Today, supercomputers are used for a wide range of tasks: from various mathematical calculations and data processing to modeling pharmaceutical compounds and artificial intelligence. However, there are computers aimed at the most accurate reproduction of the "architecture" of the human brain. And the most powerful such neuromorphic supercomputer to date was recently launched for the first time.
This computer is called Spiking Neural Network Architecture (SpiNNaker) and can perform more than 200 million operations per second, and each of the chips has 100 million parts. The machine is assembled at the School of Computer Science at the University of Manchester and spent £ 15 million on this project. But this amount is not as impressive as the time it took to develop: the concept of a supercomputer was worked out for 20 years and it took another 10 years to assemble.
SpiNNaker supercomputer microcircuit.
SpiNNaker can simulate the behavior of human neurons in real time and it is unique in that, unlike traditional computers, it does not transmit large amounts of information from point A to point B, but sends data simultaneously in thousands of different directions. This is how our brain cells work. According to one of the project's ideologists, professor of computer science Steve Furber,
It should be noted that even despite the fact that the "real artificial brain" has started working, the number of nuclei that it currently contains simulates the brain, which is equivalent to 1% of the human brain. However, already now, this neuromorphic supercomputer will help understand how our central nervous system functions, and will also provide the ability to conduct large-scale simulations that are not available on traditional machines.
Vladimir Kuznetsov