The human brain is very powerful for something we all carry about without giving it much consideration. Its capabilities are so remarkable that not even the most advanced computers can fully emulate them at this time. However, things could be about to change. DeepSouth, a powerful supercomputer developed by Western Sydney University, is the first to be able to simulate a whole human brain.
DeepSouth will be able to execute an astounding 228 trillion synaptic actions per second once it is operational. Thanks to its creative neuromorphic architecture, this is equivalent to the activity of all the many linked neurons in the brain.
The International Centre for Neuromorphic Systems (ICNS) at Western Sydney’s Professor André van Schaik said in a statement that “our inability to simulate brain-like networks at scale hampered progress in our understanding of how brains compute using neurons.” It is just too sluggish and energy-intensive to simulate spiking neural networks on conventional computers with Graphics Processing Units (GPUs) and multicore Central Processing Units (CPUs). That will alter our system.
Scientists have had a difficult time imitating the brain’s exceptional energy efficiency in a synthetic computer thus far. According to Domenico Vicinanza, an associate professor of intelligent systems and data science at Anglia Ruskin University, the Oak Ridge National Laboratory’s Frontier supercomputer—which is widely regarded as the fastest computer in the world right now—needs 22.7 megawatts to operate. This information was provided to The Conversation.
In comparison, the human brain requires just 20 watts to function at the same pace as a billion-billion computations per second, or exaflop.
Thus, DeepSouth will enable researchers to investigate computers in a less energy-intensive manner.
Artist’s rendition of DeepSouth, a red-accented supercomputer on a red backdrop that symbolizes human brain networks
Additionally, the neuromorphic design is essentially different from that of conventional electrical computers, which hasn’t altered much in many years. Computers have traditionally been distinguished by discrete memory and processor units; data is handled in one location and saved in another.
Even while there’s probably still a lot we don’t know about how the human brain processes memories, scientists are drawing inspiration from the computers inside our skulls when they create the devices of the future.
The neuromorphic circuitry of DeepSouth is built on networks of parallel-capable basic processors. It simulates the ability of various synaptically linked neurons in the brain to fire concurrently. Researchers will be able to utilize the technology without needing to have a deep grasp of the hardware itself since it will be scalable and easily reprogrammable from the front end using the well-known Python programming language.
However, specifically what types of applications would we be discussing?
Professor van Schaik stated, “This platform will advance our understanding of the brain and develop brain-scale computing applications in diverse fields including sensing, biomedical, robotics, space, and large-scale AI applications.” He went on to mention several potential applications, including more effective artificial intelligence (AI) platforms, smart devices, and agricultural sensors.
Ralph Etienne-Cummings of Johns Hopkins University, who is not directly involved in the DeepSouth project, spoke with New Scientist about how his research, which has focused on legged locomotion and mobile robotics for three decades, could benefit from the supercomputer. He also mentioned how his work has advanced neuroprostheses and brain-computer interfaces.
He said, “This will be the hardware to do it on if you are trying to understand the brain.”
Hopefully going up in April 2024, DeepSouth is named after two of the titans of the supercomputing industry, IBM’s Deep Blue and TrueNorth, as well as a reference to its location in Sydney, Australia. We’ll have to wait and see what science can accomplish before supercomputers with the computational capacity of human brains are introduced.