I’m not sure I care.
For example if I had my evil way and I went FOOM then part of my optimization process would involve mind control and somewhat deviant roleplay with certain porno actresses. Would I want those actresses to be controlled against their will? Probably not. But at the same time it would be good enough if they were able to simulate being the actresses in a way that I could not tell the difference between the original and the simulated.
You wouldn’t prefer to forego the deviant roleplay for the sake of, y’know, not being evil?
But that’s not the point, I suppose. It sounds like you’d take the Experience Machine offer. I don’t really know what to say to that except that it seems like a whacky utility function.
How is the deviant roleplay being evil if the participants are not being coerced or are catgirls? And if it’s not being evil then how would I be defined as evil just because I (sometimes—not always) like deviant roleplay?
That’s the cruz of my point. I don’t reckon that optimizing humanity’s utility function is the opposite of unfriendly AI (or any individual’s for that matter) and I furthermore reckon that trying to seek that goal is much, much harder than trying to create an AI that at a minimum won’t kill us all AND might trade with us if it wants to.
Oh, sorry, I interpreted the comment incorrectly—for some reason I assumed your plan was to replace the actual porn actresses with compliant simulations. I wasn’t saying the deviancy itself was evil. Remember that the AI doesn’t need to negotiate with you—it’s superintelligent and you’re not. And while creating an AI that just ignores us but still optimises other things, well, it’s possible, but I don’t think it would be easier than creating FAI, and it would be pretty pointless—we want the AI to do something, after all.
Therein lies the crux: you want the AI to do stuff for you.
EDIT: Oh yeah I get you. So it’s by definition evil if I coerce the catgirls by mind control.
I suppose logically I can’t have my cake and eat it since I wouldn’t want my own non-sentient simulation controlled by an evil AI either.
So I guess that makes me evil. Who would have thunk it. Well I guess strike my utility function of the list of friendly AIs. But then again I’ve already said that elsewhere that I wouldn’t trust my own function to be the optimal.
I doubt, however, that we’d easily find a candidate function from a single individual for similar reasons.
I think we’ve slightly misunderstood each other. I originally thought you were saying that you wanted to destructively upload porn actresses and then remove sentience so they did as they were told—which is obviously evil. But I now realise you only want to make catgirl copies of porn actresses while leaving the originals intact (?) - the moral character of which depends on things like whether you get the consent of the actresses involved.
But yes! Of course I want the AGI to do something. If it doesn’t do anything, it’s not an AI. It’s not possible to write code that does absolutely nothing. And while building AGI might be a fun albeit stupidly dangerous project to pursue just for the heck of it, the main motivator behind wanting the thing to be created (speaking for myself) is so that it can solve problems, like, say, death and scarcity.
Correct. I (unlike some others) don’t hold the position that a destructive upload and then a simulated being is exactly the same being therefore destructively scanning the porn actresses would be killing them in my mind.
Non destructively scanning them and them using the simulated versions for “evil purposes”, however, is not killing the originals. Whether using the copies for evil purposes even against their simulated will is actually evil or not is debatable. I know some will take the position that the simulations could theoretically be sentient, If they are sentient then I am therefroe de facto evil.
And I get the point that we want to get the AGI to do something, just that I think it will be incredibly difficult to get it to do something if it’s recursively self improving and it becomes progressively more difficult to do the further away you go from defining friendly as NOT(unfriendly).
Why is it recursively self-improving if it isn’t doing anything? If my end goal was not to do anything, I certainly don’t need to modify myself in order to achieve that better than I could achieve it now.
Well, I would argue that if the computer is running a perfect simulation of a person, then the simulation is sentient—it’s simulating the brain and is therefore simulating consciousness, and for the life of me I can’t imagine any way in which “simulated consciousness” is different from just “consciousness”.
I think it will be incredibly difficult to get it to do something if it’s recursively self improving and it becomes progressively more difficult to do the further away you go from defining friendly as NOT(unfriendly).
I disagree. Creating a not-friendly-but-harmless AGI shouldn’t be any easier than creating a full-blown FAI. You’ve already had to do all the hard working of making it consistent while self-improving, and you’ve also had the do the hard work of programming the AI to recognise humans and to not do harm to them, while also acting on other things in the world. Here’s Eliezer’s paper.
