Intelligent Computers and Visual Artists

In order to be of value to artists, computers must be perceptive and knowledgeable in visual matters. Being ‘perceptive’ means that they should be programmed to deal with phenomena that artists perceive and find interesting. Being ‘knowledgeable’ means here that the computer can use information stored in it to take appropriate courses of action.

This paper deals with techniques that might be employed rather than discussing what aesthetic phenomena might be of interest. The techniques described are drawn from the so-called ‘linguistic’ approach to graphics, which is interpreted very broadly to include descriptive, hierarchical, structural, grammatical, and proceedural methods. In short, nonnumerical rather than numerical methods based on statistical criteria, information theory, evaluation of mathematical functions or optimization.

The paper refers to the generation of shapes using grammars, analysis using grammars, picture description languages, explicit representation of semantic information using semantic nets and implicit representation using procedures. Human subjective observation of computer behavior is suggested as a sufficient guide to the valid application of these techniques.