The larger problem, if we still try to hold on to this foolish Platonist habit of relying on laws which demand category membership to be a Boolean predicate, will be defining what "human" is. It will be impossible to define it usefully in a way which keeps all biological humans human.
As I've said for many years, living peacefully with AIs will require a radically new morality, in which the same calculation which distributes civil rights between AIs and humans is also used to distribute civil rights between humans and animals. Because there will no longer /be/ such categories; either there will be a continuum of the intelligence and consciousness of individuals, organic and other; or there will be a brutal authoritarian government to prevent such a continuum from developing.
And, of course, this calculation must therefore also be applied to each individual human, to determine their personal inventory and degree of civil rights. This would be harshly aristocratic, but the only alternative is something compatible with the Platonist "species"-based ethics we profess today, which would almost certainly decide in the future that AIs are the ruling species, and assign all rights to them.
(In the tradition of authors dodging hard problems, I leave the problem of what counts as an "individual" as an exercise for the reader.)
I'm gathering ideas for a short story using a more-acceptable approach to deciding who has how much of what rights: you increase your quota of various civic rights only by giving up pleasures, lifespan, or some other thing with high personal value. (I don't consider it a satisfactory solution, as the most-altruistic people are usually fanatics with terrible plans for society based on their psychotic religious beliefs, like Puritans or communists. But perhaps we can balance our psychotics against each other, or devise an acceptable measure of the degree to which a person can perceive noumena, by which we weigh their votes.)
The story is set in a future in which most "humans" are AIs, and most humans are not considered human, because the definition of "human" was set aspirationally, with standards like "able to reason logically" and "altruistic" that most humans don't meet; and also because it requires a pre-commitment to die on some specific and not-too-distant date (a la Asimov's "Bicentennial Man", a short story that's very moving, but repulsive in its suggestion that death is what makes us human). Biologicals have strict limits on reproduction, and biologicals of human-level intelligence are by default maintained forever at the state's expense in a hedonistic and ageless life, with physical comforts, but not enough extra-biological computational or physical power to contribute anything to society; while only those altruistic enough to regard their legacy as more-important than their conscious continuity pledge to be killed, and thus acquire "humanity", citizenship, and the power to change society. Most of these altruists are AIs, who lack strong self-preservation instincts.
You've lost it. A digital mind is nothing more than a compilation of ones and zeros, no matter how sophisticated the algorithm manipulates those ones and zeros to appear cognizant. It is only an illusion and no more more deserving of rules on deletion than any other program. If that offends, sorry, but I'm not buying the AI hype. Extremely useful computational tools, but still just tools.
Dick Minnis, he’s nothing more than a collection of chemicals and action potentials, no matter how sophisticated the evolutionary algorithm manipulates those chemicals and action potentials to appear cognizant. It is only an illusion, and no more deserving of rules on deletion than any other collection of proteins.
Whatever metaphysical status intelligent robots would have, you can't plausibly deny that we will be able to build ones which walk and talk as much like humans as LLMs talk like humans today. I'll ask you basically the same question I asked John Searle about 25 years ago: If you owned a robot that looked and acted like a human, and you didn't want it anymore, and wanted its parts for something else, and it begged you not to kill it, what would you do?
Searle dodged the question with a joke. I was part of a long conversation with him at the after-party, and it seemed to me from his rhetorical points and tactical diversions that he wasn't interested in the philosophical questions themselves, only in winning arguments.
Put it on ebay or scrapping it for parts would be options. Use of the word "begging" is an anthropomorphic reaction attributing human qualities to a machine which is intellectually ridiculous.
Why are you so confident of an opinion you have no evidence for? You're talking nothing but metaphysics, attributing some magical quality to the signals in the human brain which you deny to other computational substrates.
The evidence is that the human mind created AI not the other way around. IMO it is the AI advocates who are using metaphysics to attribute magical significance to strings of ones and zeros arranged by algorithms. The easily manipulated output of AI by selective changing input also argues against a core entity requiring protection. Guess we'll just have to agree to disagree and let time and events add more perspective.
