Researchers at Western Sydney University in Australia are constructing the world’s first supercomputer capable of simulating network connections on a scale equivalent to the human brain. This supercomputer, named DeepSouth, is expected to go online next year.
The team named the supercomputer DeepSouth based on IBM’s TrueNorth system, which started the idea of building computers that act like large networks of neurons, and Deep Blue, the first computer to beat a world chess champion. The name also gives a nod to where the supercomputer is located geographically: Australia, which is situated in the southern hemisphere.
DeepSouth operates on a foundation of neuromorphic computing, mimicking the structure and function of the brain. In stark contrast to traditional supercomputers that rely on sequential processing, DeepSouth leverages a network of artificial neurons capable of transmitting signals akin to biological neurons. This allows DeepSouth to perform an estimated 228 trillion synaptic operations per second, mirroring the processing power of the human brain.
This speed is achieved through the use of custom-designed neuromorphic chips that utilize a technology called spiking neural networks. In these networks, information is encoded as electrical spikes transmitted between artificial neurons, mimicking the communication processes within the brain.
Objectives of the DeepSouth
DeepSouth’s approach is different. It tries to mimic the brain’s ability to learn, adapt and process information in a massively parallel and distributed way. This is a departure from the von Neumann architecture that most computers are based on.
One of the primary objectives of the DeepSouth project is to understand how the human brain, using just 20 watts of power, can process massive amounts of information. This understanding could lead to the development of more energy-efficient computing systems.
“Simulating brain-like networks is the bottleneck in our understanding of how brains compute with neurons,” said André van Schaik, director of the International Center for Neuromorphic Systems at Western Sydney University. He said simulating neural network spikes on regular computers using graphics processing units and central processing units is slow and power hungry. DeepSouth will change that.
Ralph Etienne-Cummings of Johns Hopkins University in Baltimore, who was not involved in the work, told New Scientist that DeepSouth will be a key milestone in the study of neuroscience. “If you’re trying to understand the brain, this will be the hardware to do this,” he said.
Future of the Deepsouth Computer
DeepSouth is part of a larger scientific effort to build machines that can match the human brain. Other researchers are working on “biological computers” powered by real brain cells. That’s a big step forward for our understanding of the brain and computing systems.
The human brain is estimated to have 86 billion neurons connected by trillions of synapses. While DeepSouth boasts an impressive processing power, it currently operates at a smaller scale. Scientists acknowledge the need for further advancements to fully replicate the brain’s intricate structure.
The research team say the DeepSouth Computer has a new architecture that optimizes data flow and gets rid of the bottlenecks that are common in traditional supercomputing systems. This allows the machine to do complex simulations and data analysis at speeds never before possible while using less power.
Critically, the DeepSouth Computer project has prioritized accessibility, with plans to collaborate with institutions that traditionally lack the resources to utilize such advanced computing power. This democratization of technology is expected to spur innovation across various disciplines, allowing for a more inclusive approach to research and development.
