Learning to save energy from the brain – information research department compass

Copyright: Pixabay / Gerd Altmann

Scientists from the Jülich Research Center of the Helmholtz Research Department of Information and the RWTH University in Aachen want to work with companies in the region to create a leading international location for neuromorphic AI equipment. (Source: Jülich Research Center – press release)

Building a technological base for neuroscience-inspired AI equipment from Europe – this is the goal of the future NeuroSys cluster and the NEUROTEC project in which Forschungszentrum Jülich and RWTH Aachen University are working together. A leading international neuromorphic equipment site is to be built in the Greater Jülich-Aachen area in collaboration with high-tech companies located in the region – computer chips modeled on the human brain. Following the meeting at the Jülich-Aachen Neuromorphic Computers Day on May 24, 2022 at the Forschungszentrum Jülich, scientists now provide an insight into the current state of the projects.

NEUROTEC entered its second phase at the end of 2021 and will receive a total funding of around € 36 million over five years from the Federal Ministry of Education and Research (BMBF). The future cluster “NeuroSys – neuromorphic equipment for autonomous artificial intelligence systems” won last year’s Clusters4Future ideas competition and is funded by the BMBF with up to 45 million euros.

In an interview with prof. Rainer Waser from Forschungszentrum Jülich and RWTH Aachen, coordinator of NEUROTEC (Neuro-inspired Technology of Artificial Intelligence for Future Electronics) explains together with prof. Max Christian Lemme from RWTH Aachen and AMO GmbH, coordinator of the future of NeuroSys cluster (Neuromorphic Hardware for Autonomous Systems of Artificial Intelligence), current state of development.

Prof. Rainer Waser, why exactly such neuromorphic computer chips, based on a model of the human brain?

At NEUROTEC, we mean a promising area, namely equipment for artificial intelligence applications. With our approach, we tackle a very basic problem, the energy problem: because the use of AI is still very energy intensive today. Models are typically trained on supercomputers and require more and more computing time. The computational effort doubles every 3-4 months, at least this is the trend in recent years.

Neuromorphic systems with artificial synapses hold the hope of solving these tasks much more efficiently – by several orders of magnitude – than is possible with conventional digital computers. In the long term, many applications can be imagined: from the smallest nanosensors, “intelligent dust”, to intelligent implants with energy self-sufficient AI control, pattern recognition chips in mobile devices, online control in Industry 4.0, vehicle AI-based electronics for autonomous driving to to mainframe systems, which in turn emulate the brain or solve complex artificial intelligence tasks centrally for the surrounding economy.

NEUROTEC is funded under the Crisis Program for Structural Change. It aims to help create new jobs in the Rhenish mining area before lignite mining is completed. What exactly do you mean?

We plan to support the local industry, especially in the area of ​​basic technologies. The professional sectors we cover extend far beyond the mere production of chips. This also includes plant construction, measurement technology and the development of electronics. The NEUROTEC II project phase started very well from November 2021 and almost all work packages are on schedule. Some work packages are even ahead of their time.

The technology is currently still in the research and development phase. Nevertheless, the first concrete successes have already been achieved. Of course, these are initially primarily jobs for specialized, skilled workers, before later, as technology matures, the focus on industrial production will increase when the technology is applied on a large scale. Current example: During the Neuromorphic Calculation Day in Jülich-Aachen, prof. Heuken from AIXTRON SE showed what prospects are opening up in the next few years for new jobs at AIXTRON due to new 2D materials. These materials are being studied by the NEUROTEC project for their suitability for neuromorphic calculations.

What are the advantages of this neuromorphic hardware that you are developing?

Conventional computer chips are based on transistors. We would like to add a new type of memristic element to these transistors. Such “memory immunity” is similar to synapses in natural nerve cells and is therefore particularly suitable for artificial neural networks such as those used in artificial intelligence applications.

An important feature of neuromorphic chips is that computational processes and information storage are no longer physically separated. The data transfer between the processor and memory, which takes place continuously on conventional computers with the so-called Von Neumann computer architecture, is extremely energy-intensive and slows down computation. In contrast, we envision neuromorphic chips with artificial synapses that can simultaneously store information and process it. The computation is then performed directly in – non-volatile – memory, which is also referred to as in-memory computation. And it enables information to be processed in a highly parallel manner. The model here is actually the human brain, which requires an average of just 20 watts. This is several orders of magnitude below the energy requirements of a supercomputer that performs similar functions using artificial intelligence.

Prof. Max Christian Lemme, in the NeuroSys Future Cluster, you lead the development of neuromorphic hardware in a complementary way. This topic is also intensively developed around the world. What is the position of the region in relation to global competition?

The BMBF Future Cluster Initiative has the clear goal of rapidly transferring excellent, cutting-edge research to applications, and in regional networks. To achieve this, in the future NeuroSys cluster, we have brought together scientists from across the value chain, each with global visibility in their discipline. There are also regional companies and start-ups from the high-tech sector. We now cover the entire scope, from new materials to image and speech processing and medical technology, and extend the concept to socio-economic questions – i.e. research into AI ethics, labor market implications and possible business models for ‘AI equipment Made in Europe “.

The cluster is complemented by an advisory board composed of scientists and international corporations. So I dare say we are on par with global competition. The region lacks a modern semiconductor factory for AI chips, the location of which we have formulated in our vision. Thanks to the well-trained engineers and scientists in the region, including Belgium, the Netherlands and the Ruhr, the proximity to the IMEC research institute in Belgium and the leading manufacturer of ASML lithographic systems in the Netherlands, we have excellent arguments.

How far is the technology?

As is often the case, there is no easy answer. It is already possible to manufacture special neuromorphic chips using conventional technology. However, the energy efficiency of the brain is still a long way off. Here, new technologies made of resistive switching oxides, phase change materials and even 2D materials can take us much further. However, their use is always a question of industrial production. This varies from material to material, and now it looks like each generation of new materials will also bring an increase in performance. Therefore, it is very important to work closely with manufacturers of material separation systems such as Aixtron from Herzogenrath or deep tech start-ups such as Black Semiconductor or Aixscale Photonics on both projects.

At the same time, the future cluster also cooperates with industry at higher levels of the value chain. There are several start-ups in the region, such as Clinomic, Gremse-IT or the company that my colleague from RWTH Aachen, prof. Rainer Leupers, he founded NeuroSys shortly after launching. Therefore, we are already working on achieving our goals at all technology levels and are still in the acceleration phase after a perfect flight start!

The original press release can be found at:

Learning to save energy from the brain

Location within the Helmholtz Research Area Information:

Helmholtz Research Area Information, Program 2: Natural, Artificial and Cognitive Information Processing, Theme 3: Neuromorphic Computation and Network Dynamics

Contact:

Prof. Dr. Rainer Waser
NEUROTEC coordinator, head of the Peter Grünberg Institute of Electronic Materials (PGI-7)
Julich research center
Telephone: +49 2461 61-5811
E-mail: r.waser@fz-juelich.de

Prof. Dr. Eng. Max C. Lemme
NeuroSys Coordinator, Department of Electronic Components
RWTH Aachen
Telephone: +49 241 80-20280
E-mail: neurosys@eld.rwth-aachen.de

Contact for this press release:

Dr. Regina Pankin
Press officer
Julich research center
Telephone: +49 2461 61-9054
E-mail: r.panknin@fz-juelich.de

Leave a Comment