Print

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

[Orange Wizard]

* Orange Wizard pulls all the fuses out of the fuse box.

What's this about world-eating AI?

[ggamer]

But one of the reasons why we won't actually have a rogue AI gaining sentience and laying waste to the human race is because nobody will actually want to make one like that.
I think the most we'll do is have an extremely advanced chatterbot - nothing more.

In a world with billions of people, millions of which are psychopaths, and a good number that are omnicidal ^{citation needed}, there's always gonna be someone. And these sorts of people are also more likely to to claw their way into positions of power. It only takes one guy who rolls a critical success, and pushing society as a whole down that path will only make it easier for that to happen.

If the AI is intelligent, it will eventually start doing unexpected things. That's the entire point of developing a true AI, that it can think for itself. Even if not explicitly told to do something, it has the capacity to learn.

[Graknorke]

Why would an AI want to kill everyone anyway? An AI with even the smallest insight into the world would know that's an awful idea.

[SirQuiamus]

I sincerely hope that humanity will eventually cripple itself into a state of sustainable development and carbon neutrality. I just don't consider it very likely.

I hope that one day, after the technology has been developed to the point that we can travel among the stars in big colony ships, the Earth is left behind stripped bare of resources and a workable atmosphere. Then there will be two camps: one will say "look at the evidence of our mistakes, we should ground ourselves and go back to Stone Age hunter-gatherer societies," whereas the other will say "who the hell cares, there's plenty of planets for us to fuck in the galaxy." Then we realize that, somewhere along the line, we were so focused on the long view that we forgot to find a middle ground, and we realize that the only people left are extremists, who will never agree with each other and will most likely drive the rest of human society into the ground with conflict.

tl;dr holy shit there's a middle ground here between the salt-the-earth and murder-society-to-hobble-growth parties, and it's not all that hard to find

But can the middle ground involve computers, cars, and T-bone steaks? One party says it can, another party says it can't, and in the end, both parties end up continuing the debate on facebook while driving to a steak joint (this applies to countries as well as individuals). The problem is that everyone is currently at the extreme, "fuck-all" end of the spectrum, and no-one seems particularly interested in moving towards the center.   

[Baffler]

Who says we have to mess up the nice planets? If we can get between planets that easily it would be better to do most resource extraction and manufacturing on lifeless planets to avoid shitting up the ones we can actually live on.

[Eric Blank]

There's also smaller bodies we can use.

[Lagslayer]

Why would an AI want to kill everyone anyway? An AI with even the smallest insight into the world would know that's an awful idea.

Why would a human want to kill everyone anyway?

[Descan]

The odds of there being other sentient lifeforms in the universe: Pretty close to 100%
The odds of humanity ever making contact with them*: Pretty damn close to 0%

*assuming that FTL travel/communication is impractical.   

It would only take... about 50 million years? or so, for a single civilization to colonize the entire galaxy at sub-light speeds, assuming I think a few centuries between planets being seeded. So uh. Yeah, not 0%. We only need one interstellar civilization to arise sometime after the dinosaurs went extinct in our galaxy and we'd see 'em.

[Graknorke]

Why would an AI want to kill everyone anyway? An AI with even the smallest insight into the world would know that's an awful idea.

Why would a human want to kill everyone anyway?

Generally they don't. And the ones that do want to kill a lot of people generally do it based on incomplete/false information.

[Gentlefish]

There's also smaller bodies we can use.

Thiiiiiis. A medium-sized platinum-rich asteroid will have so much more platinum than is actually harvestable on earth.

[Bohandas]

Who says we have to mess up the nice planets? If we can get between planets that easily it would be better to do most resource extraction and manufacturing on lifeless planets to avoid shitting up the ones we can actually live on.

That's always been the issue I've had with alien invasion movies.

[i2amroy]

In a world with billions of people, millions of which are psychopaths, and a good number that are omnicidal, there's always gonna be someone. And these sorts of people are also more likely to to claw their way into positions of power. It only takes one guy who rolls a critical success, and pushing society as a whole down that path will only make it easier for that to happen.

If the AI is intelligent, it will eventually start doing unexpected things. That's the entire point of developing a true AI, that it can think for itself. Even if not explicitly told to do something, it has the capacity to learn.

That's like saying that upon us developing a weapon that lets you kill easily everyone is going to die because of some maniac, without considering the fact that it also lets us arm the police as well. As we can see with the development of the gun, this is not the case. Even if there are a few omnicidal AIs that show up, we'll have many more who's sole purpose is to hunt down and destroy said omnicidal AIs. And considering that most omnicidal AIs will be created with a single person's budget, while the hunter packs will be operating on equipment which we all have a vested interest in, I'm gonna bet on the hunters. (Not to even look at the fact that security AIs are going to have much more development going into them in the early stages of AI development than killer omnicidal AIs will.)

[TheDarkStar]

On the other hand, it only takes one successful omnicidal AI to end the world.

[Lagslayer]

And there aren't necessarily any do-overs.

Also consider that intelligent AI would not necessarily be governable by the humans that made them.

[Spehss _]

Why would an AI want to kill everyone anyway? An AI with even the smallest insight into the world would know that's an awful idea.

Generally the go-to answer for this is "logical bugs in the programming". Say an AI is programmed with a simple worded goal in mind, like "make people happy" or "protect people". As the AI grows in capacity and intelligence, it could develops new logical ways to fulfill its goal in the most efficient way possible. If its goal is to make people happy, maybe it would kidnap everyone, put every human's brain in a jar, and hook all the jars up to a machine to keep them living forever and constantly stimulate the regions of the brain responsible for pleasure, so everyone is always happy. If its goal is to protect people, it could lock everyone inside their homes, constantly provide all their needs, and make sure no one is ever in danger.