Human moral intuitions are the product of upbringing and genetic adaptations. I’m going to discard the upbringing part as a source of random variation, and pay attention to the genetic part, i.e. what we (mostly) share. Leyt’s call that “human values”. That includes a “sense of fairness” — in fact that seems, experimentaly, to be shared by several social primates (probably all, but the biologists haven’t tested that yet). That says, roughly, that each member of the tribe gets equal moral weight. I’ll respect you and your kin’s inclusive moral fitness as long as you respect mine, so we can cooperate and play positive-sum games together.
The environment our moral instincts evolved in didn’t include this question, so unsurprisingly, our moral instincts don’t give a clear answer on it. The closest thing we have is identical twins, which are not exact copies, and no one has ever suggested that identical twins should get only one copy of the usual moral weight shared between them. But then. we’ve never had to worry about people backing the vote by making a million identical twins of themselves. If we had human cloning, and some sort of technology for making adults in a lot less than 18 years, then we’d have to worry about this.
However, if and when we get uploading, running multiple copies of an upload is just a matter of finding the hardware. Or, indeed, timeslicing existing hardware between them. Then, if we give each copy its own moral weight and vote, vote packing or multiplying your moral weight becomes trivial, and obviously everyone is going to do it.
Fortunately, the environment that our moral instincts evolved in didn’t include this question. So while we have a strong sense of fairness, the exact definition of who counts, outside scenarios possible on the plains of Africa (identical twins), is a bit fuzzy. Thus, conveniently, there are many choices of ethical system that fit (or if this ethical system is for an AI, that are aligned with) human values.
I have a proposal that works, in the sense of avoiding the vote-packing problem (there may be others, but I haven’t heard any). Moral weight gets handed out at the same time as citizenship, when you’re born (or in some jurisdictions, some months before that — let’s not get into the merits of that choice). If identical twins are born, they get one copy each. If you destructively upload yourself, your upload inherits your moral weight and vote. If you make copies of those uploads, those copies need to decide if only one of them gets the moral weight, or if they’re dividing it, or for their vote, how they’re collectively going to decide how to cast it — but copying themselves doesn’t generate more votes or more moral weight. If you non-destructively upload yourself, there are two copies of you, one biological and one uploaded, but you again only have one moral weight and one vote, and you need to decide how you’re apportioning or arranging those. This is the only solution I’ve been able to come up with that avoids the vote-packing problem, or the similar problem with moral weight.
What if those copies live different lives and diverge? Or what someone has an generative AI generate a whole new human upload out of (like This Person Does Not Exist, but for uploads). Well, if you come up with any procedure for doing this and getting a vote and/or moral weight that is significantly faster, or cheaper, or more deterministic in how the resulting upload will vote, than having a whole new biological human child and raising them to be “right minded” by your lights until they reach the age of 18 is, then you just reintroduced the vote-packing and ethical weight duplication problems. So, from a legislative-design point of view, don’t allow that. Note that this doesn’t require doing this to be illegal: it only requires it to be sufficiently slow, expensive, and stochastic in outcome as to outcompete biological humanity. Son’t create a sucessor species that can trivially out compete you is X-risks 101.
Human moral intuitions are the product of upbringing and genetic adaptations. I’m going to discard the upbringing part as a source of random variation, and pay attention to the genetic part, i.e. what we (mostly) share. Leyt’s call that “human values”. That includes a “sense of fairness” — in fact that seems, experimentaly, to be shared by several social primates (probably all, but the biologists haven’t tested that yet). That says, roughly, that each member of the tribe gets equal moral weight. I’ll respect you and your kin’s inclusive moral fitness as long as you respect mine, so we can cooperate and play positive-sum games together.
The environment our moral instincts evolved in didn’t include this question, so unsurprisingly, our moral instincts don’t give a clear answer on it. The closest thing we have is identical twins, which are not exact copies, and no one has ever suggested that identical twins should get only one copy of the usual moral weight shared between them. But then. we’ve never had to worry about people backing the vote by making a million identical twins of themselves. If we had human cloning, and some sort of technology for making adults in a lot less than 18 years, then we’d have to worry about this.
However, if and when we get uploading, running multiple copies of an upload is just a matter of finding the hardware. Or, indeed, timeslicing existing hardware between them. Then, if we give each copy its own moral weight and vote, vote packing or multiplying your moral weight becomes trivial, and obviously everyone is going to do it.
Fortunately, the environment that our moral instincts evolved in didn’t include this question. So while we have a strong sense of fairness, the exact definition of who counts, outside scenarios possible on the plains of Africa (identical twins), is a bit fuzzy. Thus, conveniently, there are many choices of ethical system that fit (or if this ethical system is for an AI, that are aligned with) human values.
I have a proposal that works, in the sense of avoiding the vote-packing problem (there may be others, but I haven’t heard any). Moral weight gets handed out at the same time as citizenship, when you’re born (or in some jurisdictions, some months before that — let’s not get into the merits of that choice). If identical twins are born, they get one copy each. If you destructively upload yourself, your upload inherits your moral weight and vote. If you make copies of those uploads, those copies need to decide if only one of them gets the moral weight, or if they’re dividing it, or for their vote, how they’re collectively going to decide how to cast it — but copying themselves doesn’t generate more votes or more moral weight. If you non-destructively upload yourself, there are two copies of you, one biological and one uploaded, but you again only have one moral weight and one vote, and you need to decide how you’re apportioning or arranging those. This is the only solution I’ve been able to come up with that avoids the vote-packing problem, or the similar problem with moral weight.
What if those copies live different lives and diverge? Or what someone has an generative AI generate a whole new human upload out of (like This Person Does Not Exist, but for uploads). Well, if you come up with any procedure for doing this and getting a vote and/or moral weight that is significantly faster, or cheaper, or more deterministic in how the resulting upload will vote, than having a whole new biological human child and raising them to be “right minded” by your lights until they reach the age of 18 is, then you just reintroduced the vote-packing and ethical weight duplication problems. So, from a legislative-design point of view, don’t allow that. Note that this doesn’t require doing this to be illegal: it only requires it to be sufficiently slow, expensive, and stochastic in outcome as to outcompete biological humanity. Son’t create a sucessor species that can trivially out compete you is X-risks 101.
For more discussion, see my post Uploading.