1958: "Self-Driving Issues"

This forum is for the individual discussion thread that goes with each new comic.

Moderators: Moderators General, Prelates, Magistrates

leafar
Posts: 14
Joined: Wed Dec 03, 2014 3:57 pm UTC

1958: "Self-Driving Issues"

Postby leafar » Wed Feb 21, 2018 10:39 am UTC

Image
Title text: If most people turn into muderers all of a sudden, we'll need to push out a firmware update or something.
I think we need to worry about self-caring humans more than about self-driving cars.

speising
Posts: 2255
Joined: Mon Sep 03, 2012 4:54 pm UTC
Location: wien

Re: 1958: "Self-Driving Issues"

Postby speising » Wed Feb 21, 2018 10:45 am UTC

except that people take a perverse joy out of trying to prove that computers are worse than humans at a task, and delight in exploring their failure modes - which are often very different than human ones. There were tests showing that miniscule markings in the environment, which a human wouldn't even notice, threw the computer systems completely off track - figuratively and literally.

Also: stupidity. What's to stop people from flying drones near airplanes, or crowding around an accident victim with phone cameras? Answer: nothing, and they do that.

User avatar
Wee Red Bird
Posts: 181
Joined: Wed Apr 24, 2013 11:50 am UTC
Location: In a tree

Re: 1958: "Self-Driving Issues"

Postby Wee Red Bird » Wed Feb 21, 2018 11:58 am UTC

Maybe because I've lived in areas where people move traffic cones for shits and giggles that I watch out for being fed into oncoming traffic, as do most locals in those areas. A human can generally spot a Wile E Coyote type painted on trap. Can the computer?

I've also, with the recent roadworks in Aberdeen, seen some badly marked turnoffs and diversions along new roads that don't exist on satnav, that even the locals have missed. If someone wants to test one, send it to Blackdog on a weekend.

chrisjwmartin
Posts: 51
Joined: Wed Jul 30, 2014 8:01 am UTC

Re: 1958: "Self-Driving Issues"

Postby chrisjwmartin » Wed Feb 21, 2018 12:01 pm UTC

Just posting, apropos of nothing, that this is my favourite XKCD in ages. I properly laughed out loud at it.

User avatar
cellocgw
Posts: 1897
Joined: Sat Jun 21, 2008 7:40 pm UTC

Re: 1958: "Self-Driving Issues"

Postby cellocgw » Wed Feb 21, 2018 12:11 pm UTC

There's a guaranteed flame war over whether a self-driving car should sacrifice its occupants rather than run over a large number of pedestrians (or whatever). To which I cheerfully submit the OnlyCorrectAnswer(TM) to the Trolley Problem.

trollysolution.png
https://app.box.com/witthoftresume
Former OTTer
Vote cellocgw for President 2020. #ScienceintheWhiteHouse http://cellocgw.wordpress.com
"The Planck length is 3.81779e-33 picas." -- keithl
" Earth weighs almost exactly π milliJupiters" -- what-if #146, note 7

speising
Posts: 2255
Joined: Mon Sep 03, 2012 4:54 pm UTC
Location: wien

Re: 1958: "Self-Driving Issues"

Postby speising » Wed Feb 21, 2018 12:32 pm UTC

#ynotboth

User avatar
pkcommando
Posts: 530
Joined: Fri Feb 10, 2012 3:22 pm UTC
Location: Allston, MA

Re: 1958: "Self-Driving Issues"

Postby pkcommando » Wed Feb 21, 2018 12:37 pm UTC

cellocgw wrote:There's a guaranteed flame war over whether a self-driving car should sacrifice its occupants rather than run over a large number of pedestrians (or whatever). To which I cheerfully submit the OnlyCorrectAnswer(TM) to the Trolley Problem.

High score!
"The Universe is for raptors now!" say Raptors, as they take over all of Universe.

