Here, evidently, politicians in control of the authority behind standards could corrupt the continuity of a good standard, but depending on how things were fundamentally arranged, the probabilities of serious damage through political corruption might becomes as small as the probabilities that the values of the standard meter and kilogram will be corrupted through the actions of politicians.
“Depending on how things were fundamentally arranged” is doing an implausible amount of work in that quote. Depending on how the atoms in my desk are fundamentally arranged, my desk may be a fully general nanoassembler, but the scientific and engineering work needed to turn my desk into a nanoassembler hasn’t been done.
Also, why do you think it is so unlikely that the value of the standard kilogram could be corrupted by the actions of politicians? It is improbable but not impossible. In fact, the definition of the “standard units” has evolved over time.
If
economists figure out a way of defining a standard portfolio of commodities that is inherently resilient against technological change, global disasters, market fluctuations, Soros-esque currency manipulation, or any other conceivable force that causes it to deviate from a perfect reflection of human value and
some entity implements this scheme in such a way that the standard is perfectly immune to institutional corruption, and
assuming the prior two (extremely difficult and maybe impossible) objectives have been accomplished, we further assume that, at the moment we turn on an AI, all knowledge potentially relevant to human value has already been in some sense encoded in this magical currency,
… then this might have a shot of working.
However, if there is any chance at all of the system that defines the currency being manipulated in any way, shape, or form, it’s pretty likely that a superintelligent AI will just manipulate the currency such that the most “economically correct” actions happen to coincide with the actions that are simplest and most certain, and in about three steps you have a paperclip maximizer.
“Depending on how things were fundamentally arranged” is doing an implausible amount of work in that quote. Depending on how the atoms in my desk are fundamentally arranged, my desk may be a fully general nanoassembler, but the scientific and engineering work needed to turn my desk into a nanoassembler hasn’t been done.
Firstly you are calling Nash dumb, which is stupid and ignorant. Do you think maybe you don’t understand what he is saying and especially WHY he is saying?
Also, why do you think it is so unlikely that the value of the standard kilogram could be corrupted by the actions of politicians? It is improbable but not impossible. In fact, the definition of the “standard units” has evolved over time.
Are you seriously addressing me as if I wrote Nash’s Ideal Money? Listen, Nash isn’t dumb, he explains how the standard units evolved over time and he is making fun of your for not considering a unit of value could do the same. He also explains how the standard unit of value needs to be adjusted over time.
economists figure out a way of defining a standard portfolio of commodities that is inherently resilient against technological change, global disasters, market fluctuations, Soros-esque currency manipulation, and
Nash perfectly defined it, don’t be silly.
some entity implements this scheme in such a way that the standard is perfectly immune to institutional corruption, and
Yes Ideal Money is immune (so is bitcoin I might add)
assuming the prior two (extremely difficult and maybe impossible) objectives have been accomplished, we further assume that, at the moment we turn on an AI, all knowledge potentially relevant to human value has already been in some sense encoded in this magical currency,
Your assumption AI won’t also evolve into existence is asinine.
… then this might have a shot of working.
Would you dare say that to Nash, that his idea MIGHT work? I don’t believe you, because no one had the guts to say when he was alive. You don’t look tough here, you look like a coward for this.
However, if there is any chance at all of the system that defines the currency being manipulated in any way, shape, or form, it’s pretty likely that a superintelligent AI will just manipulate the currency such that the most “economically correct” actions happen to coincide with the actions that are simplest and most certain, and in about three steps you have a paperclip maximizer.
No this is silly and irrational. If Nash showed that the levation of ideal money is the intelligent thing to do, then why would and AI corrupt the process considering its smarter than us? You are afraid of technological advance. Any intelligent life form, knows AI is the first goal of a level 1 civilization.
AI will just manipulate the currency such that the most “economically correct”
The currency itself defines what is economically correct. You are misunderstanding the nature of money and the problem it solves. The AI cannot do what you are saying because its not logical.
As general feed back: On Less Wrong we actually tend to have pretty thick skins, so it doesn’t bother me too much when you call me stupid. It only makes you look bad. Furthermore, we are uniquely unimpressed by appeals to authority around here. Isaac Newton did a tremendous amount of work in the occult. Despite Isaac Newton’s obviously superior intellect, I see no reason to drop what I’m doing and become an occultist. Likewise, I will defer to Nash on the particulars of abstract Game Theory, but see no strong reason to accept claims beyond his domain of expertise. I suggest you stop insulting people and stop ostentatiously worshiping John Nash, and I make this suggestion because it’s inhibiting your ability to communicate in general, not because it particularly effects me one way or the other.
You did not actually provide any arguments in your post except this line:
The currency itself defines what is economically correct. You are misunderstanding the nature of money and the problem it solves. The AI cannot do what you are saying because its not logical.
If the currency defines what is economically correct, and but does not perfectly capture human value, then human value will not be preserved when the AI tries to optimize for the parameter defined by the currency.
