They might not be able to make moral choices but humans aren't going to be making many meaningful moral choices either during the few seconds of a traffic accident. Also, many humans will choose wrong. What is probably a given though is that autonomous driving will likely drastically reduce the number of deaths and injuries overall.Who exactly is "we" and how do "they" program "their" ethics? What is "right" and what is "wrong" and do we as individuals have any say besides "well just don't drive then"?
Explain how a car will be able to make moral choices in objective reality than you or I? No AI of any magnitude in the real world has become a superior moral being. How does that work in less than vague terms? What effective philosophy will your car have?
These aren't questions I alone am asking, and no not by crockpots. This is one facet of a larger concern in terms of AI's who control us.
Can we make improvements? Sure. Is automated driving "evil"? Of course not, but I see a lack of appreciation for the complexities in more than just faster computational ability terms.
When should your car tell you when to die and why is it morally superior to you to make that decision?
Questions that need to be answered.
The one our laws set forth.
Frankly, it'd be foolish to design a vehicle to purposefully swerve out of control to avoid a person, even a child. You'd be putting everyone else around you at risk and still potentially failing at the original objective. OTOH, I can see future generations of this automation being capable of regaining control of the vehicle in situations that would have been impossible for humans to survive. But I digress....
You think it's wrong to have a cold, calculated, and socially agreed upon answer to no win scenarios, in exchange for saving 40,000 annually? Let's say you will not change the no win scenario, but you CAN save those 40,000 people.
What is probably a given though is that autonomous driving will likely drastically reduce the number of deaths and injuries overall.
Who exactly is "we" and how do "they" program "their" ethics? What is "right" and what is "wrong" and do we as individuals have any say besides "well just don't drive then"?
Explain how a car will be able to make moral choices in objective reality than you or I? No AI of any magnitude in the real world has become a superior moral being. How does that work in less than vague terms? What effective philosophy will your car have?
These aren't questions I alone am asking, and no not by crockpots. This is one facet of a larger concern in terms of AI's who control us.
Can we make improvements? Sure. Is automated driving "evil"? Of course not, but I see a lack of appreciation for the complexities in more than just faster computational ability terms.
When should your car tell you when to die and why is it morally superior to you to make that decision?
Questions that need to be answered.
They might not be able to make moral choices but humans aren't going to be making many meaningful moral choices either during the few seconds of a traffic accident. Also, many humans will choose wrong. What is probably a given though is that autonomous driving will likely drastically reduce the number of deaths and injuries overall.
Humans will not make many purposeful choice that surface to awareness, yet the mind isn't that simplistic. A part fails, a crowd looms and into a wall the car goes. That was a response, a choice, not mulled over but made at some level. It seems the answer is to avoid the question.
Interesting questions but given my experience on the road, I don't think humans make these snap decisions. In fact I don't think a lot of human drivers even think about what is happening to them on the road, they do as they please and make everyone else acquiesce to them. If the program was told to make the decision that took the least amount of life, I would expect the AI to react much better than a human, who is likely to choose their own life over others.
Tell me, what is the correct choice in your scenario. The wall, or the crowd?
Yes, that's correct, and generally where the idea of self driving cars is headed.
And that will come, probably sooner than later. Technology like this advanced at a exponential rate. It will be only a few years before antonymous cars are considerably better drivers than even the best trained human drivers. They have a lot of advantages over humans after all. Better sensors, 360 awareness, no distractions, faster decision making, faster reactions.
The rule of no jaywalking is a logical conclusion based on the fact that a heavy vehicle traveling at speed is very limited in it's ability to stop or avoid a pedestrian that can enter the road at any point with no warning.
Interesting questions but given my experience on the road, I don't think humans make these snap decisions. In fact I don't think a lot of human drivers even think about what is happening to them on the road, they do as they please and make everyone else acquiesce to them. If the program was told to make the decision that took the least amount of life, I would expect the AI to react much better than a human, who is likely to choose their own life over others.
In my stated scenario, having little driver intervention is enough.Yeah, but the last part of that transition is a huge leap. There's a qualitative difference between full autonomy and some sort of driver assistance. And, as with automation of aircraft, that transition zone is quite a dangerous one.
The AI has no self awareness, but it must make choices. I don't think I'm unique but I have had the situation arise where I risked my life to avoid killing a pedestrian. Again I didn't ponder the situation, but me, as an entity, did just that. No one else has? I have a hard time believing that.
There is a percentage of the population that would act like that, there's also a percentage of the population who drive with disregard for any life other than their own, and in the worst cases, complete disregard for all life including their own. I'm very familiar with the latter two, I live 2 miles from the university campus here in San Antonio and I run into maybe 10-15 of these drivers every day on the drive home.
Your question is a bit of an ambiguous one, there isn't a right or wrong decision to make when the loss of life is guaranteed. There are only options and consequences, as a human we decide which consequence is more palatable. From a neutral perspective minimizing the loss of life would be the most ethical, but from a highly subjective perspective I would choose my life and my family's life over a random pedestrian. Is either better than the other? That's going to depend on which party is on the end of either set of consequences.
So lets say the car harms someone. If the occupant has not control he bears no responsibility any more than any non-driving passenger. Who then, the car? What justice can be had from the machine? The programmers? They designed an imperfect system so perhaps, but then they had no ability to decide at the moment of impact. The Board? I think that unlikely. Stockholders?
Who is ultimately accountable an a material way?
I think you might find that paper I linked to interesting. There's much that goes beyond getting from point A to B and I see no wisdom in pretending otherwise.
Accountability remains a serious concern for all forms of automation. It's a far more complex issue than safety, and has implications that spread beyond just the issue at hand. I don't doubt that a GOP controlled congress and conservative SCOTUS would take actions to protect corporate interests and shield them from legal repercussions.
The long term solution is of course to completely be rid of human driven cars and switch to full automation. That would allow the algorithms used in these cars to operate at peak efficiency and hopefully ultimately be orders of magnitude safer than human drivers. What to do during the transition is the really difficult question to answer, and it's a question our leaders need to start asking themselves now and not later. The transition to automation needs to be led by those with the public interest in mind, before the financial power is allowed to flex it's muscles and control the dialogue.
Did you happen to read the paper I linked to? Might be worth the effort to help understand the topic in more terms than automation.
How does a human decide who to kill when here is no alternative? What does that question really have to do with self-driving cars?How does a machine decide who to kill when there is no alternative?
How does a human decide who to kill when here is no alternative? What does that question really have to do with self-driving cars?
It will decide based on a human's programming.So the car will run into a wall first?
Yeah that has everything to do with the discussion.
It will decide based on a human's programming.
Done.
Soon after the tipping point, the number of times when such a decision will need to be made drops to 0.
You'll have mostly-automated construction of a transportation network that consists of interconnected underground grids.
Not infallible. Inevitable.You have begged the question. You say that there won't be problems and then reference yourself by saying there won't be problems.
You are implying infallibility, and minimizing the effects on real people. I want to meet this perfect programmer with the infallibility of God.
Not infallible. Inevitable.
It's an annoying discussion. We know there will be those situations. There absolutely will be. Engineers will do their best to account for it.
Progress will happen regardless. That dilemma has been discussed to death and it's tiresome.
Soon after the tipping point and the rapid transformation of our transportation infrastructure, there will be virtually no pedestrian / vehicle interaction during a trip. It will be comparable to subway travel in that regard.
Automation will happen. Safety, efficiency, and convenience will be significantly improved for riders and for pedestrians.