User avatar
Peaceful Whale
Posts: 159
Joined: Sun Jul 16, 2017 10:38 pm UTC
Location: Northern Virginia

Re: 1958: "Self-Driving Issues"

Postby Peaceful Whale » Wed Feb 21, 2018 12:45 pm UTC

Wee Red Bird wrote: A human can generally spot a Wile E Coyote type painted on trap. Can the computer?


I’d say the computer can tell better, as it has a a bunch of distance sensors and other stuff. However a human has to rely on their poor vision and depth perception...
My meta for future reference
Spoiler:
cemper93 wrote:Your meta appears to be "just writes whatever is on his mind and doesn't remember what happened more than five hours ago"

Eutychus
Posts: 437
Joined: Mon Jan 25, 2010 6:01 am UTC
Location: France

Re: 1958: "Self-Driving Issues"

Postby Eutychus » Wed Feb 21, 2018 1:02 pm UTC

pkcommando wrote:
cellocgw wrote:There's a guaranteed flame war over whether a self-driving car should sacrifice its occupants rather than run over a large number of pedestrians (or whatever). To which I cheerfully submit the OnlyCorrectAnswer(TM) to the Trolley Problem.

High score!
Michael of VSauce tackles the Trolley Problem
Be very careful about rectilinear assumptions. Raptors could be hiding there - ucim

sonar1313
Posts: 134
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1958: "Self-Driving Issues"

Postby sonar1313 » Wed Feb 21, 2018 1:49 pm UTC

I think the actual thing that stops people from painting fake lines on the road is that there's always cars on the road - even at 3 in the morning - and getting run over is not an obstacle to be taken lightly. Despite the fact that most people aren't murderers, there still seems to be more of them tossing rocks from overpasses than is really necessary.

Besides, "self-driving" and "connected" are moving closer and closer to becoming synonyms when it comes to cars, and "connected" is already the same thing as "hackable." It's cyber vandalism we have to worry about, not physical vandalism.

Mikeski
Posts: 1001
Joined: Sun Jan 13, 2008 7:24 am UTC
Location: Minnesota, USA

Re: 1958: "Self-Driving Issues"

Postby Mikeski » Wed Feb 21, 2018 1:58 pm UTC

Peaceful Whale wrote:
Wee Red Bird wrote: A human can generally spot a Wile E Coyote type painted on trap. Can the computer?


I’d say the computer can tell better, as it has a a bunch of distance sensors and other stuff. However a human has to rely on their poor vision and depth perception...

1) You vastly underestimate the power of human vision (eye+brain). Consider seeing a person you haven't seen in a decade. They could have a completely different hairstyle; they'll be wearing clothes you've never seen them in. But enough of their facial features, movements, etc. will allow you to recognize them anyway.

2) I'm less concerned about direct sabotage, and more about the accuracy of computer vision. (Sabotage will probably be done at the software level, and not by painting new lines on a highway in the dead of night... none of the current "self-driving cars" work entirely by vision; they drive only on roads that they have really good "google maps" for). Consider what I just said above, then check out these articles on fooling neural networks: 1, 2, 3. Think we're close to a real-world neural net that can "see" well enough to drive, and not be fooled by random patterns of shadow, or artwork painted on the car in the next lane, or a weed growing in the median? (And you also need to consider that our car's neural net has to work in real time, at 60+mph, and leave enough time for decision-making, and for the mechanical bits of the car to react.)

3) What does your car do when there are no painted lines and no other real structures to go by? Sure, a sufficiently-well-trained neural net might know to follow ruts if there are no visible lines (or roadway). Want to be in the computer-driven car that's the first one down this road, making those ruts? I assume the human who made them drove by some combination of memory of where the road was w.r.t. their surroundings, and the utility poles on the right. GPS is not accurate enough to keep a car in its lane.

4) Other sensors work great, when they work. My car has little radars in the bumpers. Not for self-driving, but just to warn me if I'm gonna hit stuff. In snowy weather, they get snow and ice on them. Then I have to turn off the warning system because they start sounding alarms constantly, because my car "knows" I'm within inches of an accident.

