Sinfest

[DeD CHiKn]

[picturesofsky]

And to power the evil killer robot dogs from hell, it's the world's first deliberately evil Artificial Intelligence:

http://www.sciam.com/article.cfm?id=defining-evil
_________________
Ironically, Halen's one of the few people here I wouldn't worry about terrifying my friends and family. In my head he ends every real life conversation stroking his chin and saying, "well yes, that sounds reasonable."

[Pontupo]

Robots are coming... so we should talk about it

[tinkeringIdiot]

So there are a lot of gallys out there?

[Pontupo]

I'm glad you noted that... the very same passage stuck out in my mind as well =) Maybe we _should_ replace him with a robot?

[tinkeringIdiot]

It's certainly an interesting debate. At the moment I fail to see the dark side, but then I'm an engineer and therefore have more faith in engineered systems than is probably healthy.

I'd say robots could be fine for patrol duty where they would be easy to train and have fairly repeatable operating conditions. Sending them in to do raids or other offensive maneuvers would pose more of an ethical issue to me. Though they can be built with a fairly large safety factor since they are disposable.

[Pontupo]

I would worry about about sending them into very complicated recognition scenarios where they would have to do something like identify a "civilian." From what I understand, our visual recognition software isn't anywhere near the level required for that kind of analysis.

The border guard scenario seems promising... but somehow I still worry a bit, given the current status of our AI abilities, about arming them with lethal weapons. I'd rather they had like stun guns or something. It'd suck ass to get pwned by a robot with an AK-47 just for trespassing.

[Zasabi]

They could, y'know, do a warning first or something. Assuming they're all linked, if one goes down, you'd probably know right away and could rectify the situation. Hell, you could make it so that if it ever has to give out a warning, it sends out a flag or something in the system, so you can check out the area with a camera or something, see what's going on for yourself.

*shrugs* Maybe.

[tinkeringIdiot]

So you wouldn't be comfortable without a human in the loop, Zasabi? What about less-lethal weaponry? Crazy knockout gas and such?

If you ran them in teams, you could have one give a warning and the other use lethal or less-lethal armament if the first one is destroyed. How about that?

[Kilgore]

You guys can have all the faith you want, but when one of your robot border guards shoots up a deaf kid there's going to be hell to pay.
_________________
"Whatever afflicts thee, their asses I shall kick"

-Slick

[tinkeringIdiot]

Naturally. But the question posed acknowledges that there will be mistakes. What I'm trying to decide is if I'm more comfortable with mistakes made by machines or mistakes made by humans. In my mind, a machine that makes a mistake can be more easily corrected than a person. No trial, no counseling, no blame...just reprogram/rebuild the bot and try again. That combined with the repeatability of machinated actions has me leaning toward embracing the technology.

[Zasabi]

The problem is that in the case of an actual attack, such gear is not as effective, especially if the enemy know. Non-lethal gases can be defeated with the proper equipment, and I'm sure there's some method of blocking off taser-like hits.

Your idea is a good one, and is kind of linked to what I had in mind. I was more going for just a slightly less damaging scenarios- say it's just some random guy who stumbles onto the base. In the case of the robot being destroyed or injured, then it's probably better for them to respond without the need for a human approver. Still, an alarm should go up at the least to keep humans in the loop.

I guess the perfect way it'd work to me is

1. Verbal Warning, flag system and offer visual feed to support staff.
If ignored and approved of a response, go to step 2 or 3. If attacked, go straight to 4.
2. use non-lethal weaponry, preferably something that has the least chance of accidentally killing someone.
3. Signal a human squad to assist
If approved or, as above, attacked, commence 4.
4. Begin use of lethal weaponry, signal the alert, possibly allow use of other sentries to aid the machine as to attempt to solve the problem before any more of our lives are lost. Our being the side utilizing them.

Something like that. Not too complex, methinks.

Edit- As for dealing with the Deaf, we can also have powerful spotlight-type displays along with some sort of bright red "Alert! Turn back now!" type thing upon spotting someone. If they're Deaf and Blind, then they're screwed. Sucks to be them, at that point, if the coordinator decides they're a threat.

[Mindslicer]

[WheelsOfConfusion]

And here I was debating whether to put a Mindstorms NXT kit, Robosapien V2 or a Bladestar on my holiday Wishlist.
_________________

[tinkeringIdiot]

1.) Robosapien, duh.
2.) That guy in the three piece suit is scary.