According to computationalist theories of the mind, consciousness does not depend on any specific physical substrate, such as carbon-based biological material, but automatically arises out of the right kind of computational structure. Even though this thesis has become an unquestioned assumption in most of current AI literature, there exist only few direct arguments in favor of it. One, and probably the most prominent, argument for computationalism from the philosophy of mind is David Chalmers’ dancing-qualia argument. The aim of this paper is to challenge this argument from a hitherto neglected angle, by arguing that it is undermined by some experimental results in neurobiology regarding the workings of phenomenal memory. However, I will argue that Chalmers’ overall case for the possibility of conscious AI can still be vindicated.