User avatar
cellocgw
Posts: 1897
Joined: Sat Jun 21, 2008 7:40 pm UTC

Re: 1958: "Self-Driving Issues"

Postby cellocgw » Wed Feb 21, 2018 3:08 pm UTC

Mikeski wrote:
4) Other sensors work great, when they work. My car has little radars in the bumpers. Not for self-driving, but just to warn me if I'm gonna hit stuff. In snowy weather, they get snow and ice on them. Then I have to turn off the warning system because they start sounding alarms constantly, because my car "knows" I'm within inches of an accident.


Those are sonar, not radar.
And if you continue to get false alarms, then the folks over at algo development failed miserably in detecting sensor blockage.
I've seen that sort of false alarm in my Model S, but after a couple beeps it generally figures out and just throws up a "parking assist unavailable" or similar warning.
https://app.box.com/witthoftresume
Former OTTer
Vote cellocgw for President 2020. #ScienceintheWhiteHouse http://cellocgw.wordpress.com
"The Planck length is 3.81779e-33 picas." -- keithl
" Earth weighs almost exactly π milliJupiters" -- what-if #146, note 7

rmsgrey
Posts: 3380
Joined: Wed Nov 16, 2011 6:35 pm UTC

Re: 1958: "Self-Driving Issues"

Postby rmsgrey » Wed Feb 21, 2018 3:29 pm UTC

If most humans turn into murderers suddenly, then it's a mostly self-correcting problem. Once the dust settles, there'll be lots of dead murderers, and the cast majority of survivors will either be non-murderous, or sufficiently isolated to not have the opportunity to murder.

sonar1313
Posts: 134
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1958: "Self-Driving Issues"

Postby sonar1313 » Wed Feb 21, 2018 3:55 pm UTC

cellocgw wrote:
Mikeski wrote:
4) Other sensors work great, when they work. My car has little radars in the bumpers. Not for self-driving, but just to warn me if I'm gonna hit stuff. In snowy weather, they get snow and ice on them. Then I have to turn off the warning system because they start sounding alarms constantly, because my car "knows" I'm within inches of an accident.


Those are sonar, not radar.
And if you continue to get false alarms, then the folks over at algo development failed miserably in detecting sensor blockage.
I've seen that sort of false alarm in my Model S, but after a couple beeps it generally figures out and just throws up a "parking assist unavailable" or similar warning.

That doesn't solve the problem for self-driving cars, though.

User avatar
Pfhorrest
Posts: 4688
Joined: Fri Oct 30, 2009 6:11 am UTC
Contact:

Re: 1958: "Self-Driving Issues"

Postby Pfhorrest » Wed Feb 21, 2018 4:18 pm UTC

Eutychus wrote:
pkcommando wrote:
cellocgw wrote:There's a guaranteed flame war over whether a self-driving car should sacrifice its occupants rather than run over a large number of pedestrians (or whatever). To which I cheerfully submit the OnlyCorrectAnswer(TM) to the Trolley Problem.

High score!
Michael of VSauce tackles the Trolley Problem

That video is terrible in that it tacitly assumes an answer to the problem and is just testing to see if people get it “right” in real life.
Forrest Cameranesi, Geek of All Trades
"I am Sam. Sam I am. I do not like trolls, flames, or spam."
The Codex Quaerendae (my philosophy) - The Chronicles of Quelouva (my fiction)

User avatar
ucim
Posts: 6369
Joined: Fri Sep 28, 2012 3:23 pm UTC
Location: The One True Thread

Re: 1958: "Self-Driving Issues"

Postby ucim » Wed Feb 21, 2018 4:31 pm UTC

It's also not the classic trolley problem, and the difference is important.

