Friday, June 11, 2004
asimov@orkut
A new episode from asimov@orkut,...
Suggestion from Night [
But my way to rule robots as persons or not would be this:
Make a series of double blind experiments, where real people and robots interact with each other, maybe in a chat like environment, with the instructions to not have anything pertinent to the three laws (if they are implemented) discussed. If it is impossible for either robots or people to distinguish between themselves in such experiments, then you can safely say it would be adequate to consider robots people too. Also, if there are some perceived differences, maybe robots can have a different set of rights, but still more than mere "machines".(A turing test)]
Now my gripe is this..
Any Turing test that is applied suffers from two flaws, one is that they rely on reducing the inputs available (In night's case the three laws and blind folding) and secondly the amount of time permitted for conducting the test.
To make the argument clear, let us consider the same argument in a more restrictive way, Suppose that you are faced with an entity whom you are supposed to test, and the method of testing allows moving the pieces of a chess board, How long would you need to determine if the entity that you are facing is Human or not? The difference between this and the other one is just in degrees (with the corresponding restrictions - no testing on three laws, no question beyond a limited database, etc). The moment you want to remove this flaw by loosening the restrictions, time required for a decision increases exponentially, This is the reason why we need the meta rules to judge.
On Nights second opinion,
[Capacity for distinguishing "good" from "wrong". It's clear from our experience with the legal criminal system that some people don't know the difference. Those are usually declared "insane" and put aside society, receiving a different set of rights and duties.
Would not be, then, coherent to give robots rights based on that scale to? If a robot is self-aware, can conduct conversations and think the same way or better than a human being, while being able to distinguish "good" from "bad", it is perfectly clear to me "he" would have all the right to be called a person. Even more than some humans.]
Judging good and bad aside from simple ones are pretty difficult, and is dependent on the social environment prevalent at that time, And I would say that distinguishing good and bad is not a requirement for the Being in question,
As an example, in today's world, Can you say whether George Bush's war on Iraq was good/just ? what does Bush think, what does the Iraqies think, and what do you think? ¶ 4:19 PM
Suggestion from Night [
But my way to rule robots as persons or not would be this:
Make a series of double blind experiments, where real people and robots interact with each other, maybe in a chat like environment, with the instructions to not have anything pertinent to the three laws (if they are implemented) discussed. If it is impossible for either robots or people to distinguish between themselves in such experiments, then you can safely say it would be adequate to consider robots people too. Also, if there are some perceived differences, maybe robots can have a different set of rights, but still more than mere "machines".(A turing test)]
Now my gripe is this..
Any Turing test that is applied suffers from two flaws, one is that they rely on reducing the inputs available (In night's case the three laws and blind folding) and secondly the amount of time permitted for conducting the test.
To make the argument clear, let us consider the same argument in a more restrictive way, Suppose that you are faced with an entity whom you are supposed to test, and the method of testing allows moving the pieces of a chess board, How long would you need to determine if the entity that you are facing is Human or not? The difference between this and the other one is just in degrees (with the corresponding restrictions - no testing on three laws, no question beyond a limited database, etc). The moment you want to remove this flaw by loosening the restrictions, time required for a decision increases exponentially, This is the reason why we need the meta rules to judge.
On Nights second opinion,
[Capacity for distinguishing "good" from "wrong". It's clear from our experience with the legal criminal system that some people don't know the difference. Those are usually declared "insane" and put aside society, receiving a different set of rights and duties.
Would not be, then, coherent to give robots rights based on that scale to? If a robot is self-aware, can conduct conversations and think the same way or better than a human being, while being able to distinguish "good" from "bad", it is perfectly clear to me "he" would have all the right to be called a person. Even more than some humans.]
Judging good and bad aside from simple ones are pretty difficult, and is dependent on the social environment prevalent at that time, And I would say that distinguishing good and bad is not a requirement for the Being in question,
As an example, in today's world, Can you say whether George Bush's war on Iraq was good/just ? what does Bush think, what does the Iraqies think, and what do you think? ¶ 4:19 PM
