ocr: Input 0=45 Synaptic Disturbance V Iu X Outputs OIBTUKBING A NEURAL NET trained on auto body-shapes results in a progression of shapes that the network has never seen before. Turning up the magnitude and the numbers of disturbances ultimately pro- duces a nonsensical shape, a Picasso-style creation rather than a Detroit design. This progreesion from the plausible to the absurd shows that limited amounts of degradation provide useful results.
