The University of Liège sets up a neuromorphic engineering laboratory


In Research

Housed within the Montefiore Institute of the School of Engineering at the University of Liège, the new neuro-morphic engineering laboratory brings together the skills of engineers, computer scientists, and biologists interested in developing electronic components that will make it possible to reproduce the activities of a biological brain, while reducing their energy impact. These technologies could help human beings and substantially reduce our energy consumption. Here's how it works.

I

n March 2016, a Google supercomputer programme, AlphaGo, beat Lee Sedol, master of the game of Go. A feat considered at the time to be one of the most important milestones in modern artificial intelligence (AI). While AlphaGo consumed around 1 megawatt of energy to concentrate strictly on the specific task in hand, Lee Sedol's brain used 50,000 times less energy (around 20 watts), not only to concentrate on the game but also to maintain its sensory, motor, and vital functions. Similarly, modern robotic systems are capable of solving heavy tasks - often far beyond human capabilities - but these machines consume a lot of energy, are rigidly optimised to perform specific tasks, and can hardly cope with unexpected events.   

This observation has given rise to neuromorphic engineering, a new discipline that combines neuroscience, computer science, and engineering and seeks to design and develop computer systems capable of performing tasks such as perception, cognition and decision-making in a way that is close to that of the brain, while limiting the energy impact. This is a major challenge for scientists, who see this new discipline as a real opportunity to design hardware and develop algorithms and applications in various fields such as robotics, artificial intelligence, and neuroscience. It's a technology full of promise but needs to be tamed first because the stakes are so high and the risks so great.    

“We're in the early stages," explains Alessio Franci, a new lecturer at ULiège and a specialist in the field. Neuroengineering is in the process of taking off and developing in several universities around the world. However, we are still a long way from the technologies used in machine learning. Getting into neuro-engineering is risky because there is a demand and enormous potential, but we are still only at the concept stage. Despite the risks and the strength of its expertise in the field of artificial intelligence, the University of Liège has chosen to rise to the challenge by setting up a neuro-morphic engineering research centre. This laboratory, housed within the School of Engineering, is run by Guillaume Drion, Alessio Franci, Jean-Michel Redouté, and Pierre Sacré, four professors and researchers renowned for their expertise in the field of AI and neuro-engineering. Doctoral theses, such as those of Loris Mendolia, Kathleen Jacquerie and Antoine Debor, assistants in the Department of Electrical and Computer Engineering, are already looking at the development and practical application of neuro-morphic engineering. We are proud and enthusiastic that our university allows us to develop such a laboratory, despite the risks involved," adds Guillaume Drion. We've accepted a risk that will enable us to position ULiège at the world level and compete with some very big institutions because we'll be one of the few laboratories attempting to develop the technology.   

Software, hardware, wetware   

Combining teaching and research - two of the three fundamental pillars of universities - the laboratory will welcome students and young scientists to develop the discipline, supervised by their professors. Their research will focus on three main areas: software, hardware (electronics), and wetware.  

Adaptive robotics will enable the development of neuro-inspired algorithms for fine, adaptive manipulation, like that seen in humans. To achieve this, the laboratory will equip itself with a robotic arm featuring an anthropomorphic hand with sensory sensors to study grasping and movement management.   

The researchers will also be looking at electronics and signal processing in order to implement neuromorphic intelligence at the level of electronics. We're talking here about going down to the level of the transistor - the microscopic components at the base of modern integrated circuits, which control and/or amplify electrical currents - and processing signals in an analogue way," explains Jean-Michel Redouté. This is one of the major challenges of neuromorphic engineering, which aims to move away from the structure of conventional computers and automata."   

Finally, the laboratory will have a wet lab, enabling us to test the interactions between electronics and biological tissue. We will be working on Brain on Chip, a technology developed by IMEC - a pioneering Belgian research and innovation centre recognised as one of the world's leading R&D centres for nanoelectronics and digital technologies," says Pierre Sacré. This tool will enable us to develop artificial tissues on very dense multi-electrode recording matrices to develop biological neuron networks that can record and stimulate each neuron separately". To achieve this, the laboratory will work closely with GIGA Stems Cells, headed by Laurent Nguyen, a ULiège laboratory specialising in neurogenesis. This collaboration will enable them to understand better the plasticity of neurons and how they structure themselves.  

Assisting people, not replacing them  

When we talk about AI, the question always arises about the development of such technologies at a societal level. The contribution of this discipline to the future of mankind remains a real challenge. It is a real challenge to understand and translate the functioning of a biological brain into robotics scientifically and technologically. Very concrete applications, such as developing intelligent and adaptive prostheses for people who have lost a limb, or reducing the energy and physical load of onboard equipment for space missions, are just two of the many examples of what mastering this discipline could offer. We're not talking about competing with AI software such as ChatGPT," adds Guillaume Drion. "We aim to develop technology that is much closer to human intelligence so that we can interact with humans in a highly integrated and resilient way.   

After receiving federal funding to set it up, the laboratory hopes eventually to become completely autonomous in terms of its management. With industry giants such as Sony, Samsung, and Intel working on the development of this technology, the University of Liège is clearly positioning itself as a major player in this change by working on the development of systems capable of learning, adapting, and performing complex tasks with great energy efficiency, making them more suitable for applications in fields where traditional computing approaches remain limited.   

Your contacts at ULiège  

Guillaume Drion 

Alessio Franci 

Pierre Sacré 

Jean-Michel Redouté 

Loris Mendolia 

Kathleen Jacquerie 

Antoine Debor


*The term 'neuromorphic' combines 'neuro' - meaning related to the nervous system - and 'morphic' - meaning form - indicating the aim of creating systems that mimic the morphology and behaviour of neural networks.  

Image: Shutterstock


ODDs
ODD 3 : Bonne santé et bien-être
Donner aux individus les moyens de vivre une vie saine et promouvoir le bien-être à tous les âges Le troisième objectif vise à assurer la santé et le bien-être de tous, en améliorant la santé procréative, maternelle et infantile, en réduisant les principales maladies transmissibles, non transmissibles, environnementales et mentales.

Share this news