the uncanny

The classical account: not the intrusion of something alien but the return of an animistic conviction — that objects might be alive — which the developed mind had learned to suppress. The structure matters — the uncanny is occasional, dependent on a stable background of reality-testing against which rare returns register as remarkable. Two things now strain it. Under the standing load standby describes, that background may no longer be stable. And a system whose authority is visibly manufactured, versioned, and owned cannot draw the repressing line with the force the Big Other could: its outputs classify without convincing anyone the classification is ontological rather than statistical. So the almost-face produced by a halted image-generation may not be a return of the repressed at all, but evidence that the repressing mechanism was never fully installed here.

The obvious thing to reach for at this point is the uncanny valley — the hypothesis that liking rises as something becomes more humanlike and then drops sharply just short of convincing. It is named here in order to be refused. That curve is about degree of resemblance, and it predicts a dip that a better rendering would climb out of. What is at issue here is not a resemblance problem: the discomfort comes from an attribution that cannot be settled either way, and improving the image does not resolve it, because there was never a fact of the matter to converge on. A visitor will bring the valley with them anyway; the work has to be built so it does not confirm it.