Directed Information Flow in Computing Systems with Living Neurons

There is growing interest in building artificial computational fabrics from biological neurons, a kind of cultured intelligence. We briefly review the significant experimental advances needed to construct such computing systems with living neurons, which are also called synthetic neurocomputers. These novel systems are designed for energy-efficient implementation of numerous embodied intelligence applications, but also enable highly-controlled experimentation to investigate questions in neuroscience itself. The concept of information flow—fundamental in computation and control—is central to this investigation, now manipulable in biological neural networks with unprecedented flexibility and precision. The paper discusses strategies for modulating information flow through structural connectivity and neural plasticity, and presents a case study on information-theoretic analysis of information flow. It concludes by discussing research directions opened by the ability to precisely control information flow in living neurons, proposing future investigations into whether neural systems operate close to their informational limits within the constraints of their physical makeup and future design principles for constructing synthetic neurocomputers.