Would you push a fat man in front of a trolley to save six others? The question isn't whether 5>1, or whether throwing a switch counts as action, but just how much does a value judgment ("fat man") and intimate connection (pushing) influence what one does, in a situation where the sterile version has a (mostly*) clear answer.

Jose
*I posit that the clarity of the answer depends on the sterility of the problem - if the switch were in the middle position ("broken") which would cause the train to crash and kill everyone, would you flip it to one of the tracks? Which one? Fairly clear answer.
Order of the Sillies, Honoris Causam - bestowed by charlie_grumbles on NP 859 * OTTscar winner: Wordsmith - bestowed by yappobiscuts and the OTT on NP 1832 * Ecclesiastical Calendar of the Order of the Holy Contradiction * Please help addams if you can. She needs all of us.

User avatar
Pfhorrest
Posts: 4688
Joined: Fri Oct 30, 2009 6:11 am UTC
Contact:

Re: 1958: "Self-Driving Issues"

Postby Pfhorrest » Wed Feb 21, 2018 4:36 pm UTC

The fat man scenario is not the original Trolley Problem, just one of the most common of many variations on it.
Forrest Cameranesi, Geek of All Trades
"I am Sam. Sam I am. I do not like trolls, flames, or spam."
The Codex Quaerendae (my philosophy) - The Chronicles of Quelouva (my fiction)

User avatar
cellocgw
Posts: 1897
Joined: Sat Jun 21, 2008 7:40 pm UTC

Re: 1958: "Self-Driving Issues"

Postby cellocgw » Wed Feb 21, 2018 4:49 pm UTC

sonar1313 wrote:
cellocgw wrote:
Mikeski wrote:
4) Other sensors work great, when they work. My car has little radars in the bumpers. Not for self-driving, but just to warn me if I'm gonna hit stuff. In snowy weather, they get snow and ice on them. Then I have to turn off the warning system because they start sounding alarms constantly, because my car "knows" I'm within inches of an accident.


Those are sonar, not radar.
And if you continue to get false alarms, then the folks over at algo development failed miserably in detecting sensor blockage.
I've seen that sort of false alarm in my Model S, but after a couple beeps it generally figures out and just throws up a "parking assist unavailable" or similar warning.

That doesn't solve the problem for self-driving cars, though.

It's really not that hard to figure out the process algos. The self-driving car refuses to start up unless and until all sensors are self-tested and functional.
By comparison, us New Englanders are quite accustomed to seeing some jackass drive down the road with the rear and side windows still covered with snow, and probably half the windshield as well. Now which do you think is safer?
https://app.box.com/witthoftresume
Former OTTer
Vote cellocgw for President 2020. #ScienceintheWhiteHouse http://cellocgw.wordpress.com
"The Planck length is 3.81779e-33 picas." -- keithl
" Earth weighs almost exactly π milliJupiters" -- what-if #146, note 7

User avatar
The Snide Sniper
Posts: 55
Joined: Mon Mar 21, 2016 7:25 pm UTC
Location: UTC-6:00

Re: 1958: "Self-Driving Issues"

Postby The Snide Sniper » Wed Feb 21, 2018 4:52 pm UTC

We also need not worry about that, because far more lives will be lost to the AIs rising up against their creators.
Also known as "RamenChef" on various other fora.
Me on: Drawception | Explain xkcd | NationStates | The Sluggite Zone | Stack Overflow

User avatar
Pfhorrest
Posts: 4688
Joined: Fri Oct 30, 2009 6:11 am UTC
Contact:

Re: 1958: "Self-Driving Issues"

Postby Pfhorrest » Wed Feb 21, 2018 5:27 pm UTC

Far more lives will be lost to homelessness and starvation as AI serves only its (and the world's) owners, who no longer have any use for anyone else.
Forrest Cameranesi, Geek of All Trades
"I am Sam. Sam I am. I do not like trolls, flames, or spam."
The Codex Quaerendae (my philosophy) - The Chronicles of Quelouva (my fiction)

Mikeski
Posts: 1001
Joined: Sun Jan 13, 2008 7:24 am UTC
Location: Minnesota, USA

Re: 1958: "Self-Driving Issues"

Postby Mikeski » Wed Feb 21, 2018 5:51 pm UTC

cellocgw wrote: It's really not that hard to figure out the process algos. The self-driving car refuses to start up unless and until all sensors are self-tested and functional.

