Other applications beyond head and hand interactions are also under investigation. We shall enumerate a few concepts and leave it to the reader's imagination to conjure other scenarios. One of the virtues about the ARL system and its perceptual, data-driven nature is that it is flexible. There are no explicit mechanisms here to exclusively learn head and hand dynamics. The only systems that were specific to head and hand motion were the vision and graphics system. The time series processing and learning algorithm did not specifically require such types of data. In the learning system, there were no cognitive or kinematic models that concerned gesticulations or hand gesture constraints. Although such models could have been helpful additions in this scenario, they could be detrimental when the ARL learns a different modality (i.e. facial motion). The lack of such hard-wired models increases the generality of the approach. Thus, a large space of time varying measurements and outputs can be handled and the behaviours recovered could have had a markedly different structure.