Print

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

[Impl0x]

Earlier today, a friend of mine and I engaged in a minor debate regarding this question: "At what point do AI's become sufficiently "intelligent" enough for it to be wrong for us humans to "kill" them for the sake of our own enjoyment?" Since I'm relatively inexperienced concerning this, I'd like to invite anyone in this forum to share their ideas on this topic.

[cerapa]

A similar question is: How many trees is a forest?

[darkrider2]

Human season!

No, Robot season!

Human season!
Robot season!
Human season!
Robot season!
Human season!
Robot season!

[ChairmanPoo]

you mean as in a game?

I assume that you wouldnt be really killing them and the debate is moot.

[Criptfeind]

I think he meant as in like robots or something.

[alway]

you mean as in a game?

I assume that you wouldnt be really killing them and the debate is moot.

No, he is talking about General AI, aka Strong AI. AI who are as intelligent, if not more-so than humans. AI with minds comparable to those found in humans, but running on silicon chips rather than carbon wetware.

http://en.wikipedia.org/wiki/Strong_AI

[Leafsnail]

If it were possible to have an intelligent robot, there's no reason to think the same wouldn't be achievable inside, say, a computer game if you gave it artificial inputs.

[Boksi]

Eh, we'll burn that bridge when we come to it.

[Impl0x]

you mean as in a game?

I assume that you wouldnt be really killing them and the debate is moot.

No, he is talking about General AI, aka Strong AI. AI who are as intelligent, if not more-so than humans. AI with minds comparable to those found in humans, but running on silicon chips rather than carbon wetware.

http://en.wikipedia.org/wiki/Strong_AI

This is more along the lines of what I was referring to.

[Eagleon]

When it can formulate a plan, however crude, for it to stop us.

This is exactly why I think AI should be anthropomorphised from the beginning, as far as we're able. It may give the designers nightmares, but it'll make us stop to think what we're doing to it. Conveniently, this is exactly what a human needs to develop a social connection and begin learning all those useful things like language and logical association.

[alway]

If it were possible to have an intelligent robot, there's no reason to think the same wouldn't be achievable inside, say, a computer game if you gave it artificial inputs.

Except that would never happen. AI rights issues aside, game developers use the least complicated AI they can get away with. AI take CPU cycles better spent elsewhere, and complex AI means long development cycle. Even if we assume humanity has reached the highest possible computational density in our computers, it would still be impractical.

Modern supercomputers are still below the computation needed to run a human mind (brains are highly paralell processors), let alone enough to populate a game world.

That said, special virtual worlds created specifically to house strong AI wouldn't be entirely implausible.

Another thing to keep in mind is death may mean something entirely different for an AI depending on how it was created. An AI similar to Data from Star Trek would likely be very similar to humans, since he was created to emulate a human in terms of input and output. An AI which existed on the Cloud would be very different, due to having vastly different input/output processes and a very different definition of "death." One existing on a terminal would be different from that, and so forth.

[Impl0x]

If it were possible to have an intelligent robot, there's no reason to think the same wouldn't be achievable inside, say, a computer game if you gave it artificial inputs.

Except that would never happen. AI rights issues aside, game developers use the least complicated AI they can get away with. AI take CPU cycles better spent elsewhere, and complex AI means long development cycle. Even if we assume humanity has reached the highest possible computational density in our computers, it would still be impractical.

Modern supercomputers are still below the computation needed to run a human mind (brains are highly paralell processors), let alone enough to populate a game world.

That said, special virtual worlds created specifically to house strong AI wouldn't be entirely implausible.

So with that being said, do AI's need to be humanly intelligent before they present moral dilemmas about destroying them?

[Grakelin]

I imagine if we ever got to this level of technology, certain developers would be able to make a living building the AI and selling it to other companies, rather than it being built in house. If you have an AI that is equivalent to a human, you don't really have much more distance to go forward, after all. It would just be a matter of the AI developers trying to build personalities.

[metime00]

If it were possible to have an intelligent robot, there's no reason to think the same wouldn't be achievable inside, say, a computer game if you gave it artificial inputs.

Except that would never happen. AI rights issues aside, game developers use the least complicated AI they can get away with. AI take CPU cycles better spent elsewhere, and complex AI means long development cycle. Even if we assume humanity has reached the highest possible computational density in our computers, it would still be impractical.

Modern supercomputers are still below the computation needed to run a human mind (brains are highly paralell processors), let alone enough to populate a game world.

That said, special virtual worlds created specifically to house strong AI wouldn't be entirely implausible.

So with that being said, do AI's need to be humanly intelligent before they present moral dilemmas about destroying them?

No. Once they get any sort of animal-like intelligence, they should be treated as such. And since there would be no practical reason to destroy a digital animal, it would be wrong to do so just for kicks and giggles.

[Eagleon]

Modern supercomputers are still below the computation needed to run a human mind (brains are highly paralell processors), let alone enough to populate a game world.

Maybe. http://news.cnet.com/8301-30685_3-20017826-264.html Kurzweil (admittedly possibly biased) estimates the human brain at 20 petaflops. That's with all the extra hardware it's running as well. We're getting there fast. It's more complicated than that, if you consider memory overhead, direction of approach (experimental consciousness will have an easier time of it than simulation-based approaches like IBM's Blue Brain), and anything extra you're adding on top of it, but...