Thank you for your reply. I'm not sure I understand your objection perfectly.
> it sets up a future conflict based upon the state that a human has
It's not clear to me it does.
> It's based on a number of presumptions that only a particular set of arrangements of matter, well, matter.
Well, this premise must be true, I think. How could it be otherwise?
> And those that matter happen to be watery sacks of meat.
This I disagree with. My comment contains no such claims that only humans matter.
> So in your framework you setup a definition of what an external measurement of someones sentience state... Now, someone makes a machine that outputs those states and externally behaves as if it has sentience.
So, this is something we need to be very cautious about, I think. First, we don't really know what makes someone sentient. I do not claim to know, and certainly not know with a high degree of certainty (I may have a few incomplete hypothesis).
First, it's clear to me that someone's external behavior does not dictate sentience (positively or negatively). I gave some examples, but I think this can be pretty reliably established. To make things simpler, divide the candidate sentient systems by their "output quantity/quality", from less to more. (1) On the "no output side", it's clear that someone can think a lot and have a wonderful inner life without being able to speak. There are (some medical) cases of people going long periods without speaking by choice or by medical condition. Those people can be sentient just fine, obviously you shouldn't need to speak or "output" anything to be considered having an inner life. (2) On the "human-like output" (or superhuman), the argument becomes more involved. First, clearly just the output itself being indistinguishable from human is not sufficient because it could be for example a simple program that just pastes some verbatim copy of a never seen dialogue. Or very simple chatbots that are quite convincing at appearing human (can fool many judges for a while), but clearly also are not sentient. Even for a for a system that genuinely resembles a real human even with under scrutiny, (with all sorts of similar abilities across all fields), consider the case that this system could be simply one astronomically large (unrealistic, I know) lookup table. Basically, it simply works inside by fetching an indexed piece of data and returns it. This is possible, although this machine would need to be astronomical in scale (so it's more of a thought experiment). It shows to me, convincingly, the actual "internal" or "microscale" even implementation of your thoughts matter critically for sentience. I presume, for example, that even if some small portion of your mental activity (i.e. a kind of "subcircuit" in your brain) were replaced by something like a lookup table, you would lose part of the experience of that mental activity. This is clear when the content of your mind differs when you're trying a (kind of) problem for the first time and when you're solving one for the umpteenth time. The latter (possibly, not necessarily) feels more boring, also normally more automatic with less thinking required as you've built habitual responses.
> What do you then? Upturn the table and get mad because your classification system no longer just contains humans?
I would do no such thing. I would welcome any other kinds of sentience. But it would be important to me to carefully establish and understand what kind of systems are sentient, indeed in large part to protect them.
---
My main point is that you kind of need to start with sentient minds to understand what sentience is (and of course, not only that, but sentient minds that are aware of their sentience and willing/keen to study it). If you start without sentience, it doesn't seem like this particular concept would form a necessary basis for a theory of action or motivation. Like you can't prove from elementary mathematical premises that sentience is important, I don't think. We need to start from some knowledge that derives from first-hand experience of being an actual sentient being, and expand, solidify or formalize this knowledge to be able to understand it with scientific precision (or anything close to it).
> it sets up a future conflict based upon the state that a human has
It's not clear to me it does.
> It's based on a number of presumptions that only a particular set of arrangements of matter, well, matter.
Well, this premise must be true, I think. How could it be otherwise?
> And those that matter happen to be watery sacks of meat.
This I disagree with. My comment contains no such claims that only humans matter.
> So in your framework you setup a definition of what an external measurement of someones sentience state... Now, someone makes a machine that outputs those states and externally behaves as if it has sentience.
So, this is something we need to be very cautious about, I think. First, we don't really know what makes someone sentient. I do not claim to know, and certainly not know with a high degree of certainty (I may have a few incomplete hypothesis).
First, it's clear to me that someone's external behavior does not dictate sentience (positively or negatively). I gave some examples, but I think this can be pretty reliably established. To make things simpler, divide the candidate sentient systems by their "output quantity/quality", from less to more. (1) On the "no output side", it's clear that someone can think a lot and have a wonderful inner life without being able to speak. There are (some medical) cases of people going long periods without speaking by choice or by medical condition. Those people can be sentient just fine, obviously you shouldn't need to speak or "output" anything to be considered having an inner life. (2) On the "human-like output" (or superhuman), the argument becomes more involved. First, clearly just the output itself being indistinguishable from human is not sufficient because it could be for example a simple program that just pastes some verbatim copy of a never seen dialogue. Or very simple chatbots that are quite convincing at appearing human (can fool many judges for a while), but clearly also are not sentient. Even for a for a system that genuinely resembles a real human even with under scrutiny, (with all sorts of similar abilities across all fields), consider the case that this system could be simply one astronomically large (unrealistic, I know) lookup table. Basically, it simply works inside by fetching an indexed piece of data and returns it. This is possible, although this machine would need to be astronomical in scale (so it's more of a thought experiment). It shows to me, convincingly, the actual "internal" or "microscale" even implementation of your thoughts matter critically for sentience. I presume, for example, that even if some small portion of your mental activity (i.e. a kind of "subcircuit" in your brain) were replaced by something like a lookup table, you would lose part of the experience of that mental activity. This is clear when the content of your mind differs when you're trying a (kind of) problem for the first time and when you're solving one for the umpteenth time. The latter (possibly, not necessarily) feels more boring, also normally more automatic with less thinking required as you've built habitual responses.
> What do you then? Upturn the table and get mad because your classification system no longer just contains humans?
I would do no such thing. I would welcome any other kinds of sentience. But it would be important to me to carefully establish and understand what kind of systems are sentient, indeed in large part to protect them.
---
My main point is that you kind of need to start with sentient minds to understand what sentience is (and of course, not only that, but sentient minds that are aware of their sentience and willing/keen to study it). If you start without sentience, it doesn't seem like this particular concept would form a necessary basis for a theory of action or motivation. Like you can't prove from elementary mathematical premises that sentience is important, I don't think. We need to start from some knowledge that derives from first-hand experience of being an actual sentient being, and expand, solidify or formalize this knowledge to be able to understand it with scientific precision (or anything close to it).