It from bit? The future of bioinspired computing beyond ML

Solving 21st-century problems usually involves large amount of data emerging from dynamic complex systems, whose solution increasingly relies on bioinspired algorithms such as artificial neural networks (ANN) implementing machine learning (ML). However, the size and complexity of ANN are limited by the availability of computing resources. Future advances in scale of computations, data capacity, and energy efficiency may be driven by identifying ways of encoding and processing data based more closely on the task at hand. Inspiration may be gained by looking at natural (biological) systems from an information perspective. I will survey a broad selection of information processing and computational techniques inspired by biology (of which ANN is an example). Biological information processing is not confined by the limitations of digital representations but instead exploits a “dual-code” comprising an interplay of continuous and discrete strategies optimised by adaptation to environments. Future applications will include methods of computation where the hardware and software merge in new material substrates using or mimicking biological architectures, or hybrid electronic/biological architectures, such as real neuron chips. The mechanisms of quantum processes in biology may offer insights into implementing artificial quantum technologies.

Sonia Contera is a Professor of Biological Physics at the Physics Department of the University of Oxford. Her work is at the interface of physics, biology, and nanotechnology, and is interested in how biological “shape” relates to information processing. Here is her website: https://www.physics.ox.ac.uk/our-people/antoranzcontera