Im not familiar with this discourse, but why care about milliseconds? You can just keep looking at shorter and shorter time intervalls. As the time shortens, the explosion will shrink in probability, but the "normal" transitions of the brain will shrink in terms of "distance covered" while picking up the dropped probability, and this can be used to distinguish the two.
I also think there is potential in candidate #2. Evolutionary history frequency doesnt *define* normal functioning, though it is the most accessible measure for it. What you really want is something like the "selective footprint" of those conditions in your biology.
I don't think anything hangs on the thought experiments being put in terms of the bomb having a high chance of detonating at every millisecond. Shorter timescales would serve at least as well.
Interesting suggestion re candidate #2!
I suppose there's a question of whether defining normal functioning in terms of selectively imprinted conditions would result in normal functioning being implicitly dependent on evolutionary history (e.g. because the conditions wouldn't qualify as selectively imprinted without that history). If so, I think the worry that candidate solution #2 will make consciousness dependent on non-local factors will still apply.
There are two timescales: (A) the one at which the bomb likely explodes, and (B) the one we evaluate transition probabilities over. Im saying to look at the limiting behaviour of B getting shorter, for a given A. Then we see the probability of an explosion within B go down, while "normal functioning" over B increases in probability, but the changes made by it shrink. Basically, the only difference between the bomb threat and "Are you non-conscious because youll likely die in the next 200 years?" is what timescales were used to looking at.
>because the conditions wouldn't qualify as selectively imprinted without that history
What Im imagining is that the way the functional organisation "expects" certain conditions to hold has a physical correlate within it (which an atom-by-atom copy would share). For example, the brain evolved in a certain ambient pressure and expects it, but switching nitrogen in the air for helium doesnt matter, because the expectation is only for a gas inert under conditions in the body. Our ancestors didnt live in helium, so it *could* have been "abnormal" - but in fact, our construction is tolerant to it anyway. And a version of you that didnt tolerate it would have to have a different body, so this fact is somehow supervenient. Its just that actually spelling out all the details of how this works for all conditions would require a scientific understanding that we dont yet have.
At least this is the hope; Im less sure how this would work for synthetic minds that arent atomic copies for example.
>Im saying to look at the limiting behaviour of B getting shorter, for a given A.
When running the thought experiment we can stipulate that the detonation probability during a given interval is as high as we like, short of 1. So, for any given transition timescale, we can stipulate the the bomb is likely enough to explode to interfere with the probabilities at that transition timescale. In response, iou could then go to a shorter time scale at which transition probabilities aren't (much) disrupted. But then there will be a version of the thought experiment that also targets that time scale by stipulating a higher probability of detonation within the original interval.
Once we get to a sufficiently small time-scale (e.g. one well below the scale at which neurons fire) it will also become implausible that consciousness is sensitive to normal transitions at that time scale except insofar as they constrain normal transitions at are a larger timescale. So, if we stipulate a detonation probability that targets transition probabilities at a sufficiently small time scale, the response of moving to a smaller time scale loses its appeal, I think.
Thanks for further explaining the imprinting idea. I think I see the hope now, and how it could work if the "expectation" supervenes on internal conditions. It's not clear to me how much learning more scientific details would help. As an analogy, consider a simple, massive body in a Newtonian system that can only be subject to a limited range of external forces. In this case, we could specify various candidates for what counts as a normal external force on the system, all while considering only its internal features, not its history. In this case, even with complete scientific understanding it seems like the choice among candidates will be arbitrary. I could see an analogous point applying to the human brain even with full scientific understanding... though this could well be due to lack of imagination on my part.
>iou could then go to a shorter time scale at which transition probabilities aren't (much) disrupted. But then there will be a version of the thought experiment that also targets that time scale
To be clear, Im not talking about targeting some fixed short time scale, Im talking about the limiting behaviour for shorter and shorter times. Basically, instead of looking at transition probabilities over some particular finite time, Im looking at the derivative. This always outruns any higher probability.
This seems natural to me; long times are composed of many short times, and if you want to insist that something changing with time cant be evaluated over a short time, youll have a hard time defining it in a way where it can be evaluated over a long time. Eg if you know the integral of a function on any interval longer than t0, then in fact you know it over any interval, and therefore you know the whole function.
>As an analogy, consider a simple, massive body in a Newtonian system
I agree the body would have no expectation.
I think the question here is basically whether you can define the quasi-teleological properties of some system in a consistent way - but I think thats de facto a requirement for functionalism in general, even without the bomb problem. I mean, as soon as you try to evaluate a system that isnt a very close copy of one were sure is conscious, you need to define "the relevant functional features" in a more-that-ostentive way, and all the plausible candidates are these teleological properties.
