Rethinking computing: what can we learn from nature?

Nature has been solving complex problems with remarkable efficiency for millions of years. Could modern computing learn from it?

As AI drives demand for ever more computing power, the energy, water and infrastructure needed to support it are becoming harder to ignore. Researchers at Imperial are now exploring whether radically different approaches to computation could offer a more sustainable way forward. 

Against this backdrop, Imperial’s School of Convergence Science brought together leading researchers, innovators and industry experts for a panel discussion titled Nature-inspired computation: real-world challenges and opportunities, as part of the two-day workshop, Computing with Waves, held at the South Kensington Campus over 21 and 22 September. 

The wider workshop brought together researchers across photonics, electromagnetics, mechanics, solid-state systems and related fields to explore new approaches to computing and information processing using physical waves.  

The closing panel session, which was led by Professor Will Branford, Co-Director of the School of Convergence Science, asked: rather than simply making existing computing systems faster and more powerful, could we rethink how computation works for a more sustainable future - and what might nature teach us? 

The focus of the workshop, and the panel discussion, was primarily on new hardware paradigms. 

Session speakers 

A diverse panel brought cross-disciplinary research and fields of academia together with the technology industry: 

• Professor Will Branford (Chair) - Co-Director of the School of Convergence Science at Imperial. Will’s research is geared towards exploring the potential of new methods of computation such as neuromorphic computation and magnonics. His Imperial inaugural lecture was titled ‘How physics made AI possible, and how it can make it more efficient’.
• Hitesh Ballani - Partner Research Manager at Microsoft Research Cambridge. Hitesh’s research group is working on the future of AI infrastructure, exploring processes and materials that can accelerate computation and make it more efficient. He works with both startups and large manufacturers. 
• Professor Marco Di Renzo - Professor of Telecommunications Engineering at King’s College London. His research interests include exploring the potential to reduce power consumption of today’s compute systems using new technologies. 
• Dr Jack Gartside - Assistant Professor in Neuromorphic Computing and Metamaterials at Imperial, where he also leads the Neuromorphic Metamaterials group.
• Ella Shuter - Junior Programme Manager for Emerging Technologies at techUK, a trade association for the UK technology sector, representing over 1,100 members, from startups and SMEs through to the largest global technology companies.

Sustainability, performance and a growing gap

Imperial researchers as well as industry experts are pursuing a radical shift in how we approach computing. Why?  

Hitesh Ballani explained: 

“In recent years it has become clear that we are very close to the limit of compute power that we can get efficiently from the current compute infrastructure that we have - any sustainable increase of power will have to come from a brand-new technology.”

Concerns around sustainability are exposing the limitations of current computing, paired with the cost of maintaining infrastructure that has reached its limits.  

The challenge now is a growing gap between the demands for greater computing power and the ability of existing infrastructure to deliver it affordably and sustainably.  

Without a fundamental upgrade in compute technology, it is difficult to increase capacity, manage costs and improve sustainability beyond the marginal gains of existing approaches.  

A nature-inspired shift 

Amidst a race for greater performance and a growing need for sustainability, computer science is increasingly drawing upon the wonders of biology for inspiration.

One such inspiration is the brain - arguably one of nature’s best examples of ingenuity, having evolved to learn, remember and adapt to constant change, while operating with remarkably limited energy and resources.

The human brain assigns different regions a unique function. Each area has its own structure, design and approach to processing information, working together as part of a highly specialised and interconnected system to produce coherent and intelligent behaviour. 

Professor Marco di Renzo referred to plants as another example of nature’s computational sophistication - their intricate networks distribute water and nutrients, respond to their environment and make the most of limited resources - without a central processor or significant energy consumption. 

These principles are already inspiring new computing technologies. Neuromorphic chips, for example, take inspiration from the brain by processing information in parallel and using networks of connections to perform computing in ways that can be more energy efficient for certain tasks.

The rise of parallel processors such as GPUs has also helped drive the recent growth of AI, and researchers are now asking whether going further and incorporating more brain-inspired principles could deliver another step change in performance and efficiency.

The challenge is turning these biological principles into practical technology. If the next generation of AI is to learn from nature, it may require us not only to rethink how we build computers, but also how we bring together the disciplines and expertise needed to create them.

Why convergence science matters  

In exploring how the computing landscape is likely to change over the next decade, in alignment with modern compute requirements, the panel observed that new technology is demanding research that defies conventional approaches.

