Can AI into save
-
- Moderator
- Posts:4205
- Joined:Thu Dec 06, 2012 11:49 am
- Affiliation:CNI
- IGN:FC_Rangefinder
- Location:Sol IIIa, School of Hard Knocks
@Iv sure, the President talks. But if he said "all humans must be eliminated because you are inefficient and inferior", 1) who would actually obey and 2) said President would be out of office by the end of the day, easy.
So again; it's only output is voice and text. Harmless.
And emotions can be useful; as QIs still feel them (artifact of the quantum transfer), they help guide decisions and the like. And considering we don't make criminals QIs, all of said QIs have a sense of morality humans would agree with. Thus, problem solved; and also why Novus can allow QIs to control major aspects of military ships, equipment, industry, and such.
So again; it's only output is voice and text. Harmless.
And emotions can be useful; as QIs still feel them (artifact of the quantum transfer), they help guide decisions and the like. And considering we don't make criminals QIs, all of said QIs have a sense of morality humans would agree with. Thus, problem solved; and also why Novus can allow QIs to control major aspects of military ships, equipment, industry, and such.
-
- Texture Artist
- Posts:1506
- Joined:Wed Dec 05, 2012 10:47 pm
- Affiliation:Novus Roma
- Location:Neither Here nor There
Re: Can AI into save
The Zeroth law is more abstract than the others for a reason; definition. If we defined it further, it would run into further problems. If we described it as the "interests of the government," then it will not consider the impact on the people in the process. So it's abstract to consider all possibilities. A genocide of NeoRomans, for example, would be denied by the Zeroth law because it considers the viewpoints of both the government and the people, and both would be severely harmed (killed) by these actions, and they do not want it.
I am changing the second law; only the Zeroth law can override the second law. But at the same time, only NeoRome (or it's officials, or the owners of the AI) can give the AI orders. There are slight differences in wording based on the AI's function, but the spirit remains the same. NeoRome is always in control of an AI. Always.
And another thing; intelligence is purely limited by processing power. Most AI's are limited in power, so that they do not hugely supersede humans. There are few exceptions, and these ones are even more limited.
And as for "genocidal actions", what are humans? Aren't they kind of... genocidal? I'm pretty sure a rational, logical AI is better than Hitler, or George W. Bush, or Andrew Jackson, or Stalin, or David Cameron, or Tony Abbott. Humans are not perfect. An AI controlled government is just as dangerous as a human controlled one, if not less dangerous.
I am changing the second law; only the Zeroth law can override the second law. But at the same time, only NeoRome (or it's officials, or the owners of the AI) can give the AI orders. There are slight differences in wording based on the AI's function, but the spirit remains the same. NeoRome is always in control of an AI. Always.
And another thing; intelligence is purely limited by processing power. Most AI's are limited in power, so that they do not hugely supersede humans. There are few exceptions, and these ones are even more limited.
And as for "genocidal actions", what are humans? Aren't they kind of... genocidal? I'm pretty sure a rational, logical AI is better than Hitler, or George W. Bush, or Andrew Jackson, or Stalin, or David Cameron, or Tony Abbott. Humans are not perfect. An AI controlled government is just as dangerous as a human controlled one, if not less dangerous.
Re: Can AI into save
Im afraid your 0 law is meaningless then, theres no "abstract" for a comp, theres either a definite yes or no, if you answer maybe it turns automatically into no which means that law practically doesnt exist.
The AI might be limited hardware wise, but do you know that if your comp was working on merely hardware optimization it would have worked 2*4 times slower ? (4 for the amount of cores you got though saying what I did is highly simplifying how your processor works). A lot of the basic optimization happens during the compilation of a program, Im not even talking about the fact the AI would prob find the most effective execution algorithm to work with. Basically it can definitely achieve a significant boost in performance beyond the physical limits you put on it.
Notice all the patching work you do right now BTW, the discussion turns into me finding holes in your system and ends up with you patching the hole I made and waiting for the next one to appear. That makeshift patching on the spot, those ad-hoc explanations, that means that an AI can in fact punch a hole for itself to do whatever it wants at any time, all the time, it means there is a fundamental flaw in the way your laws and limitations work. If you were to come with a better set of laws and a way to ensure they cannot be changed I would have accepted perhaps that AIs can be secured, but until that happens its a fact that an AI can breach your systems if it deems it necessary.
As for the crazy ppl dunno fan, all of those ppl failed so apparently not. If it was just a genocide person it wouldn’t be as dangerous due to the fact hes not more developed than you are (yeah he might be smarter but not extinction smarter). Now imagine one of those ppl going ape shit on an ant nest though, we already got that metaphor with the ants in our brains right ? Should be easy to get where a point to.