If there is any way of manipulating an objective function, an AI will try to do that in order to conserve resources while still maximizing its objectives. If there is any way of manipulating the currency—for example, hacking into whatever global database defines the value of the currency and literally changing the numbers—the AI will do that. This will almost always be easier for the AI, and the AI will have no inclination to follow the “spirit rather than the letter” of its programming.
As general feed back: On Less Wrong we actually tend to have pretty thick skins, so it doesn’t bother me too much when you call me stupid. It only makes you look bad.
Calling you stupid is a short cut. I know it doesn’t hurt you, and I don’t care if I look bad. I bring content that is truth.
Despite Isaac Newton’s obviously superior intellect, I see no reason to drop what I’m doing and become an occultist.
I know but the thing is that Newton was an alchemist too, and their famous endeavor is to effectively turn math into gold, which I am claiming has been achieved now. Also in Ideal Money Nash notes Newton pegged the pound to gold as/when he was master of the mint.
Likewise, I will defer to Nash on the particulars of abstract Game Theory, but see no strong reason to accept claims beyond his domain of expertise. I suggest you stop insulting people and stop ostentatiously worshiping John Nash, and I make this suggestion because it’s inhibiting your ability to communicate in general, not because it particularly effects me one way or the other.
People are insulting us, and I am not worshiping him just because I am pointing out his logic here is obviously infallible. And it is not intelligent to say you defer to him on game theory and not Ideal Money. His discovery was Ideal Money all along. The others proofs and insights are artifacts of him trying to express Ideal Money. Ideal Money is what he spent his whole life on. Its the big problem he solved and all his works is related and an expression of that.
If the currency defines what is economically correct, and but does not perfectly capture human value, then human value will not be preserved when the AI tries to optimize for the parameter defined by the currency.
I don’t know what you mean to say. In the future humans will get paid the most for contributing the most value to society (and at the same time pursing their own selfish interests). AI will do the same. The prices will tell us what to do. AI won’t corrupt that because its obviously optimal and it is smarter than us.
If there is any way of manipulating an objective function, an AI will try to do that in order to conserve resources while still maximizing its objectives. If there is any way of manipulating the currency—for example, hacking into whatever global database defines the value of the currency and literally changing the numbers—the AI will do that. This will almost always be easier for the AI, and the AI will have no inclination to follow the “spirit rather than the letter” of its programming.
No you cannot manipulate ideal money because its conceptual. You are being silly. Also when Nash was a kid he sent a letter to the NSA explaining a conjecture for unbreakeable encryption. AI can’t hack bitcoin. The encryptor always wins.
You are conflating at least four senses of the word “value” into one concept and then assuming that those four things are the same thing.
What we call “human value” is the actual, abstract, difficult-to-know set of things that human beings want, and the things that we don’t know that we want but we would want if we knew about them, and maybe even the things that we wouldn’t want but that we might be glad we had after we got them. Discovering the content of “human value” is an enormous, unsolved philosophical problem.
What you might call “economic value” is purely defined in terms of trades between agents and can only be grounded in relative comparison between commodities.
In your post you throw in a reference to “value to society”. Something might be very valuable to society but individuals fail to coordinate and pay for that thing. This is known as the tragedy of the commons.
Individual humans are known to value weird, idiosyncratic things. You might call these “individual values”. Perhaps it would be valuable to society for some person to be put to death, but this would be contrary to that person’s individual values.
“Value to society” is not identical to “value to an individual” is not identical to “economic value” is not identical to “human value”, and none of these are identical to “Ideal Currency value”. There is certainly no apparent reason why Ideal Currency would converge on “human value”, which is the one that I would say should be optimized. Swapping these various senses of the word in your rhetoric as though they are interchangeable is leading you to make large errors. Assuming that an AI will just get what you mean and know instinctively which definition you mean is crippling to your thesis.
“Depending on how things were fundamentally arranged” is doing an implausible amount of work in that quote. Depending on how the atoms in my desk are fundamentally arranged, my desk may be a fully general nanoassembler, but the scientific and engineering work needed to turn my desk into a nanoassembler hasn’t been done.
Also, why do you think it is so unlikely that the value of the standard kilogram could be corrupted by the actions of politicians? It is improbable but not impossible. In fact, the definition of the “standard units” has evolved over time.
If
economists figure out a way of defining a standard portfolio of commodities that is inherently resilient against technological change, global disasters, market fluctuations, Soros-esque currency manipulation, or any other conceivable force that causes it to deviate from a perfect reflection of human value and
some entity implements this scheme in such a way that the standard is perfectly immune to institutional corruption, and
assuming the prior two (extremely difficult and maybe impossible) objectives have been accomplished, we further assume that, at the moment we turn on an AI, all knowledge potentially relevant to human value has already been in some sense encoded in this magical currency,
… then this might have a shot of working.
However, if there is any chance at all of the system that defines the currency being manipulated in any way, shape, or form, it’s pretty likely that a superintelligent AI will just manipulate the currency such that the most “economically correct” actions happen to coincide with the actions that are simplest and most certain, and in about three steps you have a paperclip maximizer.
