Slashdot Mirror
Google Secretly Tests Autonomous Cars In Traffic
Hugh Pickens writes "Autonomous cars are years from mass production, but technologists who have long dreamed of them believe that they can transform society as profoundly as the Internet has. Now the NY Times reports that Google has been working in secret on vehicles that can drive themselves, using artificial-intelligence software that can sense anything near the car and mimic the decisions made by a human driver. With someone behind the wheel to take control if something went awry and a technician in the passenger seat to monitor the navigation system, seven test cars have driven 1,000 miles without human intervention and more than 140,000 miles with only occasional human control. One even drove itself down Lombard Street in San Francisco, one of the steepest and curviest streets in the nation. The only accident, engineers said, was when one Google car was rear-ended while stopped at a traffic light."
Update: 10/09 22:37 GMT by T : Reader harrymcc points out that the dream of self-driving cars is nothing new: "Both Popular Science and Popular Mechanics have regularly reported on such experiments; I rounded up some examples dating as far back as 1933."
I have to say that I like the idea of a car driving itself. In theory it should be able to be better than any human. However, software is what I do for a living and it seems there are always circumstances that can not be predicted if software but would be easy for a human to handle. It's those situations that I would be paranoid about if the car was driving itself. The problem would be that even if the human could intervene there is no guarantee that you could intervene fast enough or if the system would even let you.
I think eventually we will get there. I mean I trust the antilock brakes and traction control on my car but those systems are very straightforward with simple goals and it still took a long time to get them right. A car driving itself is ridiculously complex.
It'd be awesome not to need a DD (or risk a DUI) to go to the bar in the many US cities with no or inadequate public transit... though I bet the MADD assholes will lobby to make it still illegal, somehow, and probably try to force a breathalyzer to turn the damn auto-drive on in the first place.
I think a lot of it is trust and acceptance. I would be willing to start small.
Imagine if the leftmost highway lane was designed "auto drive lane". This would greatly simplify the potential scenarios. Vehicle would only auto drive when in the auto drive lane.
Will it pick up hitchhikers?
Will it courteously let people pull out who have been waiting?
Will it flick-off people who drive 30 under?
Will it flick-off people who drive 30 over?
Will it flicker brights to warn of speed traps?
Will it pull over for emergency vehicles?
Will it draft large semis?
Will it bring me hookers and blackjack?
Also, who receives the citation in the event of a stop?
'We are trying to prove ourselves wrong as quickly as possible, because only in that way can we find progress.' RPF
Not everyone can afford to blow $80+ just to get to and from their night out. I practically never go out for that reason, and because I know being a DD sucks and wouldn't impose on someone like that. I go to a bar maybe a couple times a year, but I'd be far more inclined to accompany other friends who go more frequently if the transportation weren't an issue.
As for MADD, they have a history of pursuing policy that has more to do with neo-Prohibition than keeping people safe. I don't dislike them because they're against drunk driving--hooray for that, in fact--but because they appear to be anti-alcohol. My comment about them trying to find some way to make this technology not a legal option for inebriated transportation was serious; I bet they would.
They're a big part of this travesty:
The DUI Exception to the Constitution"
And if I'm not driving, is it any fucking business of yours?
I like to drink. I like to drive. It's really stupid to combine the two, so I do my driving early (to the beer store!) and get it out of the way, and when I get home, it's then that I fire up the grill and have a drink.
I oppose drunk driving. I oppose MADD. My two positions are consistent. Are yours?
I was going to reply to the GP but you said it pretty well. I think some people need a roomful of noisy, drunken strangers screaming at sports on a big-screen TV to enjoy a few drinks. Never really understood that, myself.
I also agree with you about MADD. They've gone completely around the bend, off the deep end, into a bizarre, and completely untenable Prohibitionist position.
The higher the technology, the sharper that two-edged sword.
This doesn't have anything to do with driving. It has to do with being convicted, and going to jail, without being able to mount a defense.
It also has to do with the State using highly suspect technology in order to file that DUI in the first place. The Breathalyzer should never, ever have been accepted by the government for that purpose: too many lives have been destroyed by defective, poorly-maintained, badly-designed and improperly used equipment. The same thing applies to police radar, but the difference there is that a speeding ticket is nowhere near as devastating as a drunk-while-under-the-influence.
The State sees the things as an easy way out, and is willing to tolerate a certain number of false positives (more properly termed "collateral damage" because people can be badly hurt by a false accusation.) I don't drink and drive, but I would refuse a breathalyzer test: if the cop wants to take me to a local hospital and have them give me a blood test (with a sufficient quantity of blood drawn and stored such that my defense attorney could have the test reperformed if necessary) at the State's expense, well, that would be okay. But they don't want that: they want a simple go/no-go test that effectively convicts you, and it's very hard to argue with the results in court. That's because a machine is generally considered more trustworthy and more reliable than any human being. The fact that it may or may not be even remotely accurate is much less relevant to the legal system that it should be.
There was a case a few years ago, where a man accused of a DUI got the court to force the manufacturer of the Breathalyzer unit in question to turn over the embedded controller's source code for independent review. It was apparently so badly written that not only did the man get off, but all the cases where that model was used had to be readjudicated or otherwise reviewed. Ohio, I think, but I'm too tired to look it up. I hope that outfit lost every government contract it had, and they should probably have been made to pay the legal costs of all the people their gadget fucked over.
I've been a software developer for thirty years, and I'll be damned if I'm going to allow the legal system to use someone else's drain-bamaged firmware to convict me of something I did not do. Hell, I hate the fact that cars are so totally dependent upon embedded systems nowadays: makes me more nervous the more lines of code they add.
The higher the technology, the sharper that two-edged sword.
Company filled with really smart curious guys and a bunch of cash.... I think at least one or two of the Microsoft guys are involved in commercial rocketry efforts BTW...
Build it, Drive it, Improve it! Hybridz.org
The issue is that when you have a nonzero number breaking them, you want 40% breaking them. Too many, and it's chaos. Too few, and the few rule breakers cause larger disruptions. 40% has some asses and some who break the rules to essentially cover for the damage the asses would have done. No one should have done a study with 0%, because that doesn't exist in the US, and so couldn't be accurately studied, and if it was accurately studied, wouldn't have any applications. So doing that would be a waste of time and money.
If 0% were to break the rules, then it should flow better than any other number (assuming the rules are changed to take advantage of the possibilities in a 0% breaking scenario), but we can't determine that until we have some way to enforce the rules well enough find out for sure.
Learn to love Alaska
The composition of the driving logic is the most important part. It can't be a big switch case. It has to be a bunch of interconnecting heuristics, constantly looking for every sign of trouble, being able to figure out context and priority of every such signal and also failing gracefully.
There are also tough tradeoffs: It's obvious that if someone's running out in front of the car, you can't go even if the light just turned green, but if a small animal ran out in front of the car, you're doing 110 km/h, every lane is packed and you're on a bridge, you're probably best off actually continuing. It becomes an equation with a thousand terms, solved continuously.