###Developing AI computing hardware with a magnet
Project ID: 2228bd1059 (You will need this ID for your application)
Research Theme: Advanced Materials
UCL Lead department: London Centre for Nanotechnology (LCN)
Lead Supervisor: Hidekazu Kurebayashi
Project Summary:
Why this research is important? The energy consumption of current computing systems is sky-rocketing in order to perform complex artificial intelligence (AI) applications that are becoming a part of our lives. For example, DeepMind’s AlphaGo is operated by five thousand parallel processing units, consuming approximately 1 MW to run and much more to train. This is many orders of magnitude greater than the energy consumption of the human brain (~20 W). This difference is essentially due to a consequence of digital computing systems, known as the von Neumann bottleneck, where information data are shuttled between different units consuming energy each time, whereas a human brain performs analog computation much more locally and in parallel. This project will develop AI hardware components by using the physical properties of magnets, and establish foundation of designing/creating efficient hardware devices for machine learning algorithm. We have successfully demonstrated first steps of this ambitious project by creating spin-based physical reservoirs with nano-arrays of magnets [1] as well as magnetic skyrmions [2]. [1] J. C. Gartside et al., Nat. Nanotech. 17, 460 (2022). [2] O. Lee et al., arXiv:2209.06962.
Who you will be working with? You will be working with UCL spintronics group (https://www.ucl.ac.uk/spintronics) headed by Prof. Kurebayashi, based in London Centre for Nanotechnology. This project will have close collaborators, Profs. Tony Kenyon (UCL Engineering) and Petros Dellaportas (UCL Statistical Science)
What you will be doing? You will be developing a novel physical reservoir for AI computing. You will fabricate spintronic prototype devices to perform experiments, examine their computational performance as well as their physical properties, and analyse them for generating new knowledge and values.
Who we are looking for? Ideal candidates to take up this project would be students from either physics, materials science, electronic engineering or computer science background.