the car also controls staying in the lane and reacting to slowing and accelerating traffic, active crash avoidance systems, lane drift warning and autocorrection, automatic parallel parking … these are seeping into the marketplace now, albeit still at substantial premiums.
Self-driving grows from there rather than being birthed fully formed.
Soon vehicle to infrastructure and vehicle to vehicle communication is included in all cars even those without those enhanced features listed above (and available to retrofit with an insurance discount to older models), allowing cars to be warned as a vehicle ahead of the one in front of the one in front of it is engaging its brakes and either automatically easing off speed or warning the driver or if with the features braking if necessary, warning and/or taking over to prevent collision with that car in the left lane blind spot … and allowing the car to time the lights and to avoid collisions at intersections with poor visibility. Critical mass and self-driving capable cars are allowed access to special lanes, possibly then complete streets or even city districts, within which they form ad hoc “car trains” all accelerating and decelerating at the same time with little space between, allowing many more vehicles during peak travel times driving faster more safely and for less fuel.
So on. The first stages are already happening.