Very well said. Even if we achieve a full-fledged AGI, it would be a mistake to anthropomorphize it and assign it human-like intentions, desires and behavior - unless somebody explicitly programmed it that way. The idea of an "evil" AI seems downright silly to me.
That's not to say that an AI could never be dangerous in some scenario, but the "demon" comparisons and other recent hyperbole are, IMO, a bit misplaced.
We cannot say that AI's will not be dangerous just because they will not have human-like behaviour. We shouldn't let inaccurate sci-fi movies cloud our judgement like that. Just because the future will most certainly not look like Terminator does not mean AI's are not a danger. The Elon Musk comment about the demon was not anthropomorphizing AI's; he was saying the invention of superintelligent general AI's could be like summoning the demon and trying to contain it in the sense that it could be a point of no return. We may find out after the fact that we made a huge mistake and cannot control it. I think the remarks of Elon Musk and Stephen Hawkings sound very alarmist, but they want to make sure that we will think about the implications of superintelligent AIs now, and not after we invent them. I am sure we can think of proper ways to 'contain the demon', but we need to think about that now and not when they are already 'set free'.
We cannot say that AI's will not be dangerous just because they will not have human-like behaviour.
I agree, and I'm not making a blanket statement like that. But I think that a lot of the recent hyperbole about AI seems to assume an AGI with "bad intention" or dangerous behavior rooted in some seemingly anthrocentric notion of goals, desires, etc. To the extent that people are saying any of those things, I think they're barking up the tree, since AI - no matter how intelligent - still isn't human.
Now, could an AGI still be dangerous, whether intentionally or by accident? Yes, and I'd guess the "by accident" bit is more likely. I don't think this is an issue that should be ignored, and I don't expect that it will be. But whipping up public hysteria over "unleashing the demon" and "AI could end the world" strikes me as overly reactionary. But that's just me.
IMHO it's the "artificial" part we need to worry about. Those machines that turn themselves off every time you switch them on (as invented by Minsky and Shannon)? Imagine that kind of logic accidentally controlling a crucial system, and operating too quickly to be interrupted.
I think the most important point is "Additionally, AI is not a single entity. ... AI is not a he or a she or even an it, AI is more like a 'they.'"
All the horrible "clippy" scenarios involve ONE AI that becomes super intelligent (and therefore powerful -- another fallacy) without any similarly intelligent and powerful entities around it. Instead we'll have incremental progress and if we ever do get super intelligent (but probably not super powerful) machines they'll be embedded in an ecology of other machines nearly as intelligent and quite likely more powerful.
I'm not saying this doesn't pose risks, but they aren't the risks that the AGI threat folks are studying.
7 comments
[ 4.3 ms ] story [ 17.3 ms ] threadThat's not to say that an AI could never be dangerous in some scenario, but the "demon" comparisons and other recent hyperbole are, IMO, a bit misplaced.
I agree, and I'm not making a blanket statement like that. But I think that a lot of the recent hyperbole about AI seems to assume an AGI with "bad intention" or dangerous behavior rooted in some seemingly anthrocentric notion of goals, desires, etc. To the extent that people are saying any of those things, I think they're barking up the tree, since AI - no matter how intelligent - still isn't human.
Now, could an AGI still be dangerous, whether intentionally or by accident? Yes, and I'd guess the "by accident" bit is more likely. I don't think this is an issue that should be ignored, and I don't expect that it will be. But whipping up public hysteria over "unleashing the demon" and "AI could end the world" strikes me as overly reactionary. But that's just me.
Well, yeah.
All the horrible "clippy" scenarios involve ONE AI that becomes super intelligent (and therefore powerful -- another fallacy) without any similarly intelligent and powerful entities around it. Instead we'll have incremental progress and if we ever do get super intelligent (but probably not super powerful) machines they'll be embedded in an ecology of other machines nearly as intelligent and quite likely more powerful.
I'm not saying this doesn't pose risks, but they aren't the risks that the AGI threat folks are studying.