Nowadays, the Internet has become indispensable to our daily lives. Since its birth, the Internet has been researched and developed with the primary goal of delivering bit strings quickly and without error. In other words, researchers mainly focused on achieving highly reliable and low latency communications.
However, the communication paradigm based on accurate transmission is reaching its limits. The rapid expansion of scale and the diversification of applications and environments cause various problems. We face multiple issues such as system outages and performance degradation due to unexpected events and increased power consumption.
On the other hand, the brain network realises extremely sophisticated information processing. The brain network consists of a large number of simple nonlinear elements, i. e. neurons. They are intricately connected and exchange electrical signals. Even when various errors and noises are introduced, appropriate answers (outputs) can be derived with high probability. Although there are still many unexplored aspects of the brain's robust and low-power-consumption mechanisms, rapid advances in measurement technology reveal their physical and functional structures. These mechanisms are applied to artificial intelligence technologies.
In the field of information networking, we believe that we can establish new communication systems and technologies by learning from the brain. The new communication systems and technologies can be robust, low power consumption, and sustainable by learning from the brain.
For example, we study the Brain-morphic Wireless Sensor Network (B-WSN). The BWSN is composed of nodes consisting of binary sensors and electrical circuits that mimic the behaviour of neurons. There are spiking signals exchanged between the nodes to create a spiking Neural Network. The network does not need the control of structure, signals, and routing. Therefore there is no overhead for network control. Furthermore, nodes only generate spike signals in response to changes in (virtual) membrane potential. Also, no sensing information is needed in the spike signals.
By observing signalling spikes of partial nodes, the technics of reservoir computation (which is a kind of information processing model in the brain) can accurately read when, where, and what happened. When the binary sensors are distributed, a neural network is formed and starts information processing, similar to the brain. It is a new paradigm of the information communication system that can be called "Network as a Brain".