A roadmap for autonomous adaptive systems: The brain-guided attention (BGA) system

We develop a brain-guided approach to autonomous adaptive systems (AAS). This is especially to determine the various components of information processing an AAS need possess in order to be able to be flexible and adaptive in new environments and to be able to possess values of objects in the world and actions on them. These components are based on the faculties of perception and categorization, of attention, of object and action valuations and of long-term memory. Further high level executive processes are assumed to be available to an AAS. This is so that it can be effective in its environment in terms of its guidance, through learnt valuations, of decisions to be made of actions on that environment. Finally we raise ethical questions about the value of such autonomy to modern society and the problems it may cause.