Computer Science · Artificial Intelligence
A Principled Approach for Systematic Mind Engineering
Ricardo Sanz, Jaime Fernández Gómez, Carlos Hernández, Adolfo Hernando
See on the mapAbstract
The systematic, engineering-grade construction of cognitive systems with up-to-human capabilities cannot be based on the exploratory programming methods that have populated the artificial cognitive systems effort for a simple matter of scale. We need a strong methodological focus to move from artisany into the requirements-driven engineering methods that are necessary to build systems with the required set of competences in any design dimension (any-x systems). The ASys Vision is an structured vision for any-x artificial mind engineering that is based on a universal conceptualisation of a situated, functional, self-aware cognitive system: an agent that exploits models —of the world, of the self— for the performance of situated tasks. It has been demonstrated that the performance of a controller is related to its capability of internalising models of the systems under control (Conant and Ashby 1970). This result is well known in practice in the field of predictive process control; and while recognised by philosophers as early as Craik and theoretical biologists as Rosen, the question of model-based minds is still controversial in some fields (like some areas of neuroscience or enactive cognition). The mental model then constitutes the very substrate of the agent’s epistemology and is efficacious to the extent that it is in congruence with the reality surrounding the agent. This surrounding reality includes the agent’s body, because model-based minds include models of the own body as part of their control strategical resources. That the biological mind represents and models the body is a well-known result from neuroscience and recent results and theories suggest that mental model execution sustains agent awareness. Obviously mental models are not complete; they capture only parts of the reality and in a usually simpler abstraction. Model complexity hampers scalability of the exploitation strategy and hence the economy-driven evolutionary-shaped strategy of a cognitive agent is to build limited, focused adhoc models that are activated on-demand based on mental state and current perceptual flow.