Most AI art producers begin from a model of child art.
Children’s art is knowledge-based, figure based, and emergent (unplanned).
In contrast, an “adult” (professional) figurative artist is observation-based, balances positive and negative space, and works on the composition before the content of the work.
A program that worked in this way would pre-scan the picture plane and tag its elements (pixels or pre-generated polygons) with relational information:
• close to the edge, centre, top, bottom, left, right
• on the edge of the picture plane
• proximity to other elements
• relative size
• formal qualities (jagged, smooth, triangular, square)
The next stage of the program can then work with a richer environment.
Embody compositional principles: golden sections, horizontal/vertical/diagonal, ground/sky line, etc.
Valence the plane, valence the compositional principles, then build the composition.
Give equal weight to figure and ground, to positive and negative space.
This is a technical exercise, but figure/ground relations are important in art. The boundary or distribution of shapes must be aesthetically interesting in some extra-aesthetic way. This sounds like “No Logo” or philosophy/business theory.