Experiments in the nascent field of extit{NeuroAI} suggest that some deep neural networks are capable of executing some of the same operations as the human brain. Consequently, these experiments appear to offer unprecedentedly direct evidence for a liberal version of the functionalist position in the philosophy of mind, according to which mental properties can be realized in non-biological systems. I scrutinize this hypothesis by subjecting it to a stringent test from the literature on multiple realization. It turns out that the empirical assessment of this kind of multiple realization yields indeterminate results, because it is not clear whether the mental kind in question is genuine. This indeterminacy suggests that the claim that a computer carries out the same mental operation as some component of the human brain has less empirical content than functionalists have tended to presume. I conclude by describing how this result extends and sharpens a family of older extit{triviality arguments} against functionalism.