There is a theorem in machine learning (blanking on the name) that says any "learner" will have to be biased in some sense.
Also, just because we can't expect to be free of bias doesn't mean that the bias is "proper functioning" of the hardware. An expected failure, perhaps, but still a failure.
I make a finer distinction of "failure" as something that's inefficient for it's clear purpose. E.g. Laryngeal nerve of the giraffe. Evolution will do that on occasion. Sensory interpretations that optical illusions are based on are often optimal for the environment, and are a complement to the power of evolution if anything. Viewing something that is optimal as a failure seems like wishful thinking (though I suspect this is more of a misunderstanding of neurobiology).
Take off every 'quote'! You know what you doing. For great insight. Move 'quote'.
And if you don't: