<i>"While Jalopnik believes that the fender-bender is proof that self-driving cars may not be in the best interests of society, we have a different take. There were 10.2 million traffic accidents in 2008, which results in 39,000 deaths. That’s 17.9 people per 100,000 licensed drivers. If Google self-driving cars can beat those statistics, they could actually prevent more accidents than they create. We also waste millions of hours commuting and driving through traffic; imagine if you had that time to be productive instead."</i>
Also, computers can't get drunk...
I'd love to have these brought to the masses. But, no matter what the statistical benefits are (safer and time saving), I can't picture a time in the near future when normal people will be accepting of automated vehicles. Just like this incident almost was, 1 death caused by a self-driving vehicle is going to overshadow 100,000s of deaths by human-driven vehicles. Even once proved feasible, I can't see something like this being adopted in a matter of years or even decades.
Anyone know of any historical examples of a similarly disruptive, but untrusted technology radically changing people's lives? How long did it take for adoption?
The UK "Locomotive Act" of 1865 required automobiles to carry a crew of three, one of which should walk in front of the automobile carrying a red flag. http://en.wikipedia.org/wiki/Locomotive_Acts
The law was supported by people with interests in the railway and horse-and-carriage industries. It took about 30 years, until 1896, before most of the restrictions were lifted.
I'm pretty sure automated vehicles will become standard at some point, but I expect it will take a few decades before we get there: For the technology to mature, for public opinion to accept the idea and for parties protecting other interests to give up the fight.
The place where it will really hit the fan will be the first time there is an accident where the autonomous vehicle had to choose who lives and who dies.
For instance, a child darts into the road, and due to the speed and the geometry, there are only two possible outcomes: hit the kid, or swerve hit an oncoming car in the next lane. Humans don't have time to make a cold, rational calculation in that situation--we are likely to either hit the breaks immediately (even if it isn't going to stop in time), or swerve immediately, and THEN we notice the oncoming car.
Not so with the computerized car. It will have time to calculate if it can stop in its lane in time, figure out the likely impact speed if it hits the child, forecast the severity of the injury to the child, and then compare that to the likely injury to its passengers from a collision with oncoming traffic, and even make some guesses based on statistics as to how many people are likely to be in the oncoming car. And then it can decide whether to hit the kid or hit the oncoming car.
I wouldn't want to be the engineer trying to explain that algorithm at a wrongful death trial.
This may actually end up being a positive spin for Google. "The car didn't crash when it was self-driven, but it crashed when it was in manual mode! Google technology better than humans at driving"
Prior to that update, it was quite interesting. It's inevitable that the computer controlled car will get in an accident, the only question is if they are better or worse than humans?
I can tell you one thing, they aren't driving in snow in Nevada. And while I'd gladly trust a computer on dry roads when the only things to avoid are hard objects (other cars, tress) and soft objects (people/animals), I get wary on wet or snow covered roads where you have to perceive the road surface itself very clearly.
I imagine a computer would be able to react a lot better than a human would to a hydroplaning / icing condition since the human element of reaction times are no longer a factor, and can measure the car's dynamic forces with much greater precision. I would, however, question whether the car in its current form could get itself out of a small ditch.
I would agree that a computer reaction time is better, but it's prior to the loss of traction that is the difference. I drive a lot in the snow - I spent something like 12,000 miles last year chasing snowstorms to ski in. There are some times it is obvious traction is going to get a lot worse with changing conditions, maybe you are about to drive over a roll over into shade on a borderline freezing day, or the wind has pilied up snow in a section of the road ahead. I can recognize these situations and slow down ahead of time. I find it hard to believe that the computer vision we have today is capable of this based on my experience making autonomous vehicles. Surely they will just set it to go about 5mph the whole way for safety, but that just ignores what modern snow tires & 4 wheel drive are capable of.
15 comments
[ 2.6 ms ] story [ 46.2 ms ] threadAlso, computers can't get drunk...
I'd love to have these brought to the masses. But, no matter what the statistical benefits are (safer and time saving), I can't picture a time in the near future when normal people will be accepting of automated vehicles. Just like this incident almost was, 1 death caused by a self-driving vehicle is going to overshadow 100,000s of deaths by human-driven vehicles. Even once proved feasible, I can't see something like this being adopted in a matter of years or even decades.
Anyone know of any historical examples of a similarly disruptive, but untrusted technology radically changing people's lives? How long did it take for adoption?
They can have bugs or get viruses though.
The law was supported by people with interests in the railway and horse-and-carriage industries. It took about 30 years, until 1896, before most of the restrictions were lifted.
I'm pretty sure automated vehicles will become standard at some point, but I expect it will take a few decades before we get there: For the technology to mature, for public opinion to accept the idea and for parties protecting other interests to give up the fight.
For instance, a child darts into the road, and due to the speed and the geometry, there are only two possible outcomes: hit the kid, or swerve hit an oncoming car in the next lane. Humans don't have time to make a cold, rational calculation in that situation--we are likely to either hit the breaks immediately (even if it isn't going to stop in time), or swerve immediately, and THEN we notice the oncoming car.
Not so with the computerized car. It will have time to calculate if it can stop in its lane in time, figure out the likely impact speed if it hits the child, forecast the severity of the injury to the child, and then compare that to the likely injury to its passengers from a collision with oncoming traffic, and even make some guesses based on statistics as to how many people are likely to be in the oncoming car. And then it can decide whether to hit the kid or hit the oncoming car.
I wouldn't want to be the engineer trying to explain that algorithm at a wrongful death trial.
I can tell you one thing, they aren't driving in snow in Nevada. And while I'd gladly trust a computer on dry roads when the only things to avoid are hard objects (other cars, tress) and soft objects (people/animals), I get wary on wet or snow covered roads where you have to perceive the road surface itself very clearly.
Or does the Captain manually take over control in such a case?
http://www.aopa.org/pilot/features/inflight9905.html
Otherwise, the plane will fight against the storm in ways that are detrimental to the flight.