Print

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

[Realmfighter]

But what is its purpose in Optimization?

what's the purpose of an ai?

Well, in the Context of an AI created to make Utopia, Make Utopia.

[Askot Bokbondeler]

error, does not compute
define utopia, i'll make sure to pervert that into an indirect "kill all humans" command

[Realmfighter]

To Maximize Human Pleasure and Minimize Human Pain, the death of a Human Being equal in total pain to them spending their entire lives in agony.

If you can pervert that into something I would find Morally wrong you win a cookie.

[Askot Bokbondeler]

ok, i castrate all humans painlessly. i also kill all unborn babies, if death = (agony x entire life), then entire life = 0 means agony x 0

then i proceed to administrate powerful sedatives to the surviving humans and pleasure inducing drugs, they'll spend the rest of the longevity of the universe in eternal bliss

doesn't sound half as bad as i expected

fake edit:
realize all humans eventually die, and if their life is longer, then the total of agony they will feel by the time of their deaths will be much greater, therefore exterminate all humans immediately.

somebody mentioned a cookie?

[Realmfighter]

First one, Pain is produced from being Painlessly Castrated. Just because, say, your boss Doesn't punch you in the face while he is firing you from your job that allows you to support your family it still hurts like the dickens.

Few would go for the Powerful Sedatives path, creating Pain from fighting and then having their Control over their lives, and assuming the only way for there being no psychological damage from this procedure there being no psychology in the first place, the Pain and in a way death of having your mental processes being turned off to make you accept this.

Fake Edit:
That is stupid crap that sounds like a Nihilistic Child Cartoon Villains reasoning. An AI that kills people to Save them has quite a few major programming Flaws.

[SalmonGod]

what about when a large amount of people are jerks to a smaller amount of people? prejudice against minorities arises naturally in all societies

The discrimination wouldn't be institutionalized as it is in current societies, thus far less harmful to victims.  If discrimination by the majority is well-organized and backed by pools of resources, then it is an authoritative institution outside the scope of this discussion, as you're essentially responding to the notion of "We need to avoid X" with "What if avoiding X turns into X."  Not that it can't happen, but it's a failure to hold true to the principle rather than a fault in the principle itself.

Realmfighter and Askot Bokbondeler on robots and pain

The death = agony x entire life equation would have to set entire life as a constant rather than a variable, in order to place equal value on human life.

Everything else would hinge on the AI's ability to comprehend non-physical forms of pain, which is an interesting problem.

[Realmfighter]

The death = agony x entire life equation

Basically my stupid ass quickly thought up way of saying that the AI can't kill people, outside of extreme circumstances where death now would be less painful then continued existence.

[SalmonGod]

The death = agony x entire life equation

Basically my stupid ass quickly thought up way of saying that the AI can't kill people, outside of extreme circumstances where death now would be less painful then continued existence.

It would actually be a pretty useful judgement tool, considering any benevolent robotic overlords would be required to make life & death decisions using quantifiable objective logic.  You just can't leave room for variables in the equation to be interpreted as something that makes certain people expendable.  Actually, you could make it protective of children with death = agony for remainder of expected lifespan, but then you'd have to include more computations to prevent it devalueing people who are older but highly accomplished.  You could end up with something resembling typical human ethics, but you'd still run into the problem of comprehending non-physical pain.

[Realmfighter]

I don't really think that Non-Physical would be any harder to understand then Physical Pain. I'm not saying it would be easy, just that once you are at a point where you can understand others Physical Pain that Non-Physical pain wouldn't be much harder.

Physical Pain is your body telling you something is Bad. Non-Physical Pain is you experiencing something Bad. They are both infinitely complex and equally ungraspable by today's computers. Putting Limitations in our Metaphorical AI that doesn't exist and runs on Pseudo-Logic is not the best way to spend our time.

[Askot Bokbondeler]

hey, i'd get a cookie for finding flaws in your programming, if you go around correcting the flaws you're cheating!

Fake Edit:
That is stupid crap that sounds like a Nihilistic Child Cartoon Villains reasoning. An AI that kills people to Save them has quite a few major programming Flaws.

the major programming flaws are there, yes, but it was you who programmed it. the point of this exercise was me trying to pervert simple, apparently harmless rules into a "kill all humans" command, dont say it's silly just because you made it so easy for me.
you never defined what was an entire life and you never defined pain, i took these words in the most literal sense, if death equals every minute of your life in agony, and if death is unavoidable, then more minutes of life equals more minutes of suffering.

First one, Pain is produced from being Painlessly Castrated. Just because, say, your boss Doesn't punch you in the face while he is firing you from your job that allows you to support your family it still hurts like the dickens.

Few would go for the Powerful Sedatives path, creating Pain from fighting and then having their Control over their lives, and assuming the only way for there being no psychological damage from this procedure there being no psychology in the first place, the Pain and in a way death of having your mental processes being turned off to make you accept this.

i assumed the people had given the ai absolute power, back when it was sowelu's idea, since it was providing them with every resource they had.
it would very simple for a post scarcity, all powerful ai, to chemically castrate everybody with pharmaceutical substances in their food, then artificially create replicas of human children to disguise the act and prevent unrest, then simply deactivate and dismantle these fake humans once the last human dies of old age or is entered into an eternal state of blissful slumber

even if the ai isn't all powerful, it will figure that humans cause pain on themselves and each other, and may decide that going into a war with humanity, and hurting and killing a few thousand humans may be worth it in the long run, as the rest of humanity would spend their lives in eternal cryogenic bliss

what about when a large amount of people are jerks to a smaller amount of people? prejudice against minorities arises naturally in all societies

