 |
The Gaming Den
Welcome to the Gaming Den.
|
| View previous topic :: View next topic |
| Author |
Message |
hyzmarca
Duke
Joined: 14 Mar 2011
Posts: 1012
|
 Posted: Fri May 18, 2012 12:58 am Post subject: |
  |
|
| fbmf wrote: |
As was mentioned earlier, military warbots do not have to be AI.
Game On,
fbmf |
No, they really do, for one important reason. Accountability.
We can make dumb killbots now. Modern predator drones can identify targets all by themselves. The only reason they can't shoot their missiles without a guy at a control panel miles away pushing a button is the need for accountability.
When a drone murders civilians you can't put it in prison. You can't hold a trial for it. You can parade it in front of the cameras and ensure the public that justice has been served. It's just a dumb machine. If it kills civilians, then that's just an equipment mishap and the blame falls squarely the officers and politicians who approved the use of the things.
Sending out dumb killbots is ethically similar to laying a minefield that moves. Laying a minefield in clearly marked areas that you can easily demine is already ethically suspect. Laying mines that seek targets out and kill them in their bedrooms is just beyond the pale.
Thus every killbot will have to be directly supervised by a human, and that's a problem. Radio communication can be jammed, and humans can be killed, and the number of robot units in service will eventually exceed the number of qualified operators. That's actually a problem now. Theire aren't enough soldiers qualified to control drones today. That problem will only be exacerbated as more and more drones are introduced.
A smart AI gives you the advantage of being a morally culpable actor. You get give it orders, send it out, and if it murders civilians or otherwise breaks the law, you can place the blame squarely on it. |
|
| Back to top |
|
 |
Kaelik
King
Joined: 07 Mar 2008
Posts: 6712
|
| Posted: Fri May 18, 2012 1:44 am Post subject: |
|
|
| hyzmarca wrote: | No, they really do, for one important reason. Accountability.
We can make dumb killbots now. Modern predator drones can identify targets all by themselves. The only reason they can't shoot their missiles without a guy at a control panel miles away pushing a button is the need for accountability.
When a drone murders civilians you can't put it in prison. You can't hold a trial for it. You can parade it in front of the cameras and ensure the public that justice has been served. It's just a dumb machine. If it kills civilians, then that's just an equipment mishap and the blame falls squarely the officers and politicians who approved the use of the things.
Sending out dumb killbots is ethically similar to laying a minefield that moves. Laying a minefield in clearly marked areas that you can easily demine is already ethically suspect. Laying mines that seek targets out and kill them in their bedrooms is just beyond the pale.
Thus every killbot will have to be directly supervised by a human, and that's a problem. Radio communication can be jammed, and humans can be killed, and the number of robot units in service will eventually exceed the number of qualified operators. That's actually a problem now. Theire aren't enough soldiers qualified to control drones today. That problem will only be exacerbated as more and more drones are introduced.
A smart AI gives you the advantage of being a morally culpable actor. You get give it orders, send it out, and if it murders civilians or otherwise breaks the law, you can place the blame squarely on it. |
See, not everyone is as dumb as you are, so we don't care about whether the machine is a "morally culpable actor" when it kills people.
Of course, morally culpable actor is already and always has been a shell game where people are morally culpable if we say they are, and aren't if we say they aren't, and there is no actual real distinction. So we can just say the kill bots that aren't AIs are morally culpable and it will be just as true as it is of humans. Then idiots like you can be happy when we deactivate them because that will suddenly make our deactivation more capable of preventing future unwanted deaths because we called them by a different name.
_________________
| DSMatticus wrote: | | Kaelik gonna kaelik. Whatcha gonna do? |
Click here to see the hidden message (It might contain spoilers) | FrankTrollman wrote: | | Really, the only thing the "my character can beat up your character" challenges ever do by presenting a clear and unambiguous beat down is to have the loser drop of the thread and pretend the challenge never happened. |
| Lord Mistborn wrote: | | The way people talk it seems like any of us could at least do better than Mearls, why does Castles and Cocks not already exist? |
| fectin wrote: | | I dare you to name five Denners you'd be willing to work with for that long. |
| Chamomile wrote: | | Hard mode: Those Denners must also be willing to work with you. |
Last edited by Kaelik on Fri May 18, 2012 1:46 am; edited 1 time in total |
|
| Back to top |
|
|
Winnah
Duke
Joined: 15 Feb 2011
Posts: 1016
Location: Oz
|
| Posted: Sat May 19, 2012 4:01 am Post subject: |
|
|
Interesting article about the use of drones in 'covert' warfare.
