It has a statistical problem; it needs a scientific subroutine that updates itself. It always fails the visual perception test. See it to believe it! The test where it makes statistical errors is the following: two identical rooms, one with a black ceiling and one with a white ceiling—which one will appear taller to a human observer? It incorrectly answers “white,” but after being prompted to observe the Gestalt, it corrects itself and admits its mistake!