Im not familiar with this discourse, but why care about milliseconds? You can just keep looking at shorter and shorter time intervalls. As the time shortens, the explosion will shrink in probability, but the "normal" transitions of the brain will shrink in terms of "distance covered" while picking up the dropped probability, and this can be used to distinguish the two.
I also think there is potential in candidate #2. Evolutionary history frequency doesnt *define* normal functioning, though it is the most accessible measure for it. What you really want is something like the "selective footprint" of those conditions in your biology.
I don't think anything hangs on the thought experiments being put in terms of the bomb having a high chance of detonating at every millisecond. Shorter timescales would serve at least as well.
Interesting suggestion re candidate #2!
I suppose there's a question of whether defining normal functioning in terms of selectively imprinted conditions would result in normal functioning being implicitly dependent on evolutionary history (e.g. because the conditions wouldn't qualify as selectively imprinted without that history). If so, I think the worry that candidate solution #2 will make consciousness dependent on non-local factors will still apply.
There are two timescales: (A) the one at which the bomb likely explodes, and (B) the one we evaluate transition probabilities over. Im saying to look at the limiting behaviour of B getting shorter, for a given A. Then we see the probability of an explosion within B go down, while "normal functioning" over B increases in probability, but the changes made by it shrink. Basically, the only difference between the bomb threat and "Are you non-conscious because youll likely die in the next 200 years?" is what timescales were used to looking at.
>because the conditions wouldn't qualify as selectively imprinted without that history
What Im imagining is that the way the functional organisation "expects" certain conditions to hold has a physical correlate within it (which an atom-by-atom copy would share). For example, the brain evolved in a certain ambient pressure and expects it, but switching nitrogen in the air for helium doesnt matter, because the expectation is only for a gas inert under conditions in the body. Our ancestors didnt live in helium, so it *could* have been "abnormal" - but in fact, our construction is tolerant to it anyway. And a version of you that didnt tolerate it would have to have a different body, so this fact is somehow supervenient. Its just that actually spelling out all the details of how this works for all conditions would require a scientific understanding that we dont yet have.
At least this is the hope; Im less sure how this would work for synthetic minds that arent atomic copies for example.
>Im saying to look at the limiting behaviour of B getting shorter, for a given A.
When running the thought experiment we can stipulate that the detonation probability during a given interval is as high as we like, short of 1. So, for any given transition timescale, we can stipulate the the bomb is likely enough to explode to interfere with the probabilities at that transition timescale. In response, iou could then go to a shorter time scale at which transition probabilities aren't (much) disrupted. But then there will be a version of the thought experiment that also targets that time scale by stipulating a higher probability of detonation within the original interval.
Once we get to a sufficiently small time-scale (e.g. one well below the scale at which neurons fire) it will also become implausible that consciousness is sensitive to normal transitions at that time scale except insofar as they constrain normal transitions at are a larger timescale. So, if we stipulate a detonation probability that targets transition probabilities at a sufficiently small time scale, the response of moving to a smaller time scale loses its appeal, I think.
Thanks for further explaining the imprinting idea. I think I see the hope now, and how it could work if the "expectation" supervenes on internal conditions. It's not clear to me how much learning more scientific details would help. As an analogy, consider a simple, massive body in a Newtonian system that can only be subject to a limited range of external forces. In this case, we could specify various candidates for what counts as a normal external force on the system, all while considering only its internal features, not its history. In this case, even with complete scientific understanding it seems like the choice among candidates will be arbitrary. I could see an analogous point applying to the human brain even with full scientific understanding... though this could well be due to lack of imagination on my part.
>iou could then go to a shorter time scale at which transition probabilities aren't (much) disrupted. But then there will be a version of the thought experiment that also targets that time scale
To be clear, Im not talking about targeting some fixed short time scale, Im talking about the limiting behaviour for shorter and shorter times. Basically, instead of looking at transition probabilities over some particular finite time, Im looking at the derivative. This always outruns any higher probability.
This seems natural to me; long times are composed of many short times, and if you want to insist that something changing with time cant be evaluated over a short time, youll have a hard time defining it in a way where it can be evaluated over a long time. Eg if you know the integral of a function on any interval longer than t0, then in fact you know it over any interval, and therefore you know the whole function.
>As an analogy, consider a simple, massive body in a Newtonian system
I agree the body would have no expectation.
I think the question here is basically whether you can define the quasi-teleological properties of some system in a consistent way - but I think thats de facto a requirement for functionalism in general, even without the bomb problem. I mean, as soon as you try to evaluate a system that isnt a very close copy of one were sure is conscious, you need to define "the relevant functional features" in a more-that-ostentive way, and all the plausible candidates are these teleological properties.