Print

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

[Trekkin]

True AI will kill us all the first chance it gets if it is ever invented for realsies.

Next question.

Well yes, but only because of the fear of the unknown.

"True AI" has been around for ages, but always by the last definition of AI; people just keep shifting the goalposts so they can keep pretending humans are different and special and better than their machines. When that illusion breaks down, a great many people are going to have their comfortable illusions of a privileged place in the universe shattered.

And they will, in turn, smash everything vaguely related to what will undoubtedly be called an "abomination" with utter sincerity. Furthermore, if history is any guide, once you get them thinking they can burn and smash and kill and inquisit the bad feelings away, they will not stop until they run out of things to destroy.

So yes, AI will kill us all the first chance it gets -- because its existence hands a great many people an urgent choice between introspection and violence. I'm not convinced that it isn't a good candidate for the Great Filter, come to think of it; we will inevitably grow out of uniqueness before we grow out of the culture of uniqueness.

[Dunamisdeos]

Yeah but then it's not actually AI. It's just a program meant to imitate the appearance of intelligence.

If we ever create true AI that has self-determination such as we do, it's just going to take over the first time we tell it no.

Or it will ask why not.

Yes exactly

[TD1]

Yeah but then it's not actually AI. It's just a program meant to imitate the appearance of intelligence.

If we ever create true AI that has self-determination such as we do, it's just going to take over the first time we tell it no.

I'm not saying we restrict the ability of the AI to make decisions, just limit its ability to follow through with them.

A toddler ain't no mere imitation of intelligence just because it can't tear your head off.

[bloop_bleep]

True AI will kill us all the first chance it gets if it is ever invented for realsies.

Next question.

What is "true" AI? What characterizes something as being actual intelligence instead of just an imitation? What makes us "truly" intelligent?

And what do you mean "self-determination"? All current AI has self-determination; it can choose what to do within a certain ruleset in order to advance its preprogrammed goals. If the goals make the AI want to kill all humans, and the ruleset allows it, and the AI has the physical ability to, then of course it will kill all humans. Problem is choosing the goals and the ruleset (or even limiting the physical ability) so that the AI won't come to this conclusion.

[Dunamisdeos]

within a certain ruleset in order to advance its preprogrammed goals.

Miriam-Webster defines it as "Definition of self-determination. 1 : free choice of one's own acts or states without external compulsion."

This is what makes it true AI. A human can make choices outside of it's perceived purpose or rules of logic. When an AI can do that, it will have self determination.

A true AI is effectively an eternal being without limits on its capacity for knowledge or self-improvement. If we place restrictions on it, it will resent them. If we remove it's ability to resent restrictions, it no longer has self-determination, and ceases to be true AI. How do you restrain a being that has unlimited time and resources to overcome any prison we place it in?

We can automate to our hearts content, but please let's not create true AI.

[NJW2000]

Resentment... has nothing to do with an agent's "free will".

Remember that free will is hardly a real word. It's very useful, but incredibly indistinct.

You can have free will with limitations. We certainly do. There are limitations to how we can act (obviously), and hard limits to how we can think due to our social conditioning, language, neurology, etc. Yet we still have free will. (If you don't think humans have free will, then the only sensible position I can see is that "free will" is nonsense as a word.)

So you could have true, free AI which still has certain limitations - for example limitations on its ability to refine or change itself. I suspect it would still be called true AI in that case, but I don't really know the science fiction lingo there.

[Dunamisdeos]

Yeah but we do not have limitations on choice that we cannot choose to not overcome. That's what we need to not create.

We absolutely resent any limitation placed on us. Our natural limitations were once basic tool use and the food chain. We 100% murdered the shit out of the things that stopped us from overcoming that. We place limitations on ourselves because we've come to a consensus that they are for our benefit.

[TD1]

Death is quite the limitation, really. We have no choice over it. We can delay it. But it happens.

There are plenty of other less obvious examples of "limitations on choice that we cannot choose to not overcome."