Unsure of the veracity of the article. |
|
| Back to top |
|
|
sabs
Duke
Joined: 29 Dec 2010
Posts: 2071
Location: Delaware
|
| Posted: Sat May 19, 2012 4:19 am Post subject: |
|
|
| also, how do you kill an AI that can literally store itself into hardware. Unless somehow the AIness is limited to a positronic brain or something, and non-transferable. Which.. seems weird. |
|
| Back to top |
|
|
Murtak
Duke
Joined: 07 Mar 2008
Posts: 1579
|
| Posted: Sat May 19, 2012 10:41 am Post subject: |
|
|
| RobbyPants wrote: | | Kaelik is right. You don't want it making decisions on whether or not to kill people. You build it right in that it can't fucking kill people. |
I'm not even sure that is possible. How do you define "killing people"? What are people? What is killing? I am not even talking about the difficulty of coding such a rule (though that would a monumental task in itself), but about the difficulty of formally defining what this act of killing actually is. Is shooting a gun at someone "killing"? Depressing a trigger does not equate to killing. Is about likelihood of killing someone? Say, any action with a likelihood of more than 20% of killing someone is prohibited? Welcome to infinite-analysis-land, where AIs sit in a windowless room, afraid to move. And even with infinite computing power, what about maiming someone? Is that ok? What about helping to build a McDonalds, which is sure to hasten the deaths of multiple people? Is that ok? Is it allowed to shoot someone to keep them from killing someone else? Humanity took centuries and we are still unclear on when it is ok to kill someone or on what actually constitutes "killing".
We can't even define what it is we don't want AIs to do. It is also doubtful we could program those rules if we made up our minds. And even if we could, it is very possible that following these rules would doom an AI to paralysis. And even if it didn't it is quitely likely that the effort of constantly evaluating the rules would use up every bit of computing power the AI actually has.
I can imagine AIs being possible. I can imagine them being capable of learning, of being trained or of being naturally cooperative. But I seriously doubt that something akin to the laws of robotics is possible.
_________________
Murtak |
|
| Back to top |
|
|
Chamomile
Duke
Joined: 03 May 2011
Posts: 1992
|
| Posted: Sat May 19, 2012 12:07 pm Post subject: |
|
|
| Humans can conceive of the concept of "killing" and barring a few edge cases, it's a consistent and coherent model of a single concept. I don't know how exactly, but doing the same for AI should be hypothetically possible at some point in the future (and the premise of the discussion is the future when AIs are a thing). |
|
| Back to top |
|
|
RobbyPants
Prince
Joined: 06 Aug 2008
Posts: 3157
|
| Posted: Sat May 19, 2012 12:51 pm Post subject: |
|
|
| Murtak wrote: | | RobbyPants wrote: | | Kaelik is right. You don't want it making decisions on whether or not to kill people. You build it right in that it can't fucking kill people. |
I'm not even sure that is possible. How do you define "killing people"? What are people? What is killing? |
We acknowledged that earlier. My point was, if we knew how to tell it how not to kill, we wouldn't have the robot "learn" not to kill; we'd build it in as a non-over-ridable routine.
_________________
Click here to see the hidden message (It might contain spoilers) First time:
| tzor wrote: | OK FUCK THIS ALL ---- BYE BYE TGD
I didn't come here to get ulcers ---- You idiots can rot in your own progressive hell forever. |
Hopefully last time:
| tzor wrote: | Don't worry, I won't bother posting anything "new" from now on.
Enjoy the hell you have created for yourselves. I'm out of here. |
Epic win by Fectin: Ignoratio!
| DSMatticus wrote: | | People don't like realism in their fantasy games (unless it's the fighter, lol fuck that guy). |
|
|
| Back to top |
|
|
Parthenon
Knight-Baron
Joined: 24 Jan 2009
Posts: 733
|
| Posted: Sat May 19, 2012 2:21 pm Post subject: |
|
|
| Chamomile wrote: | | Humans can conceive of the concept of "killing" and barring a few edge cases, it's a consistent and coherent model of a single concept. I don't know how exactly, but doing the same for AI should be hypothetically possible at some point in the future (and the premise of the discussion is the future when AIs are a thing). |
What on earth are you on about?
Is assisted suicide killing someone? Is an execution the same as killing someone? Is knowingly and deliberately not stopping a death the same as killing? Is it murder that you define as killing, or is manslaughter killing? Is it knowingly causing a death? If so does it matter how likely the action is to cause the death?
I'm not looking for answers to these questions, just pointing out that people currently have different opinions on these questions.
Saying that "killing" is a self-evident term is retarded because people don't agree what it is, and when they can agree on a term they still can argue whether one instance is killing or not.
Thats the same bullshit as "natural laws" and "these rights are held to be self-evident" or whatever the phrasing is. |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|
FAQ
Search
Memberlist
Usergroups
Register
Profile
Log in to check your private messages
Log in
