The European Innovation Council (EIC) Pathfinder programme has awarded the funding to the ambitious project, which aims to develop more energy-efficient AI by utilising nonlinear physical systems. The consortium, led by Professor Marco Fanciulli at the University of Torino, includes scientists from Imperial and across Europe. Imperial's Department of Physics will receive approximately €660,000.
Dr Jack Gartside, who leads Imperial’s Neuromorphic Metamaterials group, expressed excitement about the grant, highlighting it as a rare opportunity to support international collaboration. "We’re thrilled to have the support of this EIC Pathfinder Grant," he stated. "It’s one of the very few ways to support an international team working together across Europe, and they’re hard to get – just 43 were funded out of over 2000 submitted."
KAN-DO-IT will investigate how nonlinear physical systems can perform machine-learning tasks directly, offering a potential alternative to conventional AI that demands significant energy and computing resources. Current AI systems often rely on Graphics Processing Units (GPUs), influencing architecture design.
Imperial's Neuromorphic Metamaterials group has been researching new AI forms based on physical systems' natural behaviour. Their recent work demonstrated a Kolmogorov-Arnold Networks (KANs) approach, which could drastically reduce neural network size and energy consumption, with projected savings of up to 600-fold in network size and a thousand-fold in energy for certain tasks.
The new project will build on these findings by developing larger networks, smaller devices, faster operations, and advanced learning algorithms. Imperial researchers will focus on developing physics-aware training methods and experimentally testing physical learning networks, aiming to establish a new class of learning systems for low-power, low-connectivity environments.
"The technology and approach at the heart of this project represents an exciting step towards turning exotic nonlinear physical dynamics into an industrially-scaled technology that can benefit a lot of people," said Dr Gartside. "We really believe in the core science, and we think that we’re just scratching the surface of how much nonlinear physics can offer neural networks beyond what GPUs and conventional AI are capable of."
Potential applications include adaptive health-monitoring wearables, neuroprosthetic devices, autonomous robotic systems, and data-efficient computer vision technologies. These devices could process information locally, enhancing privacy and reducing energy consumption compared to systems relying on remote data centres.
Jack Gartside added, "We want to free the benefits of AI from being owned and controlled by a tiny number of big tech companies. We want to keep data, privacy and autonomy in the hands of the end-user, rather than being shipped off to massive remote data centres. We’re really excited to get started. We’d like to thank the EIC and all our collaborators across Europe for their support and partnership in making it happen."
