The premise about computation being observer-relative overestimates how much meaning we read into a system, and underestimates how much we read out of it. When two 1kg bags of sand land on a lever and counterbalance the one 2kg bag at the other end, that has an objective, logical, correspondence-based meaning that transcends observers. When a domino computer "executes" - same. When a silicon computer computes - same. As the computer gets more complex, it becomes harder to intuit how the physical working is objectively a computation, but this proceeds by straightforward extension from the bags of sand.

We can certainly attach semantics to the numbers (e.g. saying this bit pattern represents dollars, or a spaceship's shield %), and that is observer-relative. But that is completely different from categorizing or understanding a process as a computation in general, in terms of logic.

I'll stop short of implying that it's the same for the human brain, because nobody should be pretending to understand the brain at this point in history. However, this does provide a way to see how it could be true for the brain, if it is ever determined that the brain is precisely equivalent to a computer.