If you could damage wires in a certain way and make the voices forget how to pronounce nouns, eliminate their short-term but not long-term memory, damage their color words, and so on, you would have a solid case for the wires doing internal, functional information-processing in causal arrangements which permitted the final output to be permuted in ways that corresponded to perturbing particular causal nodes. In much the same way, a calculator might be thought to be a radio if you are ignorant of its internals, but if you have a hypothesis that the calculator contains a binary half-adder and you can perturb particular transistors and see wrong answers in a way that matches what the half-adder hypothesis predicts for perturbing that transistor, you have shown the answers are generated internally rather than externally. In a world where we can directly monitor a cat's thalamus and reconstruct part of its visual processing field, the radio hypothesis is not just privileging a hypothesis without evidence, it is frantically clinging to a hypothesis with strong contrary evidence in denial of a hypothesis with detailed confirming evidence.