Beyond Reactive Systems: An Associative Memory for Sensory-Driven Intelligent Behavior

Searching to reproduce intelligence by the mean of modelling cognitive abilities has been a trend of artificial intelligence research for over fifty years. With respect to this, the relatively limited advances of this last decade suggest that new areas of consideration need to be investigated. In particular, Nouvelle AI, which originates from a more ecological approach, claims a critical part for solving issues around embodiment. In this paper, we hypothesise that intelligent behaviour can be obtained without the need for cognitive modelling. Towards exploring this, we propose a model for a connectionist controller where an associative memory is used to enhance the adaptation of an autonomous agent operating in a complex environment. Initial experiments yield promising results, and we have found that simple agents show intelligent behaviour.