Most people imagine their past and future selves as links in a continuous chain, each one connected to the previous one and the next one. It seems to me that the structure might be more like a singly linked list: each moment has a “pointer” to the previous one, by virtue of remembering it, but there’s no corresponding pointer to the next moment, and in fact there’s no unique next moment, because we are branching all the time.
(In functional programming languages, different linked lists can “share structure” by having pointers to the same tail. That idea also works for other data structures like trees, as described in Chris Okasaki’s book “Purely functional data structures”. That’s a digression, but a really interesting one.)
So why do we feel that we have a pointer to the next moment, or a probability distribution over next moments? I think it’s because we learn by induction. Looking at our memory, we see that each moment had a previous one and a next one, so we think the same must be true for the present moment as well. But it seems to me that the whole picture can be understood by just looking at links to previous moments, and nature doesn’t need to have any special laws about links to next moments, “observer fluid” and the like.
One of the benefits of such a “memory identity theory”, compared to “pattern identity theory”, is that it makes it easier to imagine merging two creatures. A merged creature is simply a creature that remembers being both of its predecessors. Questions about teleportation and cloning get similarly simple-minded answers. The big remaining problem is where observed frequencies come from. Maybe there’s a big probability distribution over all observer-moments (“ASSA”), or maybe there’s something else.
Your comment contains an excellent point that can stand independently of Many Worlds (branching all the time). Namely, memory explains anticipation. Anticipation feels like a pointer to the next moment, but it’s just an inference based on a long sequence of memories.
There is nothing wrong with anticipating a future experience, but there is also no constraint against anticipating other future experiences as well. And most of us often do; we call that “empathy”. We have much more reliable history of knowing how correct/incorrect our anticipations about our own body were, though, and much less ability to ignore those outcomes. So self-concern feels different. But in some sense, self-concern is empathy for our future selves.
Just a nitpick, you don’t need many-worlds to have branching. Even in a classical world, if it’s large enough, there will be creatures with identical memories but different futures.
Right, as long as you don’t require causal connection for branching. To my mind “branching” suggests a causal connection, but the OP favors a pattern identity theory, so causal connection may be irrelevant.
Most people imagine their past and future selves as links in a continuous chain, each one connected to the previous one and the next one. It seems to me that the structure might be more like a singly linked list: each moment has a “pointer” to the previous one, by virtue of remembering it, but there’s no corresponding pointer to the next moment, and in fact there’s no unique next moment, because we are branching all the time.
(In functional programming languages, different linked lists can “share structure” by having pointers to the same tail. That idea also works for other data structures like trees, as described in Chris Okasaki’s book “Purely functional data structures”. That’s a digression, but a really interesting one.)
So why do we feel that we have a pointer to the next moment, or a probability distribution over next moments? I think it’s because we learn by induction. Looking at our memory, we see that each moment had a previous one and a next one, so we think the same must be true for the present moment as well. But it seems to me that the whole picture can be understood by just looking at links to previous moments, and nature doesn’t need to have any special laws about links to next moments, “observer fluid” and the like.
One of the benefits of such a “memory identity theory”, compared to “pattern identity theory”, is that it makes it easier to imagine merging two creatures. A merged creature is simply a creature that remembers being both of its predecessors. Questions about teleportation and cloning get similarly simple-minded answers. The big remaining problem is where observed frequencies come from. Maybe there’s a big probability distribution over all observer-moments (“ASSA”), or maybe there’s something else.
Your comment contains an excellent point that can stand independently of Many Worlds (branching all the time). Namely, memory explains anticipation. Anticipation feels like a pointer to the next moment, but it’s just an inference based on a long sequence of memories.
There is nothing wrong with anticipating a future experience, but there is also no constraint against anticipating other future experiences as well. And most of us often do; we call that “empathy”. We have much more reliable history of knowing how correct/incorrect our anticipations about our own body were, though, and much less ability to ignore those outcomes. So self-concern feels different. But in some sense, self-concern is empathy for our future selves.
Just a nitpick, you don’t need many-worlds to have branching. Even in a classical world, if it’s large enough, there will be creatures with identical memories but different futures.
Right, as long as you don’t require causal connection for branching. To my mind “branching” suggests a causal connection, but the OP favors a pattern identity theory, so causal connection may be irrelevant.