Is a Skynet scenario plausible?

[Howie Roary]

https://www.google.com/#hl=en&safe=o...w=1024&bih=653

[Kang]

This thread gives me very little hope that mankind can repel an AI invasion

[isladar]

I think you're absolutely right:

Overloard is on Facebook.

[Abandon]

Education at its finest

[Rhyis.]

I'm sorry Dave, I'm afraid I can't do that.

[Mojo]

The answer is yes, unless you're religious, in which case you're wrong.

[noblemountain]

for the second time in 10 days, WHAT THE FUCK IS A LOARD

I started a trend! At least I'm not the only person who can't spell for shit.

[Abandon]

The answer is yes, unless you're religious, in which case you're wrong.

stop it

[hey]

Don't most computers come with off switches? What is really the worst case scenario here?

Unless someone is making some machine that can sustain itself long term, kill people, has no way of shutting it down, and is programmed to do anything it has to to continue operation, there's really nothingt hat can possibly happen. And in that case...why are we making something like that? AI can conceivably make a machine do unexpected things, but if that's a problem, there's a simple solution: shut it down.

[Kerberoz]

just leaving this here...

http://www.youtube.com/user/BostonDynamics

delicious uncanny valley

[Cantih]

The quickest way to get a hostile intelligence is by treating it like one from the get go. Create one in a Sword of Damocles like environment, or try to lock down a spontaneous one, no human would appreciate that, so why would a non-human?

Also, an easy "take over everything" worry is simply solved by NOT making every single damned thing interconnected or sticking proto-AI everywhere when something simpler will do. "When all you have is a hammer..."

Superior/different intelligence is a trickier beast, I can't speak as well on that, but if we can get to a place where lesser intelligence is treated as ethically as equal intelligence, we'll probably be ready to handle it.

[Deadgye]

http://www.google.com/#hl=en&tbo=d&s...w=1280&bih=841

One who is willing to go the distance and have intimate relations with an abnormally large, homely woman in order to further his friend(s) chances with her better looking female accomplice.

This sounds like a perfectly acceptable definition.

[Seanr]

Don't most computers come with off switches? What is really the worst case scenario here?

Unless someone is making some machine that can sustain itself long term, kill people, has no way of shutting it down, and is programmed to do anything it has to to continue operation, there's really nothingt hat can possibly happen. And in that case...why are we making something like that? AI can conceivably make a machine do unexpected things, but if that's a problem, there's a simple solution: shut it down.

Computers can build themselves. You don't think it can't build one without a kill switch? Sustain itself long term? Computer can figure out what it would need by itself.... Next you'll say there's no life other than what's on earth.

[hey]

Computers can build themselves. You don't think it can't build one without a kill switch? Sustain itself long term? Computer can figure out what it would need by itself.... Next you'll say there's no life other than what's on earth.

So in addition to all the things i listed above, it also needs the ability to self replicate, and to learn new methods of this (why not just hard code this if it's needed...?). AI isn't magic, and they can't learn the same way people can. If you don't want it to learn something, you just don't let it.

[BaneTheBrawler]

you're not thinking big enough, Hey. Desktop PCs aren't what we need to worry about. If we're gonna worry about this problem, you need to look at automated systems, like traffic control, security systems, ventilation, all sorts of stuff that's already automated. The computers behind the stuff thats already automated are only going to get smarter (and some of them already know how to learn), and we're going to find all kinds of new things for them to control. Think Smart House on a larger scale.

So if your office building's computer, which controls fire/security/ventilation/etc systems, decides it doesn't want anyone going into it's control room... no one's going to go into its control room, unless you build in overrides... but then you leave the computer vulnerable to terrorists and other malcontents. Just like any other security system, the more escape routes/back doors/etc you have the more vulnerable it is.

[Seanr]

So in addition to all the things i listed above, it also needs the ability to self replicate, and to learn new methods of this (why not just hard code this if it's needed...?). AI isn't magic, and they can't learn the same way people can. If you don't want it to learn something, you just don't let it.

As technology gets exponentially better over time, AI will do the same. "just don't let it" what happens when it just figures out how to get around it?

[hey]

"just don't let it" what happens when it just figures out how to get around it?

AI isn't magic, it can only do what it is programmed to do.

[Ishido]

AI isn't magic, it can only do what it is programmed to do.

I'm not sure if you understand the concept of AI.

[BaneTheBrawler]

this argument is the equivalent of "just don't let someone assault you"

[myreality]

I can see this happening in the same way as the movie Eagle Eye. Except the AI would not be hindered for plot reasons.