4 ms·

One of Google’s Self-Driving Cars Gets into an Accident

15 points by canistr 15y ago

hundredwatt 15y ago
<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?
- nullentropy 15y ago
  - Also, computers can't get drunk... They can have bugs or get viruses though.
- j-g-faustus 15y ago
  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 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.
- tzs 15y ago
  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.
  - sixcorners 15y ago
    Then it can post to the child's mother's facebook about why it made the decision it did. Social cars! :D
ColinWright 15y ago
From the article: Update: A Google representative tells us that the accident occurred while the vehicle was in manual mode, not self-driving mode So why is this interesting?
- duck 15y ago
  Good question, maybe the lesson is to never use manual mode?
  - nextparadigms 15y ago
    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"
- krschultz 15y ago
  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.
  - canistr 15y ago
    I'm no pilot, but how well do auto-pilots handle storms for planes? Or does the Captain manually take over control in such a case?
    - artmageddon 15y ago
      Yes. Disconnecting autopilot is among the first few items a pilot does in that situation. From the AOPA Guide to In-Flight Emergencies: http://www.aopa.org/pilot/features/inflight9905.html http://www.aopa.org/pilot/features/inflight9905.html Otherwise, the plane will fight against the storm in ways that are detrimental to the flight.

15 points by canistr 15y ago

<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?

nullentropy 15y ago
- Also, computers can't get drunk... They can have bugs or get viruses though.
j-g-faustus 15y ago
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 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.
tzs 15y ago
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.
- sixcorners 15y ago
  Then it can post to the child's mother's facebook about why it made the decision it did. Social cars! :D

ColinWright 15y ago

From the article: Update: A Google representative tells us that the accident occurred while the vehicle was in manual mode, not self-driving mode So why is this interesting?

duck 15y ago
Good question, maybe the lesson is to never use manual mode?
- nextparadigms 15y ago
  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"
krschultz 15y ago
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.
- canistr 15y ago
  I'm no pilot, but how well do auto-pilots handle storms for planes? Or does the Captain manually take over control in such a case?
  - artmageddon 15y ago
    Yes. Disconnecting autopilot is among the first few items a pilot does in that situation. From the AOPA Guide to In-Flight Emergencies: http://www.aopa.org/pilot/features/inflight9905.html http://www.aopa.org/pilot/features/inflight9905.html Otherwise, the plane will fight against the storm in ways that are detrimental to the flight.