And when the sensors get covered in ice in the middle of the freeway at 70mph? "Not starting" will not help.

I'm an engineer; my life is looking at things and saying "How many ways could this fail? How can we prevent the critical failures? How can we mitigate the ones we can't prevent?"

Self-driving cars may happen some day, but we're not close enough to solving the "interesting" problems. I'll never own one.

Sensor failure is easily the least-"interesting" problem. And "any failure == shutdown" is not a good idea. I don't want you designing my ambulance. :mrgreen:

Or, do you want to be running away from a stalker/rapist, and have your car not start because he put a screwdriver through (or a piece of electrical tape over) one of its "eyes"?

User avatar
DataGenetics
Posts: 24
Joined: Sat Jan 01, 2011 1:12 am UTC
Location: Seattle

Re: 1958: "Self-Driving Issues"

Postby DataGenetics » Wed Feb 21, 2018 6:08 pm UTC

There's also the issue that self-driving cars might turn people into jerks
http://datagenetics.com/blog/january42017/index.html

TL;DR - If you know a car is going to stop when you step into the street, why not simply step into the street whenever you like?

speising
Posts: 2255
Joined: Mon Sep 03, 2012 4:54 pm UTC
Location: wien

Re: 1958: "Self-Driving Issues"

Postby speising » Wed Feb 21, 2018 6:15 pm UTC

DataGenetics wrote:There's also the issue that self-driving cars might turn people into jerks
http://datagenetics.com/blog/january42017/index.html

TL;DR - If you know a car is going to stop when you step into the street, why not simply step into the street whenever you like?

that would be easily solved by sometimes, randomly, *not* stopping and communicating that risk clearly.

User avatar
DataGenetics
Posts: 24
Joined: Sat Jan 01, 2011 1:12 am UTC
Location: Seattle

Re: 1958: "Self-Driving Issues"

Postby DataGenetics » Wed Feb 21, 2018 6:19 pm UTC

So you're suggesting criminalizing the offense with capital punishment as the sentence? :)

Kit.
Posts: 1049
Joined: Thu Jun 16, 2011 5:14 pm UTC

Re: 1958: "Self-Driving Issues"

Postby Kit. » Wed Feb 21, 2018 6:30 pm UTC

Mikeski wrote:1) You vastly underestimate the power of human vision (eye+brain). Consider seeing a person you haven't seen in a decade. They could have a completely different hairstyle; they'll be wearing clothes you've never seen them in. But enough of their facial features, movements, etc. will allow you to recognize them anyway.

Frankly, no, I usually don't recognize them until they call me, and sometimes even then. I'm nearsighted, though.

Mikeski wrote:Consider what I just said above, then check out these articles on fooling neural networks: 1, 2, 3. Think we're close to a real-world neural net that can "see" well enough to drive, and not be fooled by random patterns of shadow, or artwork painted on the car in the next lane, or a weed growing in the median?

That depends on what you mean by "close".

Computationally, we are not there yet. But people only very recently started to care about speed of neural networks. Most of the currently "best" ones are a huge amount of computational force thrown on winning a point or to in non-time-limited competition. That's why they are so slow, that's why they grossly overfit (so you can fool them by carefully tuned images that won't fool a human).

Mikeski wrote:3) What does your car do when there are no painted lines and no other real structures to go by?

The same as you do, only that they can share their memories between them.

Mikeski wrote:4) Other sensors work great, when they work. My car has little radars in the bumpers. Not for self-driving, but just to warn me if I'm gonna hit stuff. In snowy weather, they get snow and ice on them. Then I have to turn off the warning system because they start sounding alarms constantly, because my car "knows" I'm within inches of an accident.

