The problem with p-zombie argument is that it is completely inconsequential.
It is like if i say there is some property let's call it phi-factor, i am certain i have it, some people have it too, and some do not, but there is absolutely no way for others to tell who has this phi-factor thing and who does not, because it doesn't cause any difference in behavior of people.
This kind of hypothesis doesn't explain anything and does not predict anything, and therefore it is not possible to build any logical chain based on it.
There are good versions of p-zombie argument, e.g. you could say that it is very hard to find difference in behavior of people and p-zombies, but if you create planet like simulation inhabited with p-zombies they will have large differences, e.g. they will never develop art, or science, or religion, or will die out because will not want to have children.
In any case there needs to be some measurable difference in behavior, for argument to be valid. But since this is a normal testable theory, it does not lead to the same kind of self-contradictory statements as p-zombie version, and therefore is not popular with philosophers who want to talk about things they have nothing to say about.
The argument is not inconsequential, you just need to understand the work the argument is trying to do. If you are only interested in testable hypotheses, then the argument isn't relevant. But that's not the only thing that matters in Science. Science doesn't operate free of assumptions or conceptual frameworks. An experiment is situated within a set of background assumptions that determine what the experiment outcome says about the world.
Our conceptual framework for understanding consciousness also matters, and the p-zombie argument is directly relevant to what conceptual framework is a plausible basis for doing a science of consciousness. The zombie argument puts limits on what the brain sciences can say about consciousness as such, i.e. the qualitative feel of subjective experience. The very point of the argument is to elucidate the limits of scientific explanation regarding consciousness. For that it is very consequential.
Could you bring any example where non-testable hypothesis have mattered in science, or even have been proposed?
As far as i can tell the p-zombie argument is a variant 2=3 "proofs" with hidden division by 0. People claim as truth a proposition with no consequence,then claim that a proposition they want to prove follows from it. It does not elucidate anything.
The limited variant of the argument which predicts something does help in understanding the limits, but p-zombie supporters do not like that argument because it also shows the boundaries of what brain sciences can't say, and that is not much.
>Could you bring any example where non-testable hypothesis have mattered in science, or even have been proposed?
If you think of science as just a collection of empirical data, then you won't find any value in untestable hypotheses. If you think of science as an enterprise that's aimed towards explanations about the world, then conceptual data matters. Interpretations of quantum mechanics are a fruitful ground for non-testable conceptual work to be relevant to science. Here is one example[1] of the Many Worlds interpretation providing conceptual and intuitive support that helped advance quantum computing.
But beyond that, what an experiment tells us depends on our conceptual framework. When we hear a click in a Geiger counter, are we merely detecting clicks in this particular device, or are we detecting electrons out in the world? This isn't something that an experiment will decide for us. The conceptual framework we are operating with allows us to say the click isn't just the detector going off, and our model of these clicks isn't just predicting when this device will click in the future, but rather that our experiments are detecting the existence of actual electrons too small to interact with without special devices.
Interpretations of quantum mechanics are testable, they have to produce quantum mechanics, which puts a very narrow constraint on possible theories.
The equivalent of p-zombie hypothesis with quantum mechanics would be if one said that particles in addition to coordinates have to be described by one more variable, which does not affect any of other equations of motion. These type of theories do not have to obey any constraint, and you can trivially produce an infinite amount of them.
You can say that this kind of variable has different value for particles in human brain, therefore study of this additional variable is more important than the study of the rest of physics. But this won't improve your understanding of either the brain or physics, just like p-zombie hypothesis does not.
>Interpretations of quantum mechanics are testable, they have to produce quantum mechanics, which puts a very narrow constraint on possible theories.
No, this is not what anyone means by a theory being testable. An interpretation of QM by assumption reproduces all the same results of QM. That's why its an interpretation. In the same way, the p-zombie thought experiment by assumption reproduces all of physics without any deviation. QM interpretations and the p-zombie argument are on the same footing here.
>These type of theories do not have to obey any constraint, and you can trivially produce an infinite amount of them.
This just shows a lack of understanding of what the p-zombie argument is claiming. Perhaps revisit your initial misconceptions rather than piling on the mistakes?
Interpretation of QM does not "reproduce results by assumption", it provides a computational scheme for obtaining equations of QM. P-zombie argument on the other hand does not produce anything.
It is like if i say there is some property let's call it phi-factor, i am certain i have it, some people have it too, and some do not, but there is absolutely no way for others to tell who has this phi-factor thing and who does not, because it doesn't cause any difference in behavior of people.
This kind of hypothesis doesn't explain anything and does not predict anything, and therefore it is not possible to build any logical chain based on it.
There are good versions of p-zombie argument, e.g. you could say that it is very hard to find difference in behavior of people and p-zombies, but if you create planet like simulation inhabited with p-zombies they will have large differences, e.g. they will never develop art, or science, or religion, or will die out because will not want to have children.
In any case there needs to be some measurable difference in behavior, for argument to be valid. But since this is a normal testable theory, it does not lead to the same kind of self-contradictory statements as p-zombie version, and therefore is not popular with philosophers who want to talk about things they have nothing to say about.