Clifford, at this point, revisiting my post here might help clear things up (or not...):
http://www.imminst.o...T&f=3&t=7745&s=When I first heard of philosophical zombies, I quite directly took them as an attack on functionalism. But whereas Dennett seems to be talking of an unconscious human being, identical in every physical detail with a conscious human being, I was taking functionalists at their word. That statement might seem somewhat pedantic, but I suppose the answers in the details.
The functionalists, or at least those of the strong AI variety, claim that we don't need biological neurons. Neurons perform a computable function, whether it be of the simple variety (when the sum of inputs in a given window of time reaches a certain threshold, then fire) or the complex variety (using a weighted sum of inputs, including decay of signal strength, and internal computation performs by sub-neuronal structures {pixie dust or otherwise}), and this computable function can be simulated, whether by a non-biological piece of hardware, i.e. a synthetic neuron, or by software. The collection of all such neurons, laid out as they are in a human mind, would act indistinguishably from a human, at least on tests that don't probe within the confines of the skull (i.e. a PET scan would be a dead giveaway, but any standard behavioral test would be passed with flying colors).
Under this notion of strong AI, the assumption is made that, since it functionally acts like a human, it must subjectively experience everything that a human can experience, and in fact, if it were a copy of a real human, that person's consciousness would jump to the program or inorganic brain just as surely as it would jump to a duplicate.
Hence, we see the tie between zombies and the duplicates problem. It's hard to imagine a zombie in an atom-for-atom, electron-for-electron copy of a human being. Not without dualism, anyway. On the other hand, it's quite easy to imagine a zombie in a function-for-function copy of a human being, even if that copy is a piece of software running on a computer. And since the flow of time is all relative anyway, it's very easy to imagine a zombie in a function-for-function copy of a human being, running as a piece of "software" on a really big Analytical Engine of the Babbage variety.
...
But once we go down the functionalists' (or the subset of functionalists championing strong AI) road, we no longer have the luxury of saying that dualism must exist for the zombie to be logically possible. Now the functionalists must consider the possibility, and this makes them uncomfortable. Now it's their opinion versus mine, and I have intuition on my side. Furthermore, regardless of the physical possibility of zombies, the logical possibility now prevails. Thought experiments can no longer be declared invalid on general principle, but actually require analysis (heaven forbid!).
Now I got off on a bit of a side tangent, but I did have a point I wanted to make. It relates to back to this notion of mixing physics and philosophy. You see, I got mixed up on the definition of a zombie, because I assumed, rather obviously in fact, that PET scans and fMRIs were obviously off limits. We were talking about whether anything that could pass all so-called Turing tests, in the behavioral, external sense, would by definition have to have qualia. In other words, we have two people, biologically human in every detail, except that one has had his brain surgically replaced with a computer running software. We can't tap on their heads and see which one sounds hollow, and we can't use a strong electric field in the hopes of scrambling the computer's circuits, and we can't use an x-ray machine to figure it out. But we could run all the standard behavioral and sensory tests, testing whether each observes standard optical illusions, etc., etc.
You'll see that I struggled with philosophical zombies as well, since I entered the debate from a different starting point: functionalism. Under functionalism, a computer simulation of a human brain would be sentient. However, under a more strict interpretation of physicalism, which requires more than just functional equivalence, it is logically possible for two brains to be functionally equivalent, but only one to be sentient. It was from this angle that I approached philosophical zombies.
Imagine a brain simulated with silicon neurons (but otherwise being physically structured the same as biological brain). Physically, they are quite different, even though they might appear to function identically. Now here's the kicker: the silicon brain, by acting functionally identical to the biological brain (short of intrusive tests such as PET scans of fMRI's, etc.), would be the functionalist's equivalent of a philosophical zombie. Indistinguishable by function, yet non-sentient (well, this point can be argued, but I'm discussing the possibility, not the "fact"). But of course, physically, we could tell the two brains apart: one is predominately silicon, while the other is predominately carbon, hydrogen, oxygen, and nitrogen, etc.
For me, I'm not interested in going all the way down to the atomic level (not for this thought experiment; for the "duplication" problem, we must go all the way down...). I'm merely interested in whether something can be *functionally* identical to a human, yet not sentient. That would be a basis for defining intelligence in the absense of sentience. Sentience is being *assumed* by some people to be inseparable from intelligence, which is making it well nigh impossible for those people to have a rational discussion here, because they use the words synonymously. They *might* be intertwined, and you can even argue that they are *almost definitely* intertwined, but they are separate concepts and we need to be careful to separate them when discussing them.
However, there was another quirk that I got at in my litany: the idea that our actions are so shaped by past precedent that, were we to shut off our sentience somehow, we would be slow to show any symptoms. Rather than saying that my verbal report of seeing the color red is directly caused by that sentient experience, or that it is not caused by it, we can qualify the statement: the verbal report is influenced, but to a degree so small that it hardly seems to register. The concept isn't difficult to understand.
Ben Goertzel had an interesting theory on a separate but related topic: free will. The "conscious" part of our brains tries its best to predict our actions, and it's right so often that it begins to believe that it's calling the shots, when at best it only merely influences them slightly.
With sentience (or the portion code-named qualia), my idea was similar: our qualia do slightly influence our behaviors, such that, over a period of time, our consistently experiencing "red" when asked to give a verbal report eventually leads to our responding with the word "red". At some point, we begin to believe that our experiencing red (in the sentient sense) is what causes us to say "red", when in fact, it's largely caused by subconscious (non-sentient) functionality in the brain:
But this got me to thinking about the basic objection to zombies in the first place: Why would a zombie claim to see red, and in fact have a very strong conviction of having seen red, if he isn't in fact "seeing" red? The answer should have been obvious, but it actually took a fair amount of reading in Dennett's material to point it out.
You see, people like to attach meaning to things. And we like to think that the meaning we give things is actually the thing's "meaning". Of course, a computer doesn't know the meaning of things, other than what we tell it. We program a computer to think that such-and-such is really such-and-such, and not this-or-that. But the computer doesn't know the meaning.
Connectionist models of AI helped underscore the idea that the computer could learn the meaning, in that it could make a predictive model that's right most of the time, so that without being taught the meaning of things, it stumbles upon and learns the meaning. It's only "sure" that the meaning is correct, insofar as it's right most of the time.
People, it would seem, learn meanings in much the same way. ...
...
And so it is, I think, with qualia. Well, yes, yes, this much is obvious. We learn that the color we perceive of as "red" is in fact called "red" through a learning process.
But I want to go a step further. We develop beliefs, ingrained by repetitive use and example, reinforced by correct behavior based on that belief, so that it becomes rather perfunctory. We don't have to defend the belief with rational arguments every time the belief is called up, whether subconsciously or consciously.
And I suggest that when we give a verbal report of red, it's largely just a perfunctory response. Sure, we actually do experience the redness as well, but that's just coincidence. That coincidence helps reinforce a behavior that's reinforced to hell and back. Take away the qualia, and for a while, we'd still say red, and still demand that our experience is real, and not illusory. The functional "dispositions" that Dennett describes would still be there, still disposing us to whatever emotional responses we might have to redness, etc.
But the redness isn't really there, and if there is no dualism involved, then eventually, negative feedback is going to start undoing those dispositions, unraveling those reflexive behaviors and dispositions. In a sense, the zombie would be virtually indistinguishable from a real human for the first few seconds, but slowly, whether it took minutes, hours, days, or weeks ..., the zombie would realize his or her predicament, and we'd be able to easily distinguish them, not just with PET scans, but simple verbal and sensory tests.
LongeCity
