Author Topic: Self-Driving Cars  (Read 41677 times)

0 Members and 2 Guests are viewing this topic.

Offline brilligtove

  • Stopped Going Outside
  • *******
  • Posts: 4600
  • Ignorance can be cured. Stupidity, you deal with.
    • Valuum
Re: Self-Driving Cars
« Reply #30 on: April 15, 2015, 05:12:39 PM »
I sincerely doubt that an appropriately programmed computer will be able to write poetry well at all because that requires lateral thinking of the variety that computers are not good at.

With the self-driving cars issue, I can't help but think of the classic XKCD comic...



Driving strikes me as something that's stealthy-hard. You have to account for lots and lots of variables, and account for them so quickly that you don't consciously know that you accounted for them until you're after the fact. Figuring out what all of them are is going to be tough and I agree with TheLatinist that except in very specific instances (for instance, if you were to create a stretch of road that was fully automated for all vehicles involved), we're not particularly close to the point to where we've accounted for enough of said variables that it'll outdo humanity.

The comic isn't wrong, but we really have come a long way in this field.  The IEEE has an entire weekly newsletter dedicated to this space.
http://spectrum.ieee.org/blog/cars-that-think
You may be right, Johnny Slick. Then again, I can do that search for a bird in my Google Photos right now, and it did come back with a picture of a bird I took last month. The XKCD comic is #1425 from six months ago, BTW.

I'm sure that there are many specific conditions that will need to be dealt with using specific algorithms or tweaks. At the same time, the car probably doesn't need to use the kind of algorithms that humans use for a lot of that. It seems like the folks building these things are getting pretty good at building a system that can interpret realtime sensory data, and from an array of senses that are much broader and better than our own.

In any case, I'm in the camp that thinks these will become part of our culture the way mobile phones did. The adoption curve may be slower presuming that people keep cars for longer than they keep phones. I'm actually more concerned about hacking and emergent behaviours than I am about whether individual cars are safe drivers. 
evidence trumps experience | performance over perfection | responsibility – authority = scapegoat | emotions motivate; data doesn't

Offline SnarlPatrick

  • Too Much Spare Time
  • ********
  • Posts: 7315
Re: Self-Driving Cars
« Reply #31 on: April 15, 2015, 08:52:09 PM »
This can't come soon enough for me. I find driving to be enormously stressful, especially at night, when my vision is not particularly good. I don't normally do this, but "trigger warning, car accident"

When I was a teenager, I was the passenger in a car, when we struck a old women who was wandering in the road, in a dementia induced fog. We arrived at the top of a hill, and there she was right in front of us. I saw her with just enough to time scream "STOP", the driver managed to break, but only very slightly. Her head struck the windshield directly in front of my face, with a sound I will not describe, and sent cracks throughout the glass, she was thrown several meters down the road. 

I heard someone screaming, and then realized it was me. For a few moments, I simply stared ahead, at the skull shaped dent in the glass, the pieces of hair and gore, maybe 6 inches from my face. Then everything went blank, I completely failed to provide first aid, and ran several blocks away, and curled up into a quivering dissociative ball until a police officer found me.

Fortunately, there was a bus stop nearbye, with a nurse present, who was able to provide first aid, and she lived. But I see this image over and over and over again when I am at the wheel. If I had a self-driving car, I would go many more places than I do now.
« Last Edit: April 16, 2015, 05:37:06 AM by SnarlPatrick »
The smallest minority on earth is the individual. Those who deny individual rights cannot claim to be defenders of minorities. - Ayn Rand

Offline Kay-Ole

  • Brand New
  • Posts: 9
Re: Self-Driving Cars
« Reply #32 on: April 17, 2015, 06:58:11 PM »
I wrote this in the thread for episode #509, but it's relevant for this thread too, so I will do a copy/paste.

"There is one important point about driverless cars that has yet to be discussed in the SGU. It was brought to my attention by none other than Jeremy Clarkson, who usually has the scientific sophistication of an uneducated orangutang.

Anyway, the ethical questions that will be raised in programming these cars is quite substantial. In the unlikely event that five children suddenly finds their way into the path of the car, and all other measures for noticing them in time has failed; what will the car "choose" to do? Say the only option for saving these children is swaying off the road, but this will certainly lead to the death of the person in the car. This is of course a variation over the old, well known thought experiments on human ethics, but the fact of the matter is someone will actually have to program the cars computers for this eventuality."

Offline kvuo75

  • Well Established
  • *****
  • Posts: 1126
Re: Self-Driving Cars
« Reply #33 on: April 17, 2015, 09:26:05 PM »
I wrote this in the thread for episode #509, but it's relevant for this thread too, so I will do a copy/paste.

"There is one important point about driverless cars that has yet to be discussed in the SGU. It was brought to my attention by none other than Jeremy Clarkson, who usually has the scientific sophistication of an uneducated orangutang.