The discrimination wouldn't be institutionalized as it is in current societies, thus far less harmful to victims.  If discrimination by the majority is well-organized and backed by pools of resources, then it is an authoritative institution outside the scope of this discussion, as you're essentially responding to the notion of "We need to avoid X" with "What if avoiding X turns into X."  Not that it can't happen, but it's a failure to hold true to the principle rather than a fault in the principle itself.

i also think you're changing the rules. if i read you right the first time, your idea was that there shouldn't be any social mechanics by which a minority could impose their will over the majority:

This only requires one guiding principle, no opportunities within the social structure for the decisions of a small amount of people to have significant impact on a large amount of people.

there was still the possibility for a majority to impose their will over a minority, though. you may be underestimating the magnitudes of jerkiness a majority can reach

[Strife26]

Damn 0th rule of robotics!

[Askot Bokbondeler]

define "harm humanity"
define what's good and what's bad for humanity

[GlyphGryph]

OKay, here, use my AI instead:

Produce stuff. Get as much of what people want to them as possible, when they want, while minimizing risk to them in the process of creating and delivering that stuff.

[Realmfighter]

You Get a cookie.

But Back when this Conversation Stream started, you said that an AI might kill us in a effort to optimize its world. I was trying to show how this would be unlikely, given that Optimization of its World wouldn't be its Point. It's point would be something else, and assuming it was designed by Humans that Point would be benefiting Humankind. In Optimizing, it would destroy all possible reason for Optimizing. But Back to the Topic at hand!

Fake Edit:
That is stupid crap that sounds like a Nihilistic Child Cartoon Villains reasoning. An AI that kills people to Save them has quite a few major programming Flaws.

the major programming flaws are there, yes, but it was you who programmed it. the point of this exercise was me trying to pervert simple, apparently harmless rules into a "kill all humans" command, dont say it's silly just because you made it so easy for me.
you never defined what was an entire life and you never defined pain, i took these words in the most literal sense, if death equals every minute of your life in agony, and if death is unavoidable, then more minutes of life equals more minutes of suffering.

But I didn't program it. It doesn't exist. Anyone who creates an AI, gives it unlimited power and then says to it MAX PLEASURE MIN PAIN GOGOGO is setting themselves up for failure. And no one would do it for that exact reason. In fact, anything that hasn't proven itself capable of making the best decision, mote Humane decision and the most sane decision would never get complete control over humanity in the first place.

i assumed the people had given the ai absolute power, back when it was sowelu's idea, since it was providing them with every resource they had.
it would very simple for a post scarcity, all powerful ai, to chemically castrate everybody with pharmaceutical substances in their food, then artificially create replicas of human children to disguise the act and prevent unrest, then simply deactivate and dismantle these fake humans once the last human dies of old age or is entered into an eternal state of blissful slumber

even if the ai isn't all powerful, it will figure that humans cause pain on themselves and each other, and may decide that going into a war with humanity, and hurting and killing a few thousand humans may be worth it in the long run, as the rest of humanity would spend their lives in eternal cryogenic bliss

But to be able to perfectly Emulate humans, it must be able to empathise with humans. Its a vital part of who we are. And any being who can empathise with humans as another human (Which is what perfectly emulating a human would require) can understand why this would be unsatisfactory of just about anyone. I would even say that the being themselves would find this unsatisfactory.

define "harm humanity"
define what's good and what's bad for humanity

Harming Humans, as a social group spread over the world Simultaneously including and transcending the individual.
What is Bad is all that is considered bad, and good is what is considered good.
That is literally the most accurate answer you can get for that question.

[Askot Bokbondeler]

But I didn't program it. It doesn't exist. Anyone who creates an AI, gives it unlimited power and then says to it MAX PLEASURE MIN PAIN GOGOGO is setting themselves up for failure. And no one would do it for that exact reason. In fact, anything that hasn't proven itself capable of making the best decision, mote Humane decision and the most sane decision would never get complete control over humanity in the first place.

good point, and i got my cookie, so i'm dropping that part of the discussion. somebody capable of creating such an advanced ai would know much more of this stuff than you and me

But to be able to perfectly Emulate humans, it must be able to empathise with humans. Its a vital part of who we are. And any being who can empathise with humans as another human (Which is what perfectly emulating a human would require) can understand why this would be unsatisfactory of just about anyone. I would even say that the being themselves would find this unsatisfactory.

eh, i don't see it. there are human individuals uncapable of empathizing with other human beings, and often when they are arrested for being murderous psychopaths their neighbors and family often go to the television saying "he seemed such a nice person". empathy is emotional, an ai lacks emotions, by definition and nature, an ai is psychopathic.
also, these replicas could be human to all of us, but since they were created by the ai, the ai would have made just enough edits to get them out of it's "human" definition, but still fool us(or fool itself into believing the rules don't apply to these)

Harming Humans, as a social group spread over the world Simultaneously including and transcending the individual.
What is Bad is all that is considered bad, and good is what is considered good.
That is literally the most accurate answer you can get for that question.

that may be true, but it's still a terrible answer
how do you define harm caused onto an abstract concept like humanity? the destruction of humanity? one could argue convincingly that being wiped out would be the best thing that could happen to humanity, and he couldn't be proved wrong. in the end, humanity doesn't exist... we're just a bunch of things doing stuff, so before we go around telling robots not to harm humanity we first must decide, and state very clearly, what is humanity, and decide what harms it. it's not as simple as it sounds, it's not simple at all.