EXCERPT The Toy Universe of an Embodied Automaton Suppose a hypothetical embodied automaton possessing the following properties and features: Feature 1: It has been programmed to instantiate a number of diffuse and recurring goals centred on the maintenance and preservation of the system qua agent. The goals are at different levels of complexity, and exert various degrees of pressure on the system, analogous to biological needs. Furthermore, such pressures can be understood as necessary physical constraints imposed upon the agent as an information processing system, its computational capacities, its situated behavioural responses and even its range of possible behaviours. This means that our agent is essentially a limited being which possesses neither infinite information processing time nor infinite resources. We would not have been able to even characterize this agent as an information processing system, if it was not restricted by the physical and computational cost constraints existent in its sensory-behavioural matrix. In a nutshell, just as we cannot speak of computation without constraints of run time and computational cost, so we cannot speak of an agent as an information processing system without the physical constraints inherent to how it interacts with the environment by way of goal-specific behaviours. We might as well talk about supernatural beings. As we will soon see, the sense in which these goals are comparable to our own is complicated, but it is from here that we must begin. Feature 2: It has been wired to engage in activities that increase the probability of goal-fulfilment. The wiring of the automaton can be thought of in terms of interacting levels of structure that can produce transient multiple variations in the internal states of the system. The wiring or neural structure comprises primary low-level information processing modules, additional higher-level modules for intermediating primary modules, and workspaces through which processed information can be selected and made available to a higher-level global workspace that can be accessed by executive functions and goal-oriented behaviours. The wiring should be sufficient to causally mediate between complex environmental inputs and the complex behavioural outputs of the agent. In short, the agent can be defined as a teleological system with sufficient wiring (structure) to be capable of reliable differential and adaptive responsiveness with respect to its environment. […] Feature 3: With regard to its capacity for reliable differential responsiveness (i.e., the sufficient structure required for causal mediation between environmental inputs and behavioural outputs), the agent is equipped with different specialized sensors and different modules for integrating sensory data both within a specific sensory modality (multiple data associated with one sensor) and across different modalities (data associated with different sensors). This sensory integration extends the spatial and temporal coverage of the sensory information processing system and increases the robustness and reliability of sensory information. In addition, it reduces the ambiguity of sensory input, and thus increases salience. Specifically, integration across different sensory modalities expands the range of behaviours and yields a stronger effect on the behavioural output. The sensory integration is carried out at different levels, distinguished in terms of models of information processing (sequential or concurrent), temporal links between different streams of input (synchronous and asynchronous), and various frames of reference. Feature 4: The automaton is furnished with a sufficiently complex and functionally flexible memory capable of encoding, retrieval, consolidation, discarding, and transfer of sensory impressions. Rather than operating as an established storage space for the retention of past impressions, this memory is modelled as an adaptive dynamic process that plays a constructive—or more accurately, simulative—role in the behaviours of the agent…