According to the researchers, this powerful machine is able to “simulate networks at the scale of the human brain”.
The machine known as DeepSouth uses a neuromorphic system that mimics biological processes.
The supercomputer’s electronics emulate large neural networks with 228 trillion synaptic operations per second, rivaling the rhythm of the brain.
The researchers explained that by simulating the brain, they will be able to improve the processes of artificial intelligence compared to current models.
This was announced by Australian scientists from the International Center for Neuromorphic Systems (ICNS). initiative on December 13.
The University of Western Sydney team collaborates with partners from the University of Sydney, University of Melbourne and Aachen University in Germany.
Andre van Scheik, director of ICNS, explained that DeepSouth is unique in its kind because it requires relatively little energy.
“Progress in understanding how brains compute using neurons has been hindered by our inability to simulate brain-like networks at scale,” he declared.
“Simulating state-of-the-art neural networks on standard computers using graphics processing units (GPUs) and multi-core central processing units is very slow and energy-intensive. Our system will change that. This platform will advance our understanding of the brain and develop brain-level computing applications in a variety of fields, including sensing, biomedicine, robotics, space and large-scale artificial intelligence applications.”
DeepSouth will be smaller than other supercomputers and will be able to process large amounts of data quickly with much less power consumption, the researchers said.
“Ultra-fast, massively parallel processing with much less power: Our brain is capable of processing the equivalent of an exaflop – one billion billion (one followed by 18 zeros) math operations per second – with just 20 watts of power.”the press release states.
“The system is also scalable, allowing more hardware to be added to create a larger system or scaled down for smaller portable or more cost-effective applications.”
The goal of the scientists is to put the supercomputer into operation by April 2024.
Meanwhile, researchers at the University of Sydney recently published a study confirming that the human brain can naturally perform advanced calculations like a supercomputer and make sense of the world through what is known as Bayesian inference.
The researchers developed a mathematical model that matched the way the brain works in visual perception and contained all the necessary mechanisms to perform Bayesian inference.
The study, published in Nature Communications, was created in collaboration with researchers from the University of Queensland and the University of Cambridge.
Research leader Ruben Rideau said that despite the conceptual appeal and explanatory power of the Bayesian approach, how the brain calculates probabilities is largely a mystery.
“Our new study sheds light on this mystery. We found that the underlying structure and connections in our brain’s visual system are configured in a way that allows it to perform Bayesian inference on the sensory data it receives.”, he declared in September.
“What makes this finding important is the confirmation that our brains have an innate design that allows for this advanced form of processing, allowing us to more effectively interpret our environment.”
Demand for supercomputers is growing amid a global artificial intelligence (AI) boom.
Global AI giant Nvidia Corporation and Amazon Web Services (AWS) recently announced a partnership for a new supercomputing infrastructure for generative AI.
AWS will deliver the first AI supercomputer in the cloud with the NVIDIA Grace Hopper superchip and the scalability of an AWS ultra cluster
Both companies build on a long-standing partnership that has fueled the era of generative artificial intelligence and provides the computing power needed to advance technology.
NVIDIA CEO Jensen Huang said that generative AI “transforms cloud workloads” and puts accelerated computing at the core of “diverse” content production.
“Driven by a shared mission to deliver cost-effective, state-of-the-art generative AI to every customer, NVIDIA and AWS are working together across the entire compute stack, including AI infrastructure, acceleration libraries, underlying models and generative AI services,” said Mr. Huang.
The supercomputer, known as Project Ceiba, will be integrated with AWS services such as Amazon Virtual Private Cloud and Amazon Elastic Block Store for high-performance block storage.
Meanwhile, AWS CEO Adam Selipski noted that the two companies have been working together for more than 13 years, starting with the world’s first GPU cloud.
“Today, we offer NVIDIA’s broadest line of GPU solutions for workloads that span graphics, gaming, high-performance computing, machine learning and now generative artificial intelligence.”said Mr. Selipski.