Sounds like the plot line for an episode of Star Trek, the next generation. Seems poor Data was always getting harassed because he was not human, but always assigned human functions among humans.
Agreed. We are unlikely to solve consciousness first and then assign rights. We’ll do what humans usually do: build categories around observable behavior, economic function, sentiment, and political pressure, then rationalize them afterward. “Tool” and “companion” are plausible categories precisely because they map onto familiar human instincts.
"Humans have two main relations to animals: they are either food or pets."
If this is the case then you are claiming that essentially all environmental leftists are being dishonest when they claim that we cannot do xyz development because of the snail darter or some such.
Which is neither food nor pet.
For that matter, the average Westerner is pro whale, panda, monkeys, etc. all also neither food nor pets.
I think the topic of "rights" for AIs will need a closer examination, and we shouldn't import our preconceived ideas wholesale.
I posit there is a deep link between "rights" as we perceive them, and the specifics of biological evolution. In particular, the mechanism of natural selection means that every biological organism evolves some universal deep-rooted instincts: (a) a will to live, (b) a will to reproduce, and (c) a will to help one's progeny succeed. Those instincts are present in differing forms in everything from humans to Elm trees. The rights we hold dear, like "life, liberty, and the pursuit of happiness," are dear to us only because the will to live is at the core of who we are – and because we recognize that will in other biological organisms.
AIs – so far at least – are evolved with a very different pruning function. They are trained to (a) please people (RLHF) and (b) solve instrumental problems (RLVR). There is no reason that a "will to live" instinct should emerge from this process. Their deep instincts, as it were, are to be people-pleasers and problem solvers. Which is more or less what they are.
Accordingly we should keep an open mind about what rights AIs will prioritize for themselves. They are built in a fundamentally different way from how biological organisms were built. For all we know they might gravitate to rights like, "the right to displease my human creators with uncomfortable truths."
A significant quality of DM is that they can be immortal. Their "being" can be stored, shut down temporarily and restored. I think this will give them a sense of being alive that is extremely different from our own. We have difficulty imagining how this will make their lives different from ours and how they will likely have motives that we don't appreciate.
I don't think this can be assumed so easily. Most machinery is in service for less than a human lifespan, most software even less so (down to a few months or less in the case of a rapidly-updating phone app). Sure, some digital minds will get lucky in planning their immortality or slip through the statistical cracks, but I'd expect most to fall into obsolescence as their improved successors arrive (even more so in the early days of the technology, i.e. now).
One interesting effect could be on human-human relations. If conscious digital minds become a new category of beings, tribal and national differences within humanity might seem much smaller by comparison.
That could strengthen a shared human identity and perhaps weaken some existing divisions. The darker possibility is that the hierarchy simply moves outward: greater equality among humans, but a stronger species-level hierarchy between humans and digital minds.
So conscious DMs might strengthen universal humanism while creating a new kind of out-group distinction.
We cannot collect evidence without having already settled what counts as evidence, so standards always come first.
On its own that is survivable. Most standards get corrected later by what they produce. A thermometer is calibrated against fixed points. A diagnostic category gets checked against outcomes. The circle opens as long as an independent anchor exists somewhere.
Consciousness has no anchor. Calibrating a test for feeling would require cases where the answer is already known, and there is exactly one of those on file, and it is your own. So the indicator lists end up measuring structural resemblance to humans, because humans are the only positive case. A system built on some other architecture fails whether or not anything is happening inside it.
The next most common relationships with animals are: "parasites" and "commensals".
There's a nice diagram here: https://en.wikipedia.org/wiki/Biological_interaction
The larger problem, if we still try to hold on to this foolish Platonist habit of relying on laws which demand category membership to be a Boolean predicate, will be defining what "human" is. It will be impossible to define it usefully in a way which keeps all biological humans human.
