Graphic design is a very large field, essentially encompassing the planning of all communications that come to the receiver through the sense of sight. Within this large territory, the application of A. I. is making inroads in well-publicized areas such as recognition software (i.e. facial recognition) and the making of discrete objects such as logos and simple pictographs. One of the major areas of graphic design, however, involves neither pattern recognition nor the making of a graphic element, but rather the composition of many graphic elements, including text and photographs, onto a surface. Whether that surface be a screen or a paper page, the combinatorial possibilities are infinite. Yet, the manner in which the elements are controlled and relate to each other, vectorally and proximally, play a large part in the semantic meaning of the display. It is hypothesized on the basis of semiotics (especially the concept of semantic profiles) that the connection between kinds of layout and the culturally-linked semantics that result from the layout can be defined. If so, then it should be possible to devise meta-design algorithms that can be inserted into graphic design programs that would be able to produce prospective layouts of books, catalogs and websites, etc. Ideally, such algorithms would include a machine learning component so that over time they improve. Such technology would function much like a designer’s assistant and could be an important creative tool. My presentation will cover the potential of this approach and problems that will need to be overcome.