With the computers we have now, scientists and companies can perform complex calculations based on vast amounts of data. However, this requires a tremendous amount of energy and cooling water. A constantly increasing demand for capacity therefore poses a direct problem in terms of sustainability. Moreover, the costs for the associated infrastructure can only be met by big tech companies, on which we are becoming increasingly dependent. And then there are security and privacy requirements that need to be better addressed than is currently the case.
Faster computers that consume less energy
That means it’s high time to take a completely different approach: neuromorphic computing. It involves computers and computer chips inspired by the functioning of the human brain: with less energy and more effectiveness. Mentink: “Computers can do many things better than humans, who cannot remember and reproduce everything and, moreover, sometimes need to sleep. But take facial recognition: humans can do that very quickly and it consumes much less energy than a traditional data centre.”
Ideas about neuromorphic computing have been around for several decades. We are now approaching applications, and they are promising, says Mentink. ”With neuromorphic computer chips, which are already being manufactured, it’s possible to perform certain calculations faster and with less energy. We are currently investigating which chips are best suited to which applications. This also gives companies insight into how the transition to neuromorphic chips and computers can add value. We’re also working on new materials for neuromorphic chips that can reduce energy consumption even further.”
Possible applications
Neuromorphic computing seems suitable for applications involving large amounts of data. It also offers possibilities in so-called edge computing, where, for example, a lot of data processing is done close to a sensor. Mentink cites examples such as a hearing aid gaining new functionalities without the battery constantly running out. Or greenhouses where you can automatically measure which plant needs what care at what time. Applications that currently often consume too much energy, but which could become possible with neuromorphic chips.
National coalition for neuromorphic computing
The great thing, says Mentink, is that Radboud University is also conducting fundamental research that will enable even better neuromorphic computers to be developed in the future. To improve knowledge transfer, various universities, companies and (semi-)governmental organisations are working together on an ecosystem and national coalition for neuromorphic computing. In November 2024, they published a white paper calling for such a national approach. ‘Together with the Top Sector ICT, we are trying to influence politicians to take a strategic approach to this. Sustainability, autonomy and security are the most important arguments. In addition, we believe that it offers opportunities for scientists and entrepreneurs if the Netherlands continues to be at the forefront in this field.’
‘One of the aspects we would like to develop further is an application lab, in which universities and companies work together on a kind of overview map showing which neuromorphic computers are most suitable for which applications. The sooner we have that picture, the sooner the added value of neuromorphic computing will become apparent.’
Want to participate? Sign up!
Companies or organisations that believe neuromorphic computing can help them with their complex data issues should get in touch with Johan Mentink. ‘My colleagues and I are working extensively on applications for scientific calculations, but a great deal of calculations are also performed outside the university. We would like to test and further develop our work on the basis of real-world applications. Anyone who would like to participate is very welcome!’