Exploring the applications of CogniGron's revolutionary technology

CogniGron's work in the field of neuromorphic computing, including its first startup, IMChip, has the potential to transform various industries and society as a whole. In the second part of our series about the Groningen Cognitive Systems and Materials Center, we will explore the applications of its revolutionary technology and its potential social impact.

_{Text: Maud Brongers / Photos: Henk Veenstra}

IMChip: Combining current and new technologies

Prof. Tamalika Banerjee, Professor of Physics and Spintronics, introduces IMChip, CogniGron’s first start-up in progress, which stands for In-Memory Chip. In the first part of this series, CogniGron’s researchers explained that traditional computers consume most of their energy through data transfer, a problem that the brain architecture naturally avoids. IMChip aims to revolutionize neuromorphic hardware by making in-memory chips, in which memory and processing take place in the same location.

This technology could be applied in industries including automotive, mobile phones, healthcare, security, and AI. Think of medical applications and self-driving cars — they demand instantaneous decision-making without any delays. This is where neuromorphic computing comes into play. ‘Neuromorphic computing holds a significant advantage over traditional architectures,’ Banerjee explains. ‘It’s not just about in-memory computing; it encompasses communication, data security, and more. This technology offers enhanced capabilities in areas such as storage and encryption.’

Banerjee emphasizes the need for a combination of different materials that can function as a network of neurons, sending and receiving information, and the synapses in between that transmit the information. This approach requires tailored solutions rather than a one-size-fits-all model. Banerjee works with a talented team of PhD students: Dr Anouk Goossens, Azminul Jaman, and Ishitro Bhaduri. She describes how her team is open-minded about new possibilities for future applications. Banerjee emphasizes that she learns from established technologies and embraces what is already known in conventional approaches.

Dream applications

Dr Farhad Merchant, Assistant Professor of Innovative Computer Architecture at the UG’s Bernoulli Institute, also works towards possible applications of neuromorphic computing. He joined CogniGron in July of this year. Eager to explore startup opportunities, Merchant expresses his passion for prototyping and industry-aligned research, which can provide invaluable experience for his PhD students. His research is driven by a desire to create tangible technologies. He has ideas for innovative applications, such as safety drones that monitor children and alert parents or emergency services when needed, or drones in the agricultural sector that monitor crops. Another interesting application could be a bicycle that tracks health metrics such as blood pressure and breathing patterns. According to Merchant, these advancements promise not only improved security and energy efficiency but also a reduced carbon footprint.

Two types of security

Merchant’s research focuses on computing memory and hardware security. In the case of computing memory, he tries to bring computation closer to the computers’ memory and to bridge the gap between the two, minimize data movement, and achieve better energy efficiency and improved performance. ‘For computing memory, the goals are set,’ Merchant says, ‘but when it comes to security issues, your goal depends on attack vectors, meaning that countermeasures are diverse.’ The researcher explains that improving security contradicts the goals of chip design: to improve security, you have to compromise a bit in your performance and energy efficiency.

In addition, Merchant describes two types of security. When you try to secure your chips at the time of design, it is called ‘secure by design’. Additionally, there is ‘design for security’, in which you use electronics to improve the security of your chips. His work is crucial as it lays a foundation for creating more secure, energy-efficient, and high-performance computing systems, which are essential for advancing neuromorphic technology.

Supercomputers and big data problems

Dr Dirk Pleiter, Professor at the Faculty of Science and Engineering, explains that supercomputers are essential in advancing the development and uptake of neuromorphic devices. He joined CogniGron in October of this year and brings his expertise in high-performance computing (HPC) and computer architecture to the team. HPC uses supercomputers to solve advanced computation problems. These computers are able to produce complex simulations in areas such as weather forecasting and climate research and thus require a lot of computing power.

Pleiter compares the supercomputer to a large group of people building a house: ‘Adding more workers does not necessarily mean faster construction, as communication takes time and can slow progress. It takes approximately one microsecond to communicate a message from one server to another. This does not sound like much, but in computing, even microseconds can significantly impact performance.’ This is one of the factors Pleiter takes into account in his research, where he investigates ways to build fast computers in which many calculations or processes are carried out simultaneously, also known as parallel computing.

Pleiter’s affinity with neuromorphic computing came about during his work on the Human Brain Project, which has been one of the biggest research projects ever supported by the European Union. This project pioneered in using big data and supercomputing to simulate complicated functions of the brain. Pleiter describes how this project highlighted the complexities of simulating human brain function, where creating realistic models of neurons and synapses presents considerable challenges.

Creating a new ecosystem

Developing the hardware of a neuromorphic chip or computer is only one aspect; effectively launching this technology and integrating software is another challenge altogether, says Pleiter. He aims to establish computing systems to enhance their accessibility. The process of establishing new computing paradigms takes a lot of time due to its complexity, but also because of the required extensive testing of software programs. In order to make a neuromorphic computer architecture successful, an ecosystem of engineers is needed to create the software, the researcher knows.

He ultimately hopes to enhance computational power to improve predictions not only about weather and climate change but also in health and medicine. Pleiter illustrates how these advancements could, for instance, accelerate research into complex brain disorders such as Alzheimer’s, which remain poorly understood. Moreover, the COVID-19 pandemic underscored just how essential high-performance computing is for the development of medicine, highlighting its transformative potential in medical research.

Radical change

CogniGron’s work in neuromorphic computing, including initiatives like IMChip, has the potential to transform various industries and society as a whole. Together with the CogniGron team, Banerjee, Merchant, and Pleiter are not only pushing the boundaries of technology but are also committed to creating solutions that benefit society, making CogniGron a leader in the field of computing innovation. ‘We need this radical technological change to move forward,’ concludes Pleiter.

CogniGron was founded with a significant donation from the Ubbo Emmius Foundation (UEF).

More information

• See also: 'CogniGron: A revolution in future-proof computing'
• Tamalika Banerjee
• Farhad Merchant
• Dirk Pleiter
• CogniGron