The second architecture is a combined AI system
that uses neural networks and two-dimensional filters
to do first line object detection and locations of
objects in an image input and then uses LISP type
AI relationships to do object manipulation and image
evaluation(figure 2). The advantage of this architecture
is that it can be developed one detection
algorithm and one relationship at a time. Neural networks
and self-organizing nets have a unique ability
to recognize patterns in two dimensional data (Carpenter
and Grossberg 1991). Therefore, in this architecture
the nets do the low level detection and pattern
recognition and the other AI systems make decisions
about what the patterns are based on relationships.