Dr Jack C. Gartside - Imperial College London

The energy cost of modern AI and machine learning is rapidly spiralling out of control. Computing-associated energy costs are forecast to reach 20.9% of total global energy consumption by 2030, and a recent White Paper by OpenAI shows that the energy consumption of machine learning AI is doubling every 3.4 months. This is massively unsustainable and a critical roadblock for the crucial goal of reaching zero-carbon.

By harnessing the complex Gigahertz dynamics of nanoscale magnetic arrays, Dr Gartside’s research implements neural-network-style machine learning directly in hardware – potentially dramatically reducing the energy and carbon footprint of AI.