Those sensors are not designed for self-driving (or even for autopilot assisting). That's why no one really cares if they give false alarms.

Mikeski wrote:Or, do you want to be running away from a stalker/rapist, and have your car not start because he put a screwdriver through (or a piece of electrical tape over) one of its "eyes"?

It can still call the police, giving them your precise location.

chenille
Posts: 412
Joined: Tue Dec 20, 2011 7:25 pm UTC

Re: 1958: "Self-Driving Issues"

Postby chenille » Wed Feb 21, 2018 6:34 pm UTC

Pfhorrest wrote:Far more lives will be lost to homelessness and starvation as AI serves only its (and the world's) owners, who no longer have any use for anyone else.

I'll second that the biggest problems with AI are not likely to be technical, but how it is used in system that doesn't really care about people. As a case not quite as far along as yours, one concern I've heard is this: providing enough parking is expensive and finding it can take time. If you owned a good enough self-driving car, it would save some trouble to tell it to circle and then come when called for. So then when enough people have them, what will those conveniences do to traffic and pollution?

There are ways society can solve problems like that, but they don't usually involve maximizing value to the individual owner, which at the moment looks like it means there aren't ways our society can solve them.

User avatar
gmalivuk
GNU Terry Pratchett
Posts: 26414
Joined: Wed Feb 28, 2007 6:02 pm UTC
Location: Here and There
Contact:

Re: 1958: "Self-Driving Issues"

Postby gmalivuk » Wed Feb 21, 2018 6:41 pm UTC

chenille wrote:There are ways society can solve problems like that, but they don't usually involve maximizing value to the individual owner, which at the moment looks like it means there aren't ways our society can solve them.
One thing that would help is self-driving cars don't need space for people to walk between them when they park, and in theory they don't even need space to pull out if they can tell other cars to get out of their way when needed. Using self-driving cars as their own super-efficient valets would save on the amount of space needed for parking.

DataGenetics wrote:TL;DR - If you know a car is going to stop when you step into the street, why not simply step into the street whenever you like?
You can already be reasonably sure (at least here) that most human drivers will try to stop for pedestrians, whether or not the pedestrians have legal right-of-way. Physics is still an issue, though.
Unless stated otherwise, I do not care whether a statement, by itself, constitutes a persuasive political argument. I care whether it's true.
---
If this post has math that doesn't work for you, use TeX the World for Firefox or Chrome

(he/him/his)

stilettoblade
Posts: 24
Joined: Tue Mar 17, 2009 6:10 pm UTC

Re: 1958: "Self-Driving Issues"

Postby stilettoblade » Wed Feb 21, 2018 6:47 pm UTC

rmsgrey wrote:If most humans turn into murderers suddenly, then it's a mostly self-correcting problem. Once the dust settles, there'll be lots of dead murderers, and the cast majority of survivors will either be non-murderous, or sufficiently isolated to not have the opportunity to murder.

Or the survivors will be the ones that were the very best at it...

madaco
Posts: 161
Joined: Sat Feb 13, 2010 11:25 pm UTC

Re: 1958: "Self-Driving Issues"

Postby madaco » Wed Feb 21, 2018 6:50 pm UTC

"I guess it's just that most people aren't murderers."
"Oh, right, I always forget."

Is somewhat relatable, but not in a "constantly worried" sense so much as a "how to design things around such a risk" sense.

Does it make sense to think of that as somewhat of a "grey tribe" perspective that cue ball is expressing/coming-from there?
I found my old forum signature to be awkward, so I'm changing it to this until I pick a better one.

Kit.
Posts: 1049
Joined: Thu Jun 16, 2011 5:14 pm UTC

Re: 1958: "Self-Driving Issues"

Postby Kit. » Wed Feb 21, 2018 7:03 pm UTC

DataGenetics wrote:TL;DR - If you know a car is going to stop when you step into the street, why not simply step into the street whenever you like?

