The Silicon Brain Cube is a groundbreaking 3D modular hardware architecture designed to improve energy efficiency for demanding AI workloads and support reliable processing.

This joint research project leverages the Japanese team’s expertise in energy-efficient deep learning models and 3D hardware architectures, alongside the UK team’s strengths in trustworthy AI design and multi-level static and dynamic optimization tools.

The goal is to innovate AI models that achieve optimal trade-offs among user-defined performance, required resources, energy efficiency, prediction accuracy, and trustworthiness. Improved energy efficiency will be achieved through strategies such as reducing memory access and enabling irregular/sparse processing on reconfigurable in-memory computing fabrics.