Firstly you are calling Nash dumb, which is stupid and ignorant. Do you think maybe you don’t understand what he is saying and especially WHY he is saying?
2ndly you are not using the standard accepted definition of the word Ideal: http://lesswrong.com/r/discussion/lw/ogt/do_we_share_a_defintion_for_the_word_ideal/
Are you seriously addressing me as if I wrote Nash’s Ideal Money? Listen, Nash isn’t dumb, he explains how the standard units evolved over time and he is making fun of your for not considering a unit of value could do the same. He also explains how the standard unit of value needs to be adjusted over time.
Nash perfectly defined it, don’t be silly.
Yes Ideal Money is immune (so is bitcoin I might add)
Your assumption AI won’t also evolve into existence is asinine.
Would you dare say that to Nash, that his idea MIGHT work? I don’t believe you, because no one had the guts to say when he was alive. You don’t look tough here, you look like a coward for this.
No this is silly and irrational. If Nash showed that the levation of ideal money is the intelligent thing to do, then why would and AI corrupt the process considering its smarter than us? You are afraid of technological advance. Any intelligent life form, knows AI is the first goal of a level 1 civilization.
The currency itself defines what is economically correct. You are misunderstanding the nature of money and the problem it solves. The AI cannot do what you are saying because its not logical.
As general feed back: On Less Wrong we actually tend to have pretty thick skins, so it doesn’t bother me too much when you call me stupid. It only makes you look bad. Furthermore, we are uniquely unimpressed by appeals to authority around here. Isaac Newton did a tremendous amount of work in the occult. Despite Isaac Newton’s obviously superior intellect, I see no reason to drop what I’m doing and become an occultist. Likewise, I will defer to Nash on the particulars of abstract Game Theory, but see no strong reason to accept claims beyond his domain of expertise. I suggest you stop insulting people and stop ostentatiously worshiping John Nash, and I make this suggestion because it’s inhibiting your ability to communicate in general, not because it particularly effects me one way or the other.
You did not actually provide any arguments in your post except this line:
If the currency defines what is economically correct, and but does not perfectly capture human value, then human value will not be preserved when the AI tries to optimize for the parameter defined by the currency.
If there is any way of manipulating an objective function, an AI will try to do that in order to conserve resources while still maximizing its objectives. If there is any way of manipulating the currency—for example, hacking into whatever global database defines the value of the currency and literally changing the numbers—the AI will do that. This will almost always be easier for the AI, and the AI will have no inclination to follow the “spirit rather than the letter” of its programming.
Calling you stupid is a short cut. I know it doesn’t hurt you, and I don’t care if I look bad. I bring content that is truth.
I know but the thing is that Newton was an alchemist too, and their famous endeavor is to effectively turn math into gold, which I am claiming has been achieved now. Also in Ideal Money Nash notes Newton pegged the pound to gold as/when he was master of the mint.
People are insulting us, and I am not worshiping him just because I am pointing out his logic here is obviously infallible. And it is not intelligent to say you defer to him on game theory and not Ideal Money. His discovery was Ideal Money all along. The others proofs and insights are artifacts of him trying to express Ideal Money. Ideal Money is what he spent his whole life on. Its the big problem he solved and all his works is related and an expression of that.
I don’t know what you mean to say. In the future humans will get paid the most for contributing the most value to society (and at the same time pursing their own selfish interests). AI will do the same. The prices will tell us what to do. AI won’t corrupt that because its obviously optimal and it is smarter than us.
No you cannot manipulate ideal money because its conceptual. You are being silly. Also when Nash was a kid he sent a letter to the NSA explaining a conjecture for unbreakeable encryption. AI can’t hack bitcoin. The encryptor always wins.
You are conflating at least four senses of the word “value” into one concept and then assuming that those four things are the same thing.
What we call “human value” is the actual, abstract, difficult-to-know set of things that human beings want, and the things that we don’t know that we want but we would want if we knew about them, and maybe even the things that we wouldn’t want but that we might be glad we had after we got them. Discovering the content of “human value” is an enormous, unsolved philosophical problem.
What you might call “economic value” is purely defined in terms of trades between agents and can only be grounded in relative comparison between commodities.
In your post you throw in a reference to “value to society”. Something might be very valuable to society but individuals fail to coordinate and pay for that thing. This is known as the tragedy of the commons.
Individual humans are known to value weird, idiosyncratic things. You might call these “individual values”. Perhaps it would be valuable to society for some person to be put to death, but this would be contrary to that person’s individual values.
“Value to society” is not identical to “value to an individual” is not identical to “economic value” is not identical to “human value”, and none of these are identical to “Ideal Currency value”. There is certainly no apparent reason why Ideal Currency would converge on “human value”, which is the one that I would say should be optimized. Swapping these various senses of the word in your rhetoric as though they are interchangeable is leading you to make large errors. Assuming that an AI will just get what you mean and know instinctively which definition you mean is crippling to your thesis.