Artificial intelligence is rapidly growing, but despite all brute computation power, our brain convincingly wins in a number of tasks. It recognizes and interprets complex patterns very fast, consuming a minimum of energy. What can we learn from the brain’s structure and functions, when designing next generation computers? Scientists of the University of Twente, from a wide range of disciplines, join forces in the new Center for Brain-Inspired Nano Systems (BRAINS), started March 27. There will also be a new Master specialization in neuromorphic computing as well.
Computers cannot be beaten anymore, in playing a complex board game like Go. They get better and better in recognizing faces and speech. They are typically very strong in doing many parallel logic operations at high speeds. Our brain, in turn, is capable of recognizing and interpreting patterns in a very short time, using an amount of energy that is truly neglectable compared to that of computers. Apart from that, the brain develops itself, is plastic and learns. Artificial intelligence is often compared to the brain’s functionality, but it will really get powerful if it is actually capable of mimicking the brain’s unique way of working. The new BRAINS Center will combine knowledge of nanotechnology, materials, AI-systems, deep learning, chip design and neurosciences for developing new brain-inspired systems. The research includes the societal and ethical issues that may play a role in the new technology.
One of the recent developments at the University of Twente is the design of ‘evolutionary electronics’: this doesn’t have strictly defined functionality but develops over time. One of the problems that current electronics is facing, is the classical separation of the memory and the processing power. This means a continuous transport of data to and from the processor. Our brain, with its neurons and synapses, operates in a different way, in which storage, processing and transport of information are integrated. An ‘artificial synapse’ that can do more or less the same is the so-called memristor, a resistor that works as a memory element at the same time. This is one of the promising components of a brain-inspired computer. The UT researchers also consider using materials with new properties or atomic networks. For scaling up to fully functioning networks, a lot can be learned from the production of the current generation of CMOS chips. Intel, for example, already introduced a neuromorphic processor called Loihi.
BRAINS is a collaborative effort of three faculties (TNW, EEMCS, BMS) and two research institutes (MESA+ Institute, Digital Society Institute) of the University of Twente.
Apart from initiating groundbreaking research, the Center also introduces a Master specialization in neuromorphic computing, for Electrical Engineering and Applied Physics Master students. For more information about this, contact Brigitte Tel: firstname.lastname@example.org