Researchers at Western Sydney University in Australia have teamed up with tech giants such as Intel and Dell to build a massive supercomputer capable of simulating neural networks on the scale of the human brain. This one-of-a-kind supercomputer, called DeepSouth, will be up and running very soon.
Researchers say the computer, called DeepSouth, will be able to mimic the brain’s neural networks by performing 228 trillion synaptic operations per second, matching the speed at which the human brain completes operations. The project was announced this week at the NeuroEng Workshop, a gathering of luminaries in computational neuroscience organized by the International Centre for Neuromorphic Systems (ICNS) in Western Sydney.
The mystery of the human brain could be solved
When DeepSouth becomes operational in April next year, it could give researchers a unique perspective on how the human brain processes information. Instead of aiming for DeepSouth to be the world’s most powerful conventional supercomputer, researchers say they want to simulate the brain’s natural network of neurons using a neuromorphic system that mimics biological processes.
The result, they say, will be a more efficient and less power-hungry supercomputer built from the ground up to simulate synaptic activity in the human brain. Put simply, neuromorphic computing involves performing a large number of operations simultaneously while moving very little data, resulting in much lower energy consumption.
“Simulating spiking neural networks on standard computers using graphics processing units (GPUs) and multi-core central processing units (CPUs) is very slow and power-hungry,” ICNS director and Western Sydney professor André van Schaik said in a statement. Our system will change that.
Spiking neural networks (SNNs) are artificial neural network models that more closely mimic natural neural networks.
Could Open the Door to Advanced Artificial Intelligence
By advancing our knowledge of the brain, scientists hope to develop brain-scale computing applications in fields as diverse as sensing, biomedicine, robotics, space, and large-scale AI applications. For example, this technology could enable the creation of advanced smart devices or make AI models consume less power.
Scientists not involved in the DeepSouth project are also excited about the development. “At the end of the day, there are two types of researchers who will be interested in this: those who study neuroscience or those who want to develop new engineering solutions in AI,” said Ralph Etienne-Cummings, a professor of computer engineering at Johns Hopkins: “If you’re trying to understand the brain, this is the hardware to do it.”