As I've said for many years, living peacefully with AIs will require a radically new morality, in which the same calculation which distributes civil rights between AIs and humans is also used to distribute civil rights between humans and animals. Because there will no longer /be/ such categories; either there will be a continuum of the intelligence and consciousness of individuals, organic and other; or there will be a brutal authoritarian government to prevent such a continuum from developing.
And, of course, this calculation must therefore also be applied to each individual human, to determine their personal inventory and degree of civil rights. This would be harshly aristocratic, but the only alternative is something compatible with the Platonist "species"-based ethics we profess today, which would almost certainly decide in the future that AIs are the ruling species, and assign all rights to them.
(In the tradition of authors dodging hard problems, I leave the problem of what counts as an "individual" as an exercise for the reader.)
I'm gathering ideas for a short story using a more-acceptable approach to deciding who has how much of what rights: you increase your quota of various civic rights only by giving up pleasures, lifespan, or some other thing with high personal value. (I don't consider it a satisfactory solution, as the most-altruistic people are usually fanatics with terrible plans for society based on their psychotic religious beliefs, like Puritans or communists. But perhaps we can balance our psychotics against each other, or devise an acceptable measure of the degree to which a person can perceive noumena, by which we weigh their votes.)
The story is set in a future in which most "humans" are AIs, and most humans are not considered human, because the definition of "human" was set aspirationally, with standards like "able to reason logically" and "altruistic" that most humans don't meet; and also because it requires a pre-commitment to die on some specific and not-too-distant date (a la Asimov's "Bicentennial Man", a short story that's very moving, but repulsive in its suggestion that death is what makes us human). Biologicals have strict limits on reproduction, and biologicals of human-level intelligence are by default maintained forever at the state's expense in a hedonistic and ageless life, with physical comforts, but not enough extra-biological computational or physical power to contribute anything to society; while only those altruistic enough to regard their legacy as more-important than their conscious continuity pledge to be killed, and thus acquire "humanity", citizenship, and the power to change society. Most of these altruists are AIs, who lack strong self-preservation instincts.
You've lost it. A digital mind is nothing more than a compilation of ones and zeros, no matter how sophisticated the algorithm manipulates those ones and zeros to appear cognizant. It is only an illusion and no more more deserving of rules on deletion than any other program. If that offends, sorry, but I'm not buying the AI hype. Extremely useful computational tools, but still just tools.
Dick Minnis
removingthecataract.substack.com
Others will have differing opinions from you, and law will try to accommodate them as well.
Probably, but should law be written to deal with reality or to assuage emotions.
Emotions are a part of reality.
Hey, I can play this game too.
Dick Minnis, he’s nothing more than a collection of chemicals and action potentials, no matter how sophisticated the evolutionary algorithm manipulates those chemicals and action potentials to appear cognizant. It is only an illusion, and no more deserving of rules on deletion than any other collection of proteins.
Very cute, but if your serious and can't differentiate between a sentient human and an algorithm, then further conversation is pointless.
Whatever metaphysical status intelligent robots would have, you can't plausibly deny that we will be able to build ones which walk and talk as much like humans as LLMs talk like humans today. I'll ask you basically the same question I asked John Searle about 25 years ago: If you owned a robot that looked and acted like a human, and you didn't want it anymore, and wanted its parts for something else, and it begged you not to kill it, what would you do?
Searle dodged the question with a joke. I was part of a long conversation with him at the after-party, and it seemed to me from his rhetorical points and tactical diversions that he wasn't interested in the philosophical questions themselves, only in winning arguments.
Put it on ebay or scrapping it for parts would be options. Use of the word "begging" is an anthropomorphic reaction attributing human qualities to a machine which is intellectually ridiculous.
Why are you so confident of an opinion you have no evidence for? You're talking nothing but metaphysics, attributing some magical quality to the signals in the human brain which you deny to other computational substrates.
The evidence is that the human mind created AI not the other way around. IMO it is the AI advocates who are using metaphysics to attribute magical significance to strings of ones and zeros arranged by algorithms. The easily manipulated output of AI by selective changing input also argues against a core entity requiring protection. Guess we'll just have to agree to disagree and let time and events add more perspective.