I’m not sure I care. For example if I had my evil way and I went FOOM then part of my optimization process would involve mind control and somewhat deviant roleplay with certain porno actresses. Would I want those actresses to be controlled against their will? Probably not. But at the same time it would be good enough if they were able to simulate being the actresses in a way that I could not tell the difference between the original and the simulated.
Others may have different opinions.
You wouldn’t prefer to forego the deviant roleplay for the sake of, y’know, not being evil?
But that’s not the point, I suppose. It sounds like you’d take the Experience Machine offer. I don’t really know what to say to that except that it seems like a whacky utility function.
How is the deviant roleplay being evil if the participants are not being coerced or are catgirls? And if it’s not being evil then how would I be defined as evil just because I (sometimes—not always) like deviant roleplay?
That’s the cruz of my point. I don’t reckon that optimizing humanity’s utility function is the opposite of unfriendly AI (or any individual’s for that matter) and I furthermore reckon that trying to seek that goal is much, much harder than trying to create an AI that at a minimum won’t kill us all AND might trade with us if it wants to.
Oh, sorry, I interpreted the comment incorrectly—for some reason I assumed your plan was to replace the actual porn actresses with compliant simulations. I wasn’t saying the deviancy itself was evil. Remember that the AI doesn’t need to negotiate with you—it’s superintelligent and you’re not. And while creating an AI that just ignores us but still optimises other things, well, it’s possible, but I don’t think it would be easier than creating FAI, and it would be pretty pointless—we want the AI to do something, after all.
A-Ha!
Therein lies the crux: you want the AI to do stuff for you.
EDIT: Oh yeah I get you. So it’s by definition evil if I coerce the catgirls by mind control. I suppose logically I can’t have my cake and eat it since I wouldn’t want my own non-sentient simulation controlled by an evil AI either.
So I guess that makes me evil. Who would have thunk it. Well I guess strike my utility function of the list of friendly AIs. But then again I’ve already said that elsewhere that I wouldn’t trust my own function to be the optimal.
I doubt, however, that we’d easily find a candidate function from a single individual for similar reasons.
I think we’ve slightly misunderstood each other. I originally thought you were saying that you wanted to destructively upload porn actresses and then remove sentience so they did as they were told—which is obviously evil. But I now realise you only want to make catgirl copies of porn actresses while leaving the originals intact (?) - the moral character of which depends on things like whether you get the consent of the actresses involved.
But yes! Of course I want the AGI to do something. If it doesn’t do anything, it’s not an AI. It’s not possible to write code that does absolutely nothing. And while building AGI might be a fun albeit stupidly dangerous project to pursue just for the heck of it, the main motivator behind wanting the thing to be created (speaking for myself) is so that it can solve problems, like, say, death and scarcity.
Technically, it’s still an AI, it’s just a really useless one.
Exactly.
So “friendly” is therefore a conflation of NOT(unfriendly) AND useful rather than just simply NOT(unfriendly) which is easier.
Off. Do I win?
You’re determined to make me say LOL so you can downvote me right?
EDIT: Yes you win. OFF.
Correct. I (unlike some others) don’t hold the position that a destructive upload and then a simulated being is exactly the same being therefore destructively scanning the porn actresses would be killing them in my mind. Non destructively scanning them and them using the simulated versions for “evil purposes”, however, is not killing the originals. Whether using the copies for evil purposes even against their simulated will is actually evil or not is debatable. I know some will take the position that the simulations could theoretically be sentient, If they are sentient then I am therefroe de facto evil.
And I get the point that we want to get the AGI to do something, just that I think it will be incredibly difficult to get it to do something if it’s recursively self improving and it becomes progressively more difficult to do the further away you go from defining friendly as NOT(unfriendly).
Why is it recursively self-improving if it isn’t doing anything? If my end goal was not to do anything, I certainly don’t need to modify myself in order to achieve that better than I could achieve it now.
Isn’t doing anything for us…
Well, I would argue that if the computer is running a perfect simulation of a person, then the simulation is sentient—it’s simulating the brain and is therefore simulating consciousness, and for the life of me I can’t imagine any way in which “simulated consciousness” is different from just “consciousness”.
I disagree. Creating a not-friendly-but-harmless AGI shouldn’t be any easier than creating a full-blown FAI. You’ve already had to do all the hard working of making it consistent while self-improving, and you’ve also had the do the hard work of programming the AI to recognise humans and to not do harm to them, while also acting on other things in the world. Here’s Eliezer’s paper.
OK give me time to digest the jargon.