Biologically-Inspired Massively Parallel Architectures - computing beyond a million processors

The human brain runs on a biological architecture so efficient that it outperforms the world’s most powerful supercomputers at tasks like recognising a face—yet no one understands how it actually works at the intermediate levels of processing where its true information-processing power likely resides. This project aims to build computers that mimic the brain’s massively parallel structure, using models of very large systems of spiking neurons inspired by the latest neuroscience findings. The core problem is that current machines cannot simulate brain-like networks at the scale needed to study emergent behaviours, adaptability, and fault-tolerance. Without such machines, the intermediate layers of brain function remain a black box. If successful, the research will deliver machines of unprecedented cost-effectiveness for simulating brain-scale neural networks, making them accessible to a wide range of scientists. The unique architecture that emerges could also be applied to other demanding computational tasks—such as real-time sensor processing, robotics control, or pattern recognition in infrastructure monitoring—where conventional processors fall short. This is primarily fundamental science, driven by curiosity about how the brain works, but the computing principles it uncovers may quietly reshape how machines handle complex, real-world data.