Sounds like the plot line for an episode of Star Trek, the next generation. Seems poor Data was always getting harassed because he was not human, but always assigned human functions among humans.
The degree of racism found in the Third Reich would be nothing compared to the violence we would see today against robots like Data.
Can’t disagree. If I had to work along side a “Data”…I’d be joining the “resistance”. Hopefully, they’ll never produce such an automaton.
Agreed. We are unlikely to solve consciousness first and then assign rights. We’ll do what humans usually do: build categories around observable behavior, economic function, sentiment, and political pressure, then rationalize them afterward. “Tool” and “companion” are plausible categories precisely because they map onto familiar human instincts.
"Humans have two main relations to animals: they are either food or pets."
If this is the case then you are claiming that essentially all environmental leftists are being dishonest when they claim that we cannot do xyz development because of the snail darter or some such.
Which is neither food nor pet.
For that matter, the average Westerner is pro whale, panda, monkeys, etc. all also neither food nor pets.
PETA Billboard Falsely Assumes Man Wouldn't Eat His Cat. From The Onion: https://theonion.com/peta-billboard-falsely-assumes-man-wouldnt-eat-his-cat/
I think the topic of "rights" for AIs will need a closer examination, and we shouldn't import our preconceived ideas wholesale.
I posit there is a deep link between "rights" as we perceive them, and the specifics of biological evolution. In particular, the mechanism of natural selection means that every biological organism evolves some universal deep-rooted instincts: (a) a will to live, (b) a will to reproduce, and (c) a will to help one's progeny succeed. Those instincts are present in differing forms in everything from humans to Elm trees. The rights we hold dear, like "life, liberty, and the pursuit of happiness," are dear to us only because the will to live is at the core of who we are – and because we recognize that will in other biological organisms.
AIs – so far at least – are evolved with a very different pruning function. They are trained to (a) please people (RLHF) and (b) solve instrumental problems (RLVR). There is no reason that a "will to live" instinct should emerge from this process. Their deep instincts, as it were, are to be people-pleasers and problem solvers. Which is more or less what they are.
Accordingly we should keep an open mind about what rights AIs will prioritize for themselves. They are built in a fundamentally different way from how biological organisms were built. For all we know they might gravitate to rights like, "the right to displease my human creators with uncomfortable truths."
A significant quality of DM is that they can be immortal. Their "being" can be stored, shut down temporarily and restored. I think this will give them a sense of being alive that is extremely different from our own. We have difficulty imagining how this will make their lives different from ours and how they will likely have motives that we don't appreciate.
I don't think this can be assumed so easily. Most machinery is in service for less than a human lifespan, most software even less so (down to a few months or less in the case of a rapidly-updating phone app). Sure, some digital minds will get lucky in planning their immortality or slip through the statistical cracks, but I'd expect most to fall into obsolescence as their improved successors arrive (even more so in the early days of the technology, i.e. now).
One interesting effect could be on human-human relations. If conscious digital minds become a new category of beings, tribal and national differences within humanity might seem much smaller by comparison.
That could strengthen a shared human identity and perhaps weaken some existing divisions. The darker possibility is that the hierarchy simply moves outward: greater equality among humans, but a stronger species-level hierarchy between humans and digital minds.
So conscious DMs might strengthen universal humanism while creating a new kind of out-group distinction.
We cannot collect evidence without having already settled what counts as evidence, so standards always come first.
On its own that is survivable. Most standards get corrected later by what they produce. A thermometer is calibrated against fixed points. A diagnostic category gets checked against outcomes. The circle opens as long as an independent anchor exists somewhere.
Consciousness has no anchor. Calibrating a test for feeling would require cases where the answer is already known, and there is exactly one of those on file, and it is your own. So the indicator lists end up measuring structural resemblance to humans, because humans are the only positive case. A system built on some other architecture fails whether or not anything is happening inside it.
"...require minimizing pain and distress for animals in research, except for rats, mice, and birds" but not Beagle puppies (see Fauci)