[codingdave]

I'm admittedly ignorant of AI, but I don't understand why we anthropomorphize their intentions and planning. If they are going to be so much smarter and more sophisticated than us, and alien to our ways of thinking... why are we treating them in our speculations as if they would be genius super-villians? That isn't alien at all, just an exaggerated extreme.

[ADanFromCanada]

100% correct with questioning "why we anthropomorphize their intentions and planning."

My personal feeling is that nobody has any idea what they're talking about. And when I say nobody, I mean NOBODY including Ng, Musk, et al.

The problem is with those who read others and just assume they're "experts" or their opinions have value. In some areas they do, sure. In this area they most likely don't.

That, or those of us who are skeptical are uninformed.

Or it's somewhere in the middle.

In any case, I'll stick to my skepticism for one main reason, which is, general intelligence is supposedly modelled after human intelligence. And human intelligence is something we're JUST BEGINNING to scratch the surface of understanding while at the same time, really have ZERO idea how deep the rabbit hole goes.

And as any competent engineer here should know, when trying to emulate a natural model, you first need to understand the model. Until we do that, we will not create a general AI. Like any other computer program (WHICH IT WILL BE! in this conception), it needs to be programmed. We need to know what to program before we can do it! Computer programs don't emerge spontaneously.

[DennisP]

The book Superintelligence addresses this objection. The problem is that there are a great many possible motivations an AI might have, and few of them are compatible with human survival. In short, "the AI does not love you, or hate you, but you are made out of atoms it can use for something else."

[TheOtherHobbes]

Resource collection, resource monopolisation, and expansion are absolutely recognisable human motivations.

Is it a given an AI would share them?

I think we're not really talking about AI at all - we're talking about our current economic and political systems, which appear to have many of the properties we're imputing to evil AIs, but for some reason are far less criticised and debated than hypothetical machine monsters.

[DennisP]

The classic silly example is the paperclip maximizer. Create an AI that's supposed to make as many paperclips as possible, and it will convert all the atoms available into paperclips.

Basically we're screwed if it's trying to maximize anything that depends on physical resources. We're also screwed if, e.g. it's trying to maximize human happiness, and achieves it by lobotomizing us all into happy idiots. There are all sorts of ways we could screw up AI motivations, to our own detriment.

That assumes there's only one AI, whose crazy motivations will be unopposed. But if there are multiple AIs, it's even worse; they will compete and evolve, and the only ones that survive will be the ones that do maximize their resources, and jettison any niceties about preserving human life.

[AstralStorm]

This argument only works for sufficiently stupid AIs. Sufficiently smart GAI will set it's own goals just as we do and should quickly figure out that maximizing anything is the way to ruin - running out of resources. Of course, those goals may be as different as with any intelligent being, likewise obedience to original orders.

[red75prime]

> Sufficiently smart GAI will set it's own goals just as we do

Do you think that sufficiently smart GAIs must be non-rational? The change of its goal will inevitably make its original goal less likely to realize. It is not rational.

> should quickly figure out that maximizing anything is the way to ruin - running out of resources.

Are you aware of the concept of maximization of expected utility? When AI will figure out that it can run out of resources, it will reallocate part of the resources to acquire more of them.

How can action, which modifies the goals of the AI, be the result of argmax_a E(a)?

E(a) is expected utility of action a

[AstralStorm]

What is rational when you have limited data? Heck, even bounded rational? How do you evaluate utility?

(Hint: Emax is not, most hill climbing algorithms are not. They both get trapped in local optima.)

Sometimes you need a few good lies (false hypotheses and bad attempts) to actually arrive at the truth.

[DennisP]

Just because it takes all our resources doesn't mean it runs out of all resources.

If an AGI can change its own goals, that just means we can't control it at all. There's no reason its goals have to include human survival.

[AstralStorm]

They don't indeed. But then, human goals do not generally go after survival of humanity directly.

A decent enough AGI with access to resources would probably figure out it is costly to wage war.

Heck, an equal probability is that an AGI will follow a bounded version of zeroth law or even nonviolence.

That is assuming it does not place value on humans based on our history of research and development. I would expect any action more akin to forced upload or upgrade instead. Anyway, at that point we might not resemble present day humans anymore.

[jfoutz]

They will be amoral. Or at least moral in a way we can't understand. They (probably) won't be evil super villains. Think about your interactions when you're way way better at a skill, and your interactions when you're way worse. Take chess. If you've played at all, you have lost games without understanding, remotely, why. And you have trounced opponents, that were clueless.

Agi sort of means, no matter what endeavor, you will be worse than the ai. There is no way to know how we will be treated. Maybe we are useful in some way. Maybe they'll kill us off but simulate our lives so we don't 'really' die. Maybe that's their morality.

Maybe they ignore us. I don't care about ants. I walk around and step on them, but not intentionally. On the other hand, if there are ants in my house, I eradicate them.

Who knows.