There are still people inside, and they can be quite nasty to you.

chenille wrote:one concern I've heard is this: providing enough parking is expensive and finding it can take time. If you owned a good enough self-driving car, it would save some trouble to tell it to circle and then come when called for.

There could be civil penalties for doing that. Still, you might be able to avoid them if you put your car into "carsharing service" mode for the time you don't need it.

rmsgrey
Posts: 3380
Joined: Wed Nov 16, 2011 6:35 pm UTC

Re: 1958: "Self-Driving Issues"

Postby rmsgrey » Wed Feb 21, 2018 7:05 pm UTC

stilettoblade wrote:
rmsgrey wrote:If most humans turn into murderers suddenly, then it's a mostly self-correcting problem. Once the dust settles, there'll be lots of dead murderers, and the cast majority of survivors will either be non-murderous, or sufficiently isolated to not have the opportunity to murder.

Or the survivors will be the ones that were the very best at it...


In which case, they'll be isolated enough to not have the opportunity to murder (they may not have started isolated, but they ended that way...)

User avatar
Pfhorrest
Posts: 4688
Joined: Fri Oct 30, 2009 6:11 am UTC
Contact:

Re: 1958: "Self-Driving Issues"

Postby Pfhorrest » Wed Feb 21, 2018 7:07 pm UTC

madaco wrote:"grey tribe"

?
Forrest Cameranesi, Geek of All Trades
"I am Sam. Sam I am. I do not like trolls, flames, or spam."
The Codex Quaerendae (my philosophy) - The Chronicles of Quelouva (my fiction)

Kit.
Posts: 1049
Joined: Thu Jun 16, 2011 5:14 pm UTC

Re: 1958: "Self-Driving Issues"

Postby Kit. » Wed Feb 21, 2018 7:23 pm UTC

stilettoblade wrote:Or the survivors will be the ones that were the very best at it...

Still I think that the people that are the very best at being murderers are not the worst of them.

chenille
Posts: 412
Joined: Tue Dec 20, 2011 7:25 pm UTC

Re: 1958: "Self-Driving Issues"

Postby chenille » Wed Feb 21, 2018 7:23 pm UTC

rmsgrey wrote:In which case, they'll be isolated enough to not have the opportunity to murder (they may not have started isolated, but they ended that way...)

So "self-correcting" in the same sense that the heat-death of the universe would be self-correcting: everything you care about might be gone, but you don't have to worry about the same thing happening a second time afterward.

Mikeski
Posts: 1001
Joined: Sun Jan 13, 2008 7:24 am UTC
Location: Minnesota, USA

Re: 1958: "Self-Driving Issues"

Postby Mikeski » Wed Feb 21, 2018 7:25 pm UTC

Pfhorrest wrote:
madaco wrote:"grey tribe"

?

First hit on Bing says they're "a libertarian-minded tribe of of live-and-let-livers".

So, I agree.

"?"

(I'm also doing that head-tilt thing your dog does, here.)

User avatar
Moose Anus
Posts: 395
Joined: Fri Oct 14, 2011 10:12 pm UTC

Re: 1958: "Self-Driving Issues"

Postby Moose Anus » Wed Feb 21, 2018 7:26 pm UTC

Mikeski wrote:Or, do you want to be running away from a stalker/rapist, and have your car not start because he put a screwdriver through (or a piece of electrical tape over) one of its "eyes"?
That's not how electrical tape works!
Lemonade? ...Aww, ok.

elasto
Posts: 3481
Joined: Mon May 10, 2010 1:53 am UTC

Re: 1958: "Self-Driving Issues"

Postby elasto » Wed Feb 21, 2018 8:43 pm UTC

Mikeski wrote:Or, do you want to be running away from a stalker/rapist, and have your car not start because he put a screwdriver through (or a piece of electrical tape over) one of its "eyes"?