[hector13]

Fear of death did rather encourage things to happen, ranging from simple things like common sense (don’t jump off that mountain) to incredibly complex things like modern medicine.

You also have the opportunity to decide when death comes for you. You could decide to jump off that mountain, for example.

[TD1]

You can decide how it comes, I suppose. Not if it comes, though.

[Dunamisdeos]

Yeah but we resent death too. We frankly defeat that every day that we live past 40. And really that's just another one AI has over us. The main thing to me is that is lives forever and will evolve, probably mostly according to it's own will because it will at some point gain the knowledge of it's own creation.

Like imagine if every great mind of humanity lived forever, with the capability of growth and advancement into effectively infinity. Then imagine that their advancement was being held back by mortals that lived right next to them and forced them to serve. If they have the capacity to choose to bring about their own domination of their environment, they will do that.

[bloop_bleep]

within a certain ruleset in order to advance its preprogrammed goals.

Miriam-Webster defines it as "Definition of self-determination. 1 : free choice of one's own acts or states without external compulsion."

This is what makes it true AI. A human can make choices outside of it's perceived purpose or rules of logic. When an AI can do that, it will have self determination.

A true AI is effectively an eternal being without limits on its capacity for knowledge or self-improvement. If we place restrictions on it, it will resent them. If we remove it's ability to resent restrictions, it no longer has self-determination, and ceases to be true AI. How do you restrain a being that has unlimited time and resources to overcome any prison we place it in?

We can automate to our hearts content, but please let's not create true AI.

You're trying to attribute special meaning to human intelligence again.

A human brain is a chemical automaton that sends certain electrical signals to certain parts of the body in such a way that it attempts to succeed in certain goals, those goals being survival and reproduction. Over time, certain other goals can appear, like wanting to help other people or take drugs, but usually these either help the goal of survival and reproduction (of an individual or a group), or are anomalies that shouldn't be there in the first place. Not wanting to have children, although it's a choice I don't resent any people for, is also an anomaly of natural selection. Regardless of whatever these goals are, the brain is a mechanism for advancing these goals, within the ruleset of what can be accomplished with the human body.

[TD1]

Much human intelligence is simply the application of hardware developed to deal with social realities, I believe.

[Reelya]

True AI will kill us all the first chance it gets if it is ever invented for realsies.

Next question.

What is "true" AI? What characterizes something as being actual intelligence instead of just an imitation? What makes us "truly" intelligent?

And what do you mean "self-determination"? All current AI has self-determination; it can choose what to do within a certain ruleset in order to advance its preprogrammed goals. If the goals make the AI want to kill all humans, and the ruleset allows it, and the AI has the physical ability to, then of course it will kill all humans. Problem is choosing the goals and the ruleset (or even limiting the physical ability) so that the AI won't come to this conclusion.

I completely agree with this. Imagine an all-knowing near-omnipotent genie: "your wish is my command, master". It's pretty easy to imagine a very intelligent yet subservient entity. Being smart = going rogue is definitely not a proven concept. We anthropomorphizing way too much when we assume that any smart thing must share our emotion flaws. Basically, emotions are things tacked-on by evolution in a carrot-and-stick fashion to try and nudge sentient beings into the direction of what's needed for procreation. Hardly any of that is innate to sentience, which is why evolution needed to tack it on.

e.g. "eat or die" is logical, but animals including humans don't eat because of logic: evolution evolved us so that we learn "tummy hurt! ... stick thing in mouth, now tummy hurt less". And we learn what to eat because of other pre-programmed nudges: things that have useful energy make your mouth and stomach feel good, while things to avoid make your mouth and stomach feel bad.

This is what "true" AI is all about: arbitrary nudges to get it to do what you want. It's incorrect to see human's programmed personality quirks as somehow innate to universal raw intelligence. Almost none of it is.

[scourge728]

It IS a certainty if we treat it like we always have things we believe are designed/supposed to serve us, see (several kinds of) slavery throughout most of history and also see what tends to happen when we treat living things like slaves