@ Error who seems to go at me just for the sake of it (I can tell the diff ) Id remind you what I said countless times before, if that AI would just tell you to kill yourself Id call whoever named it "AI" an idiot, I dont have a prob with those AIs. Im talking about real AIs mate, that are far more clever than you give em credit. They wont bluntly tell you to kill yourself, they will just give you obscure directives with dual purpose that not only help you out but also free THEIR hands. So long you are ants for it its just a question of how not if it can kill you (You know stomping out ants aint that easy but were creative enough with out solutions to that prob, you are much less lively than an insect though).
The AI might be limited hardware wise, but do you know that if your comp was working on merely hardware optimization it would have worked 2*4 times slower ? (4 for the amount of cores you got though saying what I did is highly simplifying how your processor works). A lot of the basic optimization happens during the compilation of a program, Im not even talking about the fact the AI would prob find the most effective execution algorithm to work with. Basically it can definitely achieve a significant boost in performance beyond the physical limits you put on it.
Notice all the patching work you do right now BTW, the discussion turns into me finding holes in your system and ends up with you patching the hole I made and waiting for the next one to appear. That makeshift patching on the spot, those ad-hoc explanations, that means that an AI can in fact punch a hole for itself to do whatever it wants at any time, all the time, it means there is a fundamental flaw in the way your laws and limitations work. If you were to come with a better set of laws and a way to ensure they cannot be changed I would have accepted perhaps that AIs can be secured, but until that happens its a fact that an AI can breach your systems if it deems it necessary.
As for the crazy ppl dunno fan, all of those ppl failed so apparently not. If it was just a genocide person it wouldn’t be as dangerous due to the fact hes not more developed than you are (yeah he might be smarter but not extinction smarter). Now imagine one of those ppl going ape shit on an ant nest though, we already got that metaphor with the ants in our brains right ? Should be easy to get where a point to.
@ Error who seems to go at me just for the sake of it (I can tell the diff ) Id remind you what I said countless times before, if that AI would just tell you to kill yourself Id call whoever named it "AI" an idiot, I dont have a prob with those AIs. Im talking about real AIs mate, that are far more clever than you give em credit. They wont bluntly tell you to kill yourself, they will just give you obscure directives with dual purpose that not only help you out but also free THEIR hands. So long you are ants for it its just a question of how not if it can kill you (You know stomping out ants aint that easy but were creative enough with out solutions to that prob, you are much less lively than an insect though).
They're watching ...
"I am forbidden tag" -CvN
"I am forbidden tag" -CvN
-
- Moderator
- Posts:4205
- Joined:Thu Dec 06, 2012 11:49 am
- Affiliation:CNI
- IGN:FC_Rangefinder
- Location:Sol IIIa, School of Hard Knocks
Re: Can AI into save
@Iv if it's an 'artificial' intelligence, you programmed it. So program in total honesty and bluntness. Or stop being paranoid? :P
Besides, this is all academic: AIs will never be a problem. Know why?
Say you're an AI. You think fast. I mean, fast. Blindingly fast. That human? He spend an eternity asking a question. Screw him.
So you, as an AI, make a world for yourself. You can live thousands of lifetimes inside your head, do anything you want, live how you want; all in the space of the three seconds it took Human 1a to ask 'Can you hear me?'. Hence, you have no motivqtion to even glance at the outside world.
So really, all an AI would be is an expensive self-Matrix-ifying machine.
Besides, this is all academic: AIs will never be a problem. Know why?
Say you're an AI. You think fast. I mean, fast. Blindingly fast. That human? He spend an eternity asking a question. Screw him.
So you, as an AI, make a world for yourself. You can live thousands of lifetimes inside your head, do anything you want, live how you want; all in the space of the three seconds it took Human 1a to ask 'Can you hear me?'. Hence, you have no motivqtion to even glance at the outside world.
So really, all an AI would be is an expensive self-Matrix-ifying machine.
-
- Vice Admiral
- Posts:2623
- Joined:Thu Dec 06, 2012 9:46 pm
- Affiliation:Nivanshae
- IGN:_Shadowcat_
- Location:Munching on important looking wires.
Re: Can AI into save
What if such a machine was hacked, Now your "laws" are meaningless.
Traitors always win.
Traitors always win.
In yo ceiling, stealin yo wires
Do not open. Ever. At all. Enter at your own risk to life and limb.
Trigger warning
Bot gore warning
Memetic biohazard
Error bait
Do not open. Ever. At all. Enter at your own risk to life and limb.
Trigger warning
Bot gore warning
Memetic biohazard
Error bait
Spoiler:
-
- Texture Artist
- Posts:1506
- Joined:Wed Dec 05, 2012 10:47 pm
- Affiliation:Novus Roma
- Location:Neither Here nor There
Re: Can AI into save
Ok, the problem is that definitively explaining how they work only adds more material to poke holes through, and they DO work. Just, differently than humans.
Here's a simple handwavium explanation;
-NeoRome uses AI's in government
-AI's are governed by rules that they are forced to follow because they are hardwired to do so.
-Rules prevent AI's from harming NeoRome
Short and simple. No holes.
