Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Zrk2]

I'm a bit late for the spamfest, but:

IT'S SKYNET!!!

Continue your regularly scheduled discussion.

[AntiAntiMatter]

I'm a bit late for the spamfest, but:

IT'S SKYNET!!!

Continue your regularly scheduled discussion.

Incorrect. It isn't intelligent, it's just a network, not an AI. Rather, it will be the basis for a robot hive-mind.

[Zrk2]

I'm a bit late for the spamfest, but:

IT'S SKYNET!!!

Continue your regularly scheduled discussion.

Incorrect. It isn't intelligent, it's just a network, not an AI. Rather, it will be the basis for a robot hive-mind.

Close enough...

[freeformschooler]

I thought this was very cool until I realized vandalism is very real.

"you guys you have to put a lit match in the car to completely refill it after filling it up with gas"

[AntiAntiMatter]

I thought this was very cool until I realized vandalism is very real.

"you guys you have to put a lit match in the car to completely refill it after filling it up with gas"

Oh course, hopefully they won't just follow instructions blindly...

[Ochita]

maybe a hardcode
No fire
No death
No harming people

Stuff like that?

[Starver]

I'd just get a robot and then train it so it associates the wrong actions with orders so that all around the world guys get punched in the balls when they ask for a coke.

Good idea.  Of course, you need to get started on that project right away, while it's just you and your robot.  That way when the robotnet goes live it'll already have a well-reinforced database on how to accomplish the action, however much the victimhuman tries to protect himself, jump around, shout "No, no, no, you stupid robot!  I wish I'd never taught you to punch me in the balls!" and try to find its off-switch. 

(Orchita, at al: Three Laws, Asimov.  Already been thought of.  But a little bit handwavium on how it might actually be accomplished, still.)

[Zangi]

Well, if robots could calculate the consequences of its actions.  (Also knowing possible results of actions would be helpful...)

Otherwise... there is gonna be a lot of 'oops, I didn't know that could happen' murder.

[Starver]

"...and my client submits a faithfully forensic copy of its brain data to support his... er, its.... claim."

/waits for claims of "virtualisation" to arise from the Prosecution bench.

[Virex]

The really foreboding part is that there's never been a viable or realistic way to force robots into not harming humans. It requires
a) An overseer layer (either human or AI) that determines whether an action is harmful to humans (which itself requires another oversight layer to make sure it isn't making faulty or outwright falsified decisions, ad infinitum),
b) A database of every possible action that could result in the death of a human (prohibitively extensive, and prone to failure due to extreme circumstances),
or c) A highly brainwashed human-like mind (prone to a nasty thing called learning)

So no, there's really no preventing a robot from learning that drain cleaner and bleach make awesome milkshakes. The solution really is to not drink bubbly milkshakes made by robots. I can see this working as a local network, highly controlled by professionals to help in hospitals, jails, and other institutions, but not as an open database.

There is also no realistic way to prevent humans from harming other humans and that has never stopped us from creating more humans. Hell, at least a robot is unlikely to give you a car battery (assuming it's strong enough to lift a car) out of spite or jealousy.

[Akura]

Just thinking, could we teach robots to rap?

I think we already have. Not quite, but awesome nonetheless.