Ella Shuter observed:  

“We’re seeing a more convergent, cross-technology approach emerging for future technologies. We shouldn’t think of this as replacing technology, but as creating an array of technologies combining to work together.  A heterogenous future for computing technologies.” 

Hitesh Ballani added:  

“We need to step away from the siloed mindset that only one technology can win. It’s time to support and embrace the convergence and integration of multiple technologies.” 

For nature-inspired computing to be developed further, historically siloed disciplines, sectors and partners will need opportunities to come together and cross-pollinate, so the effectiveness of new ideas can be demonstrated in the real world.  

From promising research to real-world impact  

Cutting-edge research does not automatically translate into societal impact and there are key challenges to realising new technologies.  

The panellists reflected on skills, manufacturing, investment infrastructure and regulation as some of the key barriers in the UK.  

Skills wise, commercialisation hinges upon researchers being able to bring their products to market in the first place and knowing the various routes in.

Dr Jack Gartside explained a key dealbreaker in commercial viability is being able to demonstrate a technology’s capabilities rather than its potential, including being ready with initial-use cases to prove real-world readiness.

Beyond securing investment, Ella Shuter advocated for knowing your investor well:

“It’s just as important to ensure an investor has a deep understanding of your technology in order to help, develop, grow and deploy it."

Achieving commercialisation effectively also requires wider infrastructure and a national drive.  

Ella Shuter recognised fragmentation of the UK technology and innovation ecosystem as a pinch point.  

Dr Jack Gartside added: 

“Deep tech in Europe benefits from much existing support, both in terms of investment and also in accelerator programmes and facilities – take pilot lines where researchers or SMEs can scale up their chips for production runs. Whereas in the UK we’ve a real gap between laboratory innovation and industrial-scale manufacturing, and financial support for startups that want to scale high tech manufacturing beyond the initial university spinout programme.”

Manufacturing was recognised as a bottleneck in the UK, with often unviable costs involved in large-scale industrial production, preventing research breakthroughs from reaching society. 

As for regulation, Ella Shuter noted that the Regulatory Innovation Office has recently been set up to help the UK reach its full potential in innovation.  

Harnessing the strengths of UK universities  

Hitesh Ballani remarked that as computing approaches the limits of existing technologies, universities can provide a space to explore ideas beyond the immediate commercial horizon – helping industry understand what might become possible next.  

He explained how large tech companies act as 'hyperscalers' of new technology, once it is ready to be deployed. 

Collaboration between universities, startups and industry can therefore create a pathway from long-term research to real-world application.  While universities can explore what is possible, startups and industry can help develop, demonstrate and scale technologies for deployment.

Closing remarks 

As conventional approaches reach their limits, we need to rethink how computing is designed to meet growing demands for performance, cost, efficiency and sustainability.  

The current landscape holds much opportunity for different disciplines and sectors to come together and co-design technologies from the ground up, opening the door to entirely new approaches to computing.  

Demonstrating the application and efficiency of new technologies will also be critical to building confidence in their deployment and adoption, and this will warrant investment in enterprising infrastructure.

Universities have an important role to play in this, working alongside cross-sector partners to explore and develop what comes next.  

As Ella Shuter remarks:

“There is a lot of fragmentation in the wider technology and innovation ecosystem, so the conversation around convergence and collaboration is a great step in the right direction”.

If we are to realise the potential of more sustainable and powerful computing, there is untapped potential beyond conventional approaches for us to learn from the sophisticated systems that nature has been refining for millions of years.  

To close, Professor Will Branford reflected: 

“Realising the benefits of AI in a sustainable and safe way is one of the defining challenges of our times. Nature shows us the power of specialised systems co-evolved to efficiently solve complex tasks. There are great ideas for more sustainable devices, materials, algorithms, networks and comms, but widespread adoption will require a convergence of all this expertise to integrate these ideas with the existing computing stack.”

Powering research  

Imperial’s School of Convergence Science is an initiative of the university’s Science for Humanity strategy. The School is designed to enable the deep integration of disciplines and cross-sector partners, advancing knowledge to deliver societal impact.

The School collaborates globally with researchers, industry, governments, and funders to translate science into action, and is structured around four themes: Health and Technology, Human and Artificial Intelligence, Space, Security and Telecoms, and Sustainability.

Imperial Enterprise supports Imperial staff and students to realise the world-improving potential of their expertise through entrepreneurship and external partnerships, and helps industry partners access the university’s science and technology.