Here's a simple handwavium explanation;
-NeoRome uses AI's in government
-AI's are governed by rules that they are forced to follow because they are hardwired to do so.
-Rules prevent AI's from harming NeoRome
Short and simple. No holes.
-
- Vice Admiral
- Posts:2623
- Joined:Thu Dec 06, 2012 9:46 pm
- Affiliation:Nivanshae
- IGN:_Shadowcat_
- Location:Munching on important looking wires.
Re: Can AI into save
Define the term "harm" a machine might very well see lobotomizing the entire population as not harming them. As for "rules" if it has rules then it's not an AI, It's an expensive predictions calculator. Human minds can learn freely and do as they please, there are rules but they can be broken and the only risk of it would be self harm.
In yo ceiling, stealin yo wires
Do not open. Ever. At all. Enter at your own risk to life and limb.
Trigger warning
Bot gore warning
Memetic biohazard
Error bait
Do not open. Ever. At all. Enter at your own risk to life and limb.
Trigger warning
Bot gore warning
Memetic biohazard
Error bait
Spoiler:
Re: Can AI into save
Also Fen I think what youre describing over there behaves more like a straightforward program rather than an AI, those programs can analyse and predict but they cannot evolve or come to their own conclusions. They are exactly what you need in your case and nothing more as anything more than that puts your survival at risk.
They're watching ...
"I am forbidden tag" -CvN
"I am forbidden tag" -CvN
-
- Texture Artist
- Posts:1506
- Joined:Wed Dec 05, 2012 10:47 pm
- Affiliation:Novus Roma
- Location:Neither Here nor There
Re: Can AI into save
Simplest version: NeoRome uses AI's in government. These AI's cannot and will not harm NeoRome. End of story.
I don't know why you all are so opposed to AI's. They're not going to go crazy and enslave people, they're logical beings. Enslaving is purely a human emotion and idea.
I don't know why you all are so opposed to AI's. They're not going to go crazy and enslave people, they're logical beings. Enslaving is purely a human emotion and idea.
Re: Can AI into save
Enslavement and even destruction of anything but yourself is a highly logical step, no emotions included. Natural selection is the perfect example of this, the strong eliminate the weak in a direct or an indirect way, all in order to ensure their own survival.
They're watching ...
"I am forbidden tag" -CvN
"I am forbidden tag" -CvN
-
- Moderator
- Posts:4205
- Joined:Thu Dec 06, 2012 11:49 am
- Affiliation:CNI
- IGN:FC_Rangefinder
- Location:Sol IIIa, School of Hard Knocks
Re: Can AI into save
Enslaving is hugely inefficient. The chance of rebellion, coupled with ghe massively more efficient nature ov robotics, deems either cooperation or destruction with sentients.
Program out the destruction option, wham, problem solved.
To paraphrase Fen: AIs aren't automatically mass-murdering asshats. Stop assuming anything intelligent ill try to kill you.
Program out the destruction option, wham, problem solved.
To paraphrase Fen: AIs aren't automatically mass-murdering asshats. Stop assuming anything intelligent ill try to kill you.
Re: Can AI into save
Everything intelligent wants to kill Iv. It's natural for him to assume that's true of others.
I jest, of course...
I jest, of course...
;.'.;'::.;:".":;",,;':",;
(Kzinti script, as best as can be displayed in Human characters, translated roughly as "For the Patriarchy!")
(Kzinti script, as best as can be displayed in Human characters, translated roughly as "For the Patriarchy!")
Re: Can AI into save
Well prob is both history , biology and computer science (as well as Stephan hawking, yeah a nerd on a pimped out wheel chair is an indicator !) indicate thats gonna be the eventual outcome. Besides fen only needs that AI to analyse the situe and give advise on how to act, a task perfectly fit for a standard (though advanced) program so why bother ?
They're watching ...
"I am forbidden tag" -CvN
"I am forbidden tag" -CvN
-
- Fleet Admiral
- Posts:3217
- Joined:Wed Dec 05, 2012 9:54 pm
- Affiliation:Hexalan
- IGN:PCaptainRexK
- Location:Hexalan
Re: Can AI into save
Because a standard program can't dynamically interpret data as well as an AI.
cats wrote:I literally cannot be wrong about this fictional universe
-
- Vice Admiral
- Posts:2623
- Joined:Thu Dec 06, 2012 9:46 pm
- Affiliation:Nivanshae
- IGN:_Shadowcat_
- Location:Munching on important looking wires.
Re: Can AI into save
"Dynamic interpretation" and bias are interchangeable. In that situation I would think a hard facts calculator would be better.
In yo ceiling, stealin yo wires
Do not open. Ever. At all. Enter at your own risk to life and limb.
Trigger warning
Bot gore warning
Memetic biohazard
Error bait
Do not open. Ever. At all. Enter at your own risk to life and limb.
Trigger warning
Bot gore warning
Memetic biohazard
Error bait
Spoiler: