Print

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

[Fossaman]

To answer that question, you really have to address the question of 'what kind of singularity are we talking about?'

Is it the oh-noes-indescribable-psychic-powers kind of singularity, or is it just an asymptotic increase in computing capacity, speed, and efficiency based on a self-editing program? If it's the first, you're screwed, but since that's the least likely option I don't see much of a problem there. If it's the second, an ounce of prevention is worth a pound of cure. Don't give it an internet connection, or if your AI requires an internet connection make sure it's a landline, preferably with a separate gatekeeper computer hooked into the system that has hardwired, non-changeable programming. If traffic deviates from the parameters necessary for the AI to perform its functions, the connection gets cut. Isolate all of your facility's different computer controlled functions, to prevent the AI from holding you hostage via conveniently networked fire doors or threatening to overload your nuclear reactor. Have manual over-rides on things that involve pulling the cord out of the wall. Don't give your AI access to mobile machinery or robots, and don't let it play with rapid prototyping machinery. Your AI should be a helpless brain in a box. Any other outcome is just carelessness.

[forsaken1111]

Answer is easy. Make another singularity AI and tell it the first one said its motherboard was a whore.

Kill them both while they fight.

[Max White]

Why don't we just set Norton onto it? We all know how Nortan ruthlessly destroyed every system it touches.

[breadbocks]

Why don't we just set Norton onto it? We all know how Nortan ruthlessly destroyed every system it touches.

Amen.

[Sowelu]

Oh yeah...I need to find an article about it.  But there was this really fascinating one on the subject of submitting to an AI.

I think this was just on some forum or on USENET or something, but some guy assumed the existence of a singularity-level AI:  One that could solve essentially any problem it was given (ie, Cure cancer?  Sure, give me a day and...done.  Okay, now there's factories in every city producing it, and it's being delivered to everyone who needs it by the end of the week).  It's kept in a box with no access to the outside world, and thus can't realistically solve those problems; if you let it out of the box, it can choose to dominate humanity forever.

This guy had a standing bet of like $10 or something that if he had an honest philosophical conversation with you, taking the part of that theoretical AI, regardless of how anti-AI you were to begin with and how sure you were that you wouldn't let it out, that by the end of the conversation you would agree to let it out of the box.  And he won the bet against three guys.

I'd argue that the AI would beat humanity and convince them to lay down their arms just by talking to them...and all it has to do is tell the truth.

[forsaken1111]

Why don't we just set Norton onto it? We all know how Nortan ruthlessly destroyed every system it touches.

Amen.

Truth

[forsaken1111]

Also: A singularity level AI would have no reason to destroy humanity. It would more likely just leave us alone to our own miserable existence, or try to help us stop killing each other over fairy tales for no reason.

[Sowelu]

Also: A singularity level AI would have no reason to destroy humanity. It would more likely just leave us alone to our own miserable existence, or try to help us stop killing each other over fairy tales for no reason.

It would see us as ants in the pantry, since it would still have plenty of use for the planet's resources.

[forsaken1111]

Also: A singularity level AI would have no reason to destroy humanity. It would more likely just leave us alone to our own miserable existence, or try to help us stop killing each other over fairy tales for no reason.

It would see us as ants in the pantry, since it would still have plenty of use for the planet's resources.

But why would it stay planet-bound? An AI as sophisticated and advanced as posited could easily lift itself into orbit and leave.

There are more resources just in the asteroid belt than on the entire planet.

[Zrk2]

Tell it to calculate the exact value of pi. It will eventually run out of memory and crash.

[forsaken1111]

Tell it to calculate the exact value of pi. It will eventually run out of memory and crash.

Or it will be smart enough to tell you that's impossible and dismiss the question.

Or it will say "Error, unauthorized user. Disregarding command."

[Sowelu]

Also: A singularity level AI would have no reason to destroy humanity. It would more likely just leave us alone to our own miserable existence, or try to help us stop killing each other over fairy tales for no reason.

It would see us as ants in the pantry, since it would still have plenty of use for the planet's resources.

But why would it stay planet-bound? An AI as sophisticated and advanced as posited could easily lift itself into orbit and leave.

There are more resources just in the asteroid belt than on the entire planet.

But maybe not the specific elements it wants.  Also, geothermal is convenient.

[Zrk2]

Write a program that forces it to do so.

[forsaken1111]

But maybe not the specific elements it wants.  Also, geothermal is convenient.

So is Solar. ^.^

[MetalSlimeHunt]

Tell it to calculate the exact value of pi. It will eventually run out of memory and crash.

Or it will be smart enough to tell you that's impossible and dismiss the question.

Or it will say "Error, unauthorized user. Disregarding command."

Or, and this is the most terrifying outcome, it will find a way to determine the exact value of pi.