Academic Awards 2024 booklet

95 Organic neuromorphic computing: at the interface with bioelectronics Traditional computers excel at high-precision calculations, tasks that challenge human capabilities. However, there is a growing demand for technologies that operate more like humans: processing unstructured and noisy data in real time and adapting to change. These tasks can be realized by Artificial Intelligence (AI) using large artificial neural networks. While powerful, these software-based learning algorithms are energy- inefficient. In contrast, our brain is incredibly efficient at similar tasks. Neuromorphic computing aims to emulate several crucial aspects of the brain to efficiently perform AI tasks at the hardware level (see Figure 1). While various technologies and materials are being explored for neuromorphic computing, this thesis focuses on organic materials, which offer significant advantages for biological applications. First, we investigated and developed various organic materials as building blocks for neuromorphic systems. Using these building blocks, we created a modular neuromorphic system that could replicate a neural pathway based on artificial neurons and synapses. Furthermore, we developed a smart biosensor that was able to learn by itself to detect a disease from sweat samples (see Figure 2). Lastly, we designed an algorithm that enables training of large neural networks in hardware and thereby significantly improve the energy-efficiency. By leveraging the unique properties of organic materials, we can create bioelectronic devices that are more efficient, adaptive, and capable of integrating seamlessly with biological systems. Figure 1: Schematic representation of a) a biological neural network as in the brain, b) an artificial software neural network used in digital computers and, c) an artificial hardware neural network used in neuromorphic (brain-inspired) computing. Figure 2: Illustration of the neuromorphic biosensor.