An artificial intelligence technology that speeds up the development of advanced materials, from years to a single day, has won the £1 million Manchester Prize.
Polaron, an Imperial College London spinoff, uses state-of-the-art generative AI and microstructural image data in materials development. The prize they’ve won is funded by the UK government and designed to support AI breakthroughs for the public good.
The money will enable the Polaron team to continue developing AI tools to accelerate material development for applications such as high-performance alloys and batteries.
Essentially, this technology will enable the production of stronger, lighter, and more efficient materials and components for use in key infrastructure, clean energy and transport. This work will help to support the UK’s net-zero goals.
Standard material design involves engineering instinct, mixed in with experience of what does and doesn’t work. It’s a “trial and error” approach that can take years. With Polaron, the algorithms learn the microscopic structure of a material from image data. This allows an incredibly fast exploration of possible designs.
Polaron was originally set up in 2024 by a team in the Dyson School of Design and Engineering at Imperial, working closely with Imperial Enterprise. They are now working with their first client in the electric vehicle and battery space. And they’ve demonstrated a 10% improvement in the energy density of batteries, which adds roughly 20 miles of range to an electric vehicle.
Dr Isaac Squires, the third co-founder and chief executive of Polaron, says this EV work is just the beginning. “Polaron is material agnostic, and we are already bringing our rapid design capabilities to industrial manufacturing more widely.”
It’s fascinating to see that AI is being used to innovate in this way, and I’m curious as to how this could ultimately benefit our industry. I’m also concerned about any potential downsides to this approach. Clearly, it’s going to save time, but you’ll still need human input to decide which materials to test.
As an example from my own work, I’m currently using the paid version of ChatGPT, and it has definitely improved over time, but I wouldn’t want to use artificial intelligence to design a structure. AI wouldn’t know whether a design works in practice or not. I believe AI can be used as a tool and can be beneficial, but it cannot fully replace people or skills, as it’s not yet good enough. What do you think?
Meanwhile, if you need assistance with the structural elements of an upcoming project, please do get in touch.
