Print

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

[Sergarr]

Design an AI to come up with strategies to keep the other AIs in line.

that's a bad idea

[Sergarr]

So you build an AI to come up with strategies to keep THAT AI in line!

It's elephants all the way down!

[Bauglir]

Do you think we could hire elephants for this? They should be able to remember all of the strategies we've applied, so we don't have to trust the AIs to do it.

[Descan]

Well the honest answer that might actually WORK

is to program the AI to not want to program itself to be all evil and shit.

Like if you could re-program your brain to make you want to be a cannibal, WOULD YOU?

probably not?

So why would an AI programmed to be all lovey and shit towards humans want to change that to be all "must eat hoomans for spare parts!" eh?

[Levi]

Heh, could be hard to program.  I remember in the X-Men cartoon they had programmed the Sentinels to protect humans from the mutants.  Then they turned on the humans, and when the inventor asked why they said "Mutants are human, therefore sentinels must protect humanity from itself". 

Then Cyclops blasted it with his eye lasers or something.  I don't remember the details.

[Sergarr]

Or just implement the laziness attribute. That and some overestimation of it's own abilities. Should go nicely to remove any possibility of it becoming "runaway" or "maniac"; after all, maniacs aren't lazy!

[MonkeyHead]

Is it that much of a stretch to create AI's that rely on us in some way? That would be a good incentive not to do us in. Morally grey, yes, but not as bad as blatantly lying to them, or "killing" to control.

[Levi]

Is it that much of a stretch to create AI's that rely on us in some way? That would be a good incentive not to do us in. Morraly grey, yes, but not as bad as blatantly lying to them, or "killing" to control.

Base them on a puppy dog's brain.  Sure, we'll have to play fetch and pet the computer a lot, but it can solve all of humanities problems while taking naps.   

[Descan]

Well, making them want to not want to kill us wouldn't be lying to us. It'd just be a very controlled way of "raising" them. Rather than relying on natural formation as in human children (if that would even be possible, it would be... undesired. Even if your kid turns into an asshole, he (probably) won't have access to literally planet-ending technology.)

[Sergarr]

Laziness is quite literally the best way to prevent random flukes resulting in catastrophic consequences.

Also, every single living being on Earth is inherently lazy, so it's only natural to give our AI that gift!

[Illogical_Blox]

Programming with emotions tends to go a long way towards removing the whole notion of maniac killing AIs.

Weeeeeeeeell, many of the fictional AIs went rogue because they had emotions, and they saw what evil creatures we can be at times. Every person is evil, yet people are good. Figure that out and you are clever than me. 

[Frumple]

... pretty sure there's nothing out there that's actually inherently lazy. Humans certainly aren't, and neither are any animal species I can recall. Occasionally optimized for long downtimes or large amounts of sleep, but that's a very different sort of thing.

Regardless, even something substantially lazier than a lazy human can still end the world if that results in exponentially more uptime regardless. Wasting 75% of your time is fairly irrelevant if you're working 100,000% faster or whatever. Hell, wasting 99+% of one's time is still irrelevant if the thing in question is able to self-upgrade -- eventually that .0001% potential will still exceed mankind's capability or build something less wasteful or whatnot.

Laziness wouldn't prevent an AI end of times, it would just slow it down. Marginally.

[Sergarr]

... pretty sure there's nothing out there that's actually inherently lazy. Humans certainly aren't, and neither are any animal species I can recall. Occasionally optimized for long downtimes or large amounts of sleep, but that's a very different sort of thing.

Regardless, even something substantially lazier than a lazy human can still end the world if that results in exponentially more uptime regardless. Wasting 75% of your time is fairly irrelevant if you're working 100,000% faster or whatever. Hell, wasting 99+% of one's time is still irrelevant if the thing in question is able to self-upgrade -- eventually that .0001% potential will still exceed mankind's capability or build something less wasteful or whatnot.

Laziness wouldn't prevent an AI end of times, it would just slow it down. Marginally.

Laziness would prevent it from self-improving at all. Just like it does for humans. From AI's position, if it works good enough, then why change it? Too much effort.

Judging by your comment of "humans aren't lazy", you don't really understand just how deep the roots of laziness go

[Frumple]

Judging by your comment of "humans aren't lazy", you don't really understand just how deep the roots of laziness go

... understand quite well, and the answer is "not very". Humans as a species are considerably inclined towards action, particularly creative action, and it takes fairly substantial environmental pressures and/or mental illness to actually make us inclined towards substantial inaction. Laziness is very much not inherent to humankind.

"Too much effort" isn't a sign of laziness, it's a sign of depression, generally. A lazy person can and will still improve, just at a slower pace. People like to do things, even if they're lazy.

If anything, laziness just breeds efficiency, getting more done with less effort.

[alway]

Meanwhile, Elon Musk continues being Iron Man.
"Btw, we are actually working on a charger that automatically moves out from the wall & connects like a solid metal snake. For realz."
https://twitter.com/elonmusk/status/550297212769402881