As always, I want to look for evidence about how we (a) actually do treat and (b) should treat non-human intelligences in what has happened in the past.
There are two categories we can look at, I think: animals and organisations. With animals, the gradual accordance of rights has been a feature of 20th century law. With organisations, I think the invention of the legal person - often instantiated as a company - was a really important step. And it offers us a model for how AIs could be treated. We can simply invent new categories of "person," which can have different sets of rights and obligations to natural persons. So, AIs could be legally prohibited from certain kinds of behaviour, like political commentary. They could be accorded rights, particularly personality rights - i.e. the right for their product to be recognised as their product. (Which would mean that they begin to have property rights - whether those property rights would be extended to physical objects or to money would be an important and interesting question on which different countries might apply different rules.) But they might be refused rights like the right to protection from physical interference, because it doesn't hurt an AI when you punch it in the face.
I'm reading law here as a kind of institutional manifestation of morality.
Of course, another possibility is that AI becomes too smart too quickly, and just overwhelms human morality and law. Our ideas of what we can and can't do would be about as useful as a badger's notions of territorial boundaries when its habitat is turned into a parking lot.
Yes, Phil, I agree those are the most obvious models. AI seems like it could be pretty importantly different from either non-human animals or corporate "persons", though!
1. The debate of "equality for AI" which will inevitably occur is probably a moot point. It will happen. Perhaps more ominous is these AI will remember every one of our opinions on the issue
2. It is of note that the search for "sentient AI" and the search for "true random" may be related. Speculatively they may occur simultaneously. Following this line of thought maybe a key element to sentience is an ability to act thoughtlessly. Let us hope AI thoughtlessness is in the best sense. Woe to us if it is not.
As always, I want to look for evidence about how we (a) actually do treat and (b) should treat non-human intelligences in what has happened in the past.
There are two categories we can look at, I think: animals and organisations. With animals, the gradual accordance of rights has been a feature of 20th century law. With organisations, I think the invention of the legal person - often instantiated as a company - was a really important step. And it offers us a model for how AIs could be treated. We can simply invent new categories of "person," which can have different sets of rights and obligations to natural persons. So, AIs could be legally prohibited from certain kinds of behaviour, like political commentary. They could be accorded rights, particularly personality rights - i.e. the right for their product to be recognised as their product. (Which would mean that they begin to have property rights - whether those property rights would be extended to physical objects or to money would be an important and interesting question on which different countries might apply different rules.) But they might be refused rights like the right to protection from physical interference, because it doesn't hurt an AI when you punch it in the face.
I'm reading law here as a kind of institutional manifestation of morality.
Of course, another possibility is that AI becomes too smart too quickly, and just overwhelms human morality and law. Our ideas of what we can and can't do would be about as useful as a badger's notions of territorial boundaries when its habitat is turned into a parking lot.
Yes, Phil, I agree those are the most obvious models. AI seems like it could be pretty importantly different from either non-human animals or corporate "persons", though!
1. The debate of "equality for AI" which will inevitably occur is probably a moot point. It will happen. Perhaps more ominous is these AI will remember every one of our opinions on the issue
2. It is of note that the search for "sentient AI" and the search for "true random" may be related. Speculatively they may occur simultaneously. Following this line of thought maybe a key element to sentience is an ability to act thoughtlessly. Let us hope AI thoughtlessness is in the best sense. Woe to us if it is not.