We cannot expect to know the detailed "wiring diagram" of the nervous system for any intelligent creature for quite a long time. Even then, the true organization is likely to be incredibly complex and tangled. However, in order to build intelligent robots now, we need a plausible interim architecture. A functional model for robot organization is proposed, starting with a basic, first order model, which is gradually refined. In particular, it is proposed that associative memory provides a useful -and perhaps plausible --basis for an intelligent system.