Performing hundreds of billions of calculations per second, while consuming a minimum of energy, is what the brain can do. The DeepSouth supercomputer is inspired by it to be able to accelerate the development of technologies and in particular artificial intelligence.
This will also interest you
[EN VIDÉO] Interview: How is a quantum computer different? The quantum world is fascinating: at this scale, for example, objects can…
Artificial intelligence (AI) algorithms know how to fool us thanks to massive probability calculations, but they do not think and do not work at all like our brains can. To simulate real neural networks, drawing inspiration from brainbrainresearchers from theInternational Neuromorphic Systems (ICNS) at the University of Western Sydney, Australia, unveiled the neuromorphic supercomputer Deep South. It will be activated from April 2024 and will be capable of processing 228 trillion synaptic operations per second.
For its part, the human brain can calculate the equivalent of an exaflop (a billion billion floating point operations executed in one second) of mathematical operations per second and it only takes 20 wattswatts ofenergyenergy to achieve this. Reproducing this ability with less consumption is exactly what DeepSouth is designed for.
Imitate the brain to consume less
With this computing power, the supercomputer should make it possible to accelerate the processing of data aroundappsapps such as biomedicine, robotics, space research and artificial intelligence. In any case, its operation will contrast with current supercomputers which, when simulating neural networks with graphics processing units (GPUs) and other multi-core chips, are slow and power-hungry.
The name DeepSouph is both a winkeyeeye to the TrueNorth supercomputerIBMIBM which aimed to build a machine capable of simulating large neural networks and DeepBlue which was the first computer to beat Man at chess. The great thing about DeepSouth is that it’s also scalable. It will therefore be possible to adapt and reconfigure it to increase or reduce its capacities according to demand. If DeepSouth meets the expectations of scientists, by developing and miniaturizing itself, this type of technology could further accelerate the development of AI, while considerably reducing their energy consumption.