Anyway, the ethical questions that will be raised in programming these cars is quite substantial. In the unlikely event that five children suddenly finds their way into the path of the car, and all other measures for noticing them in time has failed; what will the car "choose" to do? Say the only option for saving these children is swaying off the road, but this will certainly lead to the death of the person in the car. This is of course a variation over the old, well known thought experiments on human ethics, but the fact of the matter is someone will actually have to program the cars computers for this eventuality."


i think as current law usually allows. roads are dangerous..  the pedestrians get the benefit of the doubt, unless they are doing something unlawful.

eg. if you are in a car and someone is standing in the road, you have to stop.  if someone jumps in front of your car, they cant expect to not get hit. 

you could just program it to behave like a typical reasonable motorist, and plow into the pedestrians.. status quo.

i might not be thinking it through enough, but i dont see it as a big thing.  a computer could certainly make whatever decision it needed to quicker, so already we're ahead of a human driver.







Offline SnarlPatrick

  • Too Much Spare Time
  • ********
  • Posts: 7315
Re: Self-Driving Cars
« Reply #34 on: April 20, 2015, 03:51:45 AM »
Yeah idk.... i think the idea that the car sensors would be so precise, that they could weigh the relative risks of different actions... i mean... you'd need sensors that counted the number of occupants of cars, and then weighed the relative value of people, injuries to pedestrians vs occupants of vehicles with restraints? I think thats hugely optimistic for the near future. People don't do a great job of this. Computers should be able to do about as well. There might well be room for an ethicist in the design process, but I'm not sure it is going to be that difficult a problem, just because we're going to have limited functionality at first. When in doubt... err in the the side of utilitarianism i suppose?
The smallest minority on earth is the individual. Those who deny individual rights cannot claim to be defenders of minorities. - Ayn Rand

Offline Kay-Ole

  • Brand New
  • Posts: 9
Re: Self-Driving Cars
« Reply #35 on: April 21, 2015, 08:36:11 PM »
Well, the computers will be as good at this as the humans that programme them. And that is the point. There is no computer code for "Asses human value correctly" that will make computers magically do the right thing. But these issues will need to be adressed when the computer for the cars behaviour are going to be programmed.

There are allready driverless cars to some degree, and I'm sure there are plenty of prototypes of computers allready out there, so it would be interesting to know how these issues are adressed.

Offline brilligtove

  • Stopped Going Outside
  • *******
  • Posts: 4600
  • Ignorance can be cured. Stupidity, you deal with.
    • Valuum
Re: Self-Driving Cars
« Reply #36 on: April 22, 2015, 09:11:05 AM »
Yeah idk.... i think the idea that the car sensors would be so precise, that they could weigh the relative risks of different actions... i mean... you'd need sensors that counted the number of occupants of cars, and then weighed the relative value of people, injuries to pedestrians vs occupants of vehicles with restraints? I think thats hugely optimistic for the near future. People don't do a great job of this. Computers should be able to do about as well. There might well be room for an ethicist in the design process, but I'm not sure it is going to be that difficult a problem, just because we're going to have limited functionality at first. When in doubt... err in the the side of utilitarianism i suppose?

It's not that the sensors will be so precise. Heck, our human sensors are crap wrapped up in a cloudy bag. The computers don't have to make great moral judgements either. Accidents happen when
  • something in a control system goes wrong (driver error)
  • something in the environment happens that the control system can't handle (moose on road)
  • something in the vehicle fails in a mode that is not safe (tire falls off on highway while changing lanes going downhill in the rain on a bend with a cliff down on one side and up on the other)
The control systems in driverless cars will be far less prone to driver error in part because of better sensors than humans have (such as radar, lidar, and IR cameras), and in part because they are much less likely to be distracted or overwhelmed.

There will always be environmental and vehicle conditions that a car can't handle, whether driven by a human or by itself. Some of these can be posed as moral questions with absolute answers. This is not a fair way to go about it, since there is no right answer to program into the car. We use avoidance and mitigation strategies as best we can; asking 'hit the child or hit the adult' can't be addressed, but minimize acceleration to anything with human body temperature and size - that could be taught.

Well, the computers will be as good at this as the humans that programme them. And that is the point. There is no computer code for "Asses human value correctly" that will make computers magically do the right thing. But these issues will need to be adressed when the computer for the cars behaviour are going to be programmed.

There are allready driverless cars to some degree, and I'm sure there are plenty of prototypes of computers allready out there, so it would be interesting to know how these issues are adressed.

There is no factual basis to assert that a computer can only be as good as it programmers at some cognitive task. Deep Blue. Google search. Watson. Humans make tools that are better than us in specific domains all the time. (Your typewriter had better handwriting than you.) It's part of what makes us human.
evidence trumps experience | performance over perfection | responsibility – authority = scapegoat | emotions motivate; data doesn't

Offline kvuo75

  • Well Established
  • *****
  • Posts: 1126
Re: Self-Driving Cars
« Reply #37 on: April 22, 2015, 09:56:48 AM »
There is no computer code for "Asses human value correctly" that will make computers magically do the right thing.

people don't magically do the right thing, either. in fact, they do the exact wrong thing, constantly.

