Researchers in the UK recently switched on the SpiNNaker, the world’s most powerful computer. Under construction for about 12 years, the supercomputer was built by the University of Manchester and is designed to mimic the human brain.
Meet the SpiNNaker
A big challenge in developing computers to mimic the human brain is that the hardware should be able to handle the massive parallel processing capacity of brains. The team at Manchester solved this through a computer chip consisting of 18 processor cores.
More than 55,000 such chips were linked to create the supercomputer featuring a million processor cores with a capacity to perform 200 trillion operations per second. This allows the machine to model far more biological neurons in real time than any other computer in history.
“SpiNNaker completely re-thinks the way conventional computers work… We’ve essentially created a machine that works more like a brain than a traditional computer, which is extremely exciting,” Steve Furber, a computer scientist at the University of Manchester, said in a statement (Motherboard).
Neurons in human brains communicate with one another using brief flashes of electrochemical energy. Though individual flashes have very little information, hundreds of millions of simultaneous flashes allow for large-scale information processing. The team used this aspect of the brain’s design in their supercomputer where the processor cores send billions of packets of information to thousands of destinations within the system.
As such, the SpiNNaker does not communicate by sending a huge amount of information between A and B through a standard network. Instead, the supercomputer sends small amounts of information to multiple destinations at the same time, perfectly mimicking the architecture of a human brain. The custom chips handling the transmission of information also allow for a drastic reduction in the packet size.
“SpiNNaker can support detailed biological models of the cortex — the outer layer of the brain that receives and processes information from the senses — delivering results very similar to those from an equivalent supercomputer software simulation… The ability to run large-scale detailed neural networks quickly and at low power consumption will advance robotics research and facilitate studies on learning and brain disorders,” Dr. Sacha van Albada, lead author of the study, said to ScienceDaily.
Future of the SpiNNaker
The research group aims to eventually model about 1 billion neurons in real time. However, considering that the human brain has about 100 billion interconnected neurons, it will be a long time before the SpiNNaker is actually able to fully mimic the human brain.
“The team agreed that this system was still in the early stages and that they will be looking to refine the software in order to enable it to provide greater efficiency, and also larger real-time simulations with these computers… Now, the SpiNNaker consists of 600 circuit boards, but the actual simulation used just 6. One of the authors, Steve Furber, speaking on behalf of his team, said that they were planning to reduce this number to just one board,” according to Evolving Science.
Efforts are being directed at upgrading the supercomputer chip. While the SpiNNaker currently uses a chip that has 19 ARM968 processor nodes, the team plans on releasing an upgraded chip by 2023 featuring 160 Arm Cortex-M4F nodes per package.