5 ms·

I don't know about you, but I know that I do not want an AI "solving" the trolley problem in real world situations.

lev99 8y ago
It's unavoidable. Let's say a self driving car realizes that it is on an unavoidable collusion course, but changes in it's behavior (steering) will change where it collides. That's the trolley problem.
- SahAssar 8y ago
  If I have understood it correctly in most self driving cars the actual AI/ML part is recognition of the scene/area, not deciding the action or outcome of such a situation. Either way this is one of the reasons I'm not wild about the crazy self-driving race.
  - cpeterso 8y ago
    An autonomous car must decide at every moment whether to brake, accelerate, swerve left, swerve right, or do nothing. Otherwise the car is not very autonomous. :) If it senses an unavoidable collusion ahead, it will need to choose from possible reactions that each have different consequences for the various people and property. Someone has to program or train the car what to value when choosing its reaction. The car will be biased, implicitly or explicitly, by the programmer's values.
    - sterlind 8y ago
      If you make the car flip a coin (perhaps a weighted coin, on the strength of the belief in its sensor data), you avoid providing a solution to the trolley problem. Alternatively, you can have the people vote on a self-driving car referendum, which specifies a DMV-enforced policy for the trolley problem. As a society we vote on such matters all the time. Nothing new here.
      - cpeterso 8y ago
        Flipping a coin is still a conscious decision by the programmer. MIT created an online survey called the Moral Machine asking people to choose who a car should hit in various (randomized and user-created) scenarios. Data like that could be used to train a "fair" AI. It's an interesting thought experiment but the survey's scenarios are unrealistic and include information that a car AI would not have available, e.g. person A is homeless and has no children while person B is a wealthy executive in good health. http://moralmachine.mit.edu/ http://moralmachine.mit.edu/
gowld 8y ago
How is that better than a human solving it? The trolley problem is unsolvable; that's the point.
- SahAssar 8y ago
  The point is that it is a philosophical problem, not a technical one. A human will see that it is unsolvable, whereas a machine will probably optimize for unethical or strictly cost-win payoffs without considering the philosophical parts. I'm not saying the decision will be better when a human makes it, I'm saying it will be more humanely considered.

It's unavoidable. Let's say a self driving car realizes that it is on an unavoidable collusion course, but changes in it's behavior (steering) will change where it collides. That's the trolley problem.

SahAssar 8y ago
If I have understood it correctly in most self driving cars the actual AI/ML part is recognition of the scene/area, not deciding the action or outcome of such a situation. Either way this is one of the reasons I'm not wild about the crazy self-driving race.
- cpeterso 8y ago
  An autonomous car must decide at every moment whether to brake, accelerate, swerve left, swerve right, or do nothing. Otherwise the car is not very autonomous. :) If it senses an unavoidable collusion ahead, it will need to choose from possible reactions that each have different consequences for the various people and property. Someone has to program or train the car what to value when choosing its reaction. The car will be biased, implicitly or explicitly, by the programmer's values.
  - sterlind 8y ago
    If you make the car flip a coin (perhaps a weighted coin, on the strength of the belief in its sensor data), you avoid providing a solution to the trolley problem. Alternatively, you can have the people vote on a self-driving car referendum, which specifies a DMV-enforced policy for the trolley problem. As a society we vote on such matters all the time. Nothing new here.
    - cpeterso 8y ago
      Flipping a coin is still a conscious decision by the programmer. MIT created an online survey called the Moral Machine asking people to choose who a car should hit in various (randomized and user-created) scenarios. Data like that could be used to train a "fair" AI. It's an interesting thought experiment but the survey's scenarios are unrealistic and include information that a car AI would not have available, e.g. person A is homeless and has no children while person B is a wealthy executive in good health. http://moralmachine.mit.edu/ http://moralmachine.mit.edu/

gowld 8y ago

How is that better than a human solving it? The trolley problem is unsolvable; that's the point.

SahAssar 8y ago
The point is that it is a philosophical problem, not a technical one. A human will see that it is unsolvable, whereas a machine will probably optimize for unethical or strictly cost-win payoffs without considering the philosophical parts. I'm not saying the decision will be better when a human makes it, I'm saying it will be more humanely considered.