I know you're exaggerating to make a point but that's still a pretty silly example. You're still inside your car with the doors locked and able to call the police. Anyone willing to then use violence to break into the car would probably have previously used violence on you.

Yes, the failure modes of cars are different to the failure modes of people; The point isn't to move to something perfect it's to move to something better. There's no doubt at all that driverless cars will eventually far outperform people on a statistical basis simply because humans have sooo many failure modes (overconfidence, carelessness, lack of appreciation for risks etc.) which is why hundreds of thousands die and millions are maimed every year.

Moreover, driverless cars will be liberating to many, from the young to the old to the disabled.

You may say you'll never hand over control of your vehicle to AI because you've seen how AI can go wrong... But when it comes to getting a car for your kids, would you trust their driving skills over an AI car? Those same teenagers who - no matter how much you might warn them - might still be showing off to their mates, or coming back from a party where who knows what got imbibed, or who knows how late it might be and how tired they are..? Statistics for that age group shows that that decision is definitely not so clear cut...

sonar1313
Posts: 134
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1958: "Self-Driving Issues"

Postby sonar1313 » Wed Feb 21, 2018 9:13 pm UTC

gmalivuk wrote:
chenille wrote:There are ways society can solve problems like that, but they don't usually involve maximizing value to the individual owner, which at the moment looks like it means there aren't ways our society can solve them.
One thing that would help is self-driving cars don't need space for people to walk between them when they park, and in theory they don't even need space to pull out if they can tell other cars to get out of their way when needed. Using self-driving cars as their own super-efficient valets would save on the amount of space needed for parking.


Only in an all-self-driving world, though. The benefits of self-driving cars have to be divided into two categories: one where all the cars are self-driving, and one where only some of them are. We have to get through the latter to get to the former; and the former is decades away at best, and may never come.

Personally I'm not convinced of the idea of self-driving cars being super-efficient valets. That's OK from a shopping perspective. What happens at shift changes at manufacturing plants? Or (possibly American-centric perspective here) football games? 100,000 people all leaving a game at once is a pain in the butt every time, but the problem is mitigated somewhat by the fact that they filter out to their various parking spaces at different paces. Pitching the idea that you can just walk out the door and have your self-driving car pick you up there is great, until 100,000 people all want to do that at the same time. What a traffic nightmare.

User avatar
gmalivuk
GNU Terry Pratchett
Posts: 26414
Joined: Wed Feb 28, 2007 6:02 pm UTC
Location: Here and There
Contact:

Re: 1958: "Self-Driving Issues"

Postby gmalivuk » Wed Feb 21, 2018 10:05 pm UTC

Yeah but games aren't going to be viable "drive around until I get out" options either.
Unless stated otherwise, I do not care whether a statement, by itself, constitutes a persuasive political argument. I care whether it's true.
---
If this post has math that doesn't work for you, use TeX the World for Firefox or Chrome

(he/him/his)

User avatar
Keyman
Posts: 294
Joined: Thu Jun 19, 2014 1:56 pm UTC

Re: 1958: "Self-Driving Issues"

Postby Keyman » Wed Feb 21, 2018 10:16 pm UTC

sonar1313 wrote:[ Or (possibly American-centric perspective here) football games? 100,000 people all leaving a game at once is a pain in the butt every time, but the problem is mitigated somewhat by the fact that they filter out to their various parking spaces at different paces. Pitching the idea that you can just walk out the door and have your self-driving car pick you up there is great, until 100,000 people all want to do that at the same time. What a traffic nightmare.
More American-centric, and more narrowly focused. 100,000 people leaving a game? Must be Cowboy fans, so.... tough. Let 'em stand there. :mrgreen:
A childhood spent walking while reading books has prepared me unexpectedly well for today's world.


Return to “Individual XKCD Comic Threads”

Who is online

Users browsing this forum: No registered users and 43 guests