Offline brilligtove

  • Stopped Going Outside
  • *******
  • Posts: 4600
  • Ignorance can be cured. Stupidity, you deal with.
    • Valuum
Re: Self-Driving Cars
« Reply #38 on: April 22, 2015, 10:33:03 AM »
There is no computer code for "Asses human value correctly" that will make computers magically do the right thing.

people don't magically do the right thing, either. in fact, we they do the exact wrong thing, constantly and consistently.

FTFY. ;) Good point to keep in mind.
evidence trumps experience | performance over perfection | responsibility – authority = scapegoat | emotions motivate; data doesn't

Offline PANTS!

  • One leg at a time.
  • Reef Tank Owner
  • *********
  • Posts: 8763
  • What seals? I auditioned for this job.
Re: Self-Driving Cars
« Reply #39 on: April 22, 2015, 11:21:17 AM »
Yeah - I think this is a non existent problem.  No way will the computer ever be given the choice to swerve off the road or otherwise put the passengers in danger.  It certainly will otherwise try to avoid any collision.  I mean what human realistically in the moment does anything else anyway.  I reject the idea that a driver when faced with a collision that is such a surprise it can not be avoided has time to asses the moral weight of thier life vs. the worth of the objects/beings they are about to strike.
Now where I come from
We don't let society tell us how it's supposed to be
-Uptown, Prince 👉

We cross our bridges when we come to them and burn them behind us, with nothing to show for our progress except a memory of the smell of smoke, and a presumption that once our eyes watered.

Offline teethering

  • 37% better than Slick
  • Poster of Extraordinary Magnitude
  • **********
  • Posts: 10813
Re: Self-Driving Cars
« Reply #40 on: April 22, 2015, 11:47:48 AM »
There are already driverless subway and metro systems in the world functioning today.  So I'm not convinced legal or liability barriers are really that big of a deal.  I think the key thing is proving the car technology to be just as safe, but as a category it should be fine as we have similar precedents.

However, apparently technology is proving to be really hard.  Harder than people estimated originally.  On the one hand that's not too surprising, given that it's in the sort of area (machine intelligence) where we've been terrible at predicting things on the too optimistic side.  On the other hand, you'd think we'd learn by now.

Offline Jeremy's Sea

  • Kintsukuroi, baby.
  • Frequent Poster
  • ******
  • Posts: 3160
  • 667 - Neighbor of the beast.
    • jeremyad [sic]
Re: Self-Driving Cars
« Reply #41 on: April 22, 2015, 03:28:17 PM »
I realize it's no cold comfort to a parent whose child has been plastered by a self driving car, but while there will still be accidents, there will be tens of thousands of lives saved by sleepy/drunk/anxious/inattentive drivers not having to operate a vehicle they have no business being in. Every year an elderly person plows into a crowd at a mall because they think they're braking while hitting the accelerator. With a multiple system redundancy this will be so vanishingly small of a chance as to not have to worry about it.
It seems to me any moral dilemma presented needs to look at the larger picture. Don't let perfection get in the way of good enough, or better?
Knowledge is power. France is bacon.

Offline Morvis13

  • Big Ol' Goober
  • Planetary Skeptic
  • *
  • Posts: 21970
  • Natural Source of Paranoia
Re: Self-Driving Cars
« Reply #42 on: April 22, 2015, 03:32:08 PM »
Don't forget that a computers reaction time is significantly better than a human. I think the trade off of eliminating human error is well worth it. 
Murphy's Law: Anything that can go wrong will go wrong.
Morvis' Law: Anything that does go wrong is my fault.

Offline Jeremy's Sea

  • Kintsukuroi, baby.
  • Frequent Poster
  • ******
  • Posts: 3160
  • 667 - Neighbor of the beast.
    • jeremyad [sic]
Re: Self-Driving Cars
« Reply #43 on: April 22, 2015, 03:35:05 PM »
Don't forget that a computers reaction time is significantly better than a human. I think the trade off of eliminating human error is well worth it.
I agree too. I still encounter people though who resist this.
Knowledge is power. France is bacon.

Offline Morvis13

  • Big Ol' Goober
  • Planetary Skeptic
  • *
  • Posts: 21970
  • Natural Source of Paranoia
Re: Self-Driving Cars
« Reply #44 on: April 22, 2015, 03:48:58 PM »
Human error accounts for like 80% of crash statistics.
Plus the car can be programmed to reduce speed in areas where children play and even during hours when they would be out.
I think stop lights and speed limits would go extinct once the humans are removed from behind the wheel.
Murphy's Law: Anything that can go wrong will go wrong.
Morvis' Law: Anything that does go wrong is my fault.

 

personate-rain