AI and IoT Architecture Based on Markov Blankets

The Internet of Things delivers data from sensors and devices around the environment. The problem for many organizations is making sense of all that data. Digital Agents (DA) can be used as a conceptual schema for modeling, planning actions, and predicting context behaviour starting from data. The Markov blanket allows to identification of the relevant characteristics for an analysis or a prediction and is connected to the principle of free energy and active inference, which are fundamental concepts to explain the autonomous systems. According to this perspective, DA systems consist of a hierarchy of nested Markov blankets, which self-organize to minimize surprise or uncertainty. It can be considered autonomous and context-independent if it respects these statistical parameters. A DA can be the basic building block to create a new generation of intelligent systems.