Using the world’s fourth most powerful supercomputer, at Riken research institute in Japan, scientists have managed to simulate 1 second of brain activity. With nearly 90,000 processors and a petabyte of memory it took nearly 40 minutes to accomplish the simulation of 1.73 billion neurons.
The brain is the most complex biologic structure known to man, where the human brain tops the list with nearly 20 billion neurons in just the cerebral cortex and 85 billion in the entire nerve system. The colossal amount of neurons and synapses makes it hard to really explore the functions of the brain, but an experiment done by a group of Japanese and German scientists hints of a not too distant future where we can digitally simulate a human brain.
The scientists at Okinawa Institute of Technology in Japan and Forschungszentrum Jülich in Germany used a framework called NEST to construct a virtual neural network consisting of 1.73 billion neurons with 10.4 trillion synapses. The simulation then ran on the supercomputer K sporting 82,944 cores.
A fraction of the brain – 40 minutes to execute
Even though it only had a tenth of the neurons the cerebral cortex the simulation needed 1 petabyte of system memory and took 40 minutes to execute, resulting in 1 second of simulated brain activity. It should be said that it is the biologic side of the activity that is studied and that the neurons were scattered randomly.
While it may sound restraint to limit the amount of neurons to only a fraction of the human brain and far from realtime it is a big step in the right direction. Not too many years ago this type of simulation was unthinkable, and 1 second of activity of a tenth of the cerebral cortex in 40 minutes can lead to much bigger and faster projects in the next ten years. Some say we may be able to simulate a whole human brain before 2050, while others are more skeptical.
At the same time there are many other properties of the human brain that are a lot harder to simulate. The ability to create memory and shape a conscience is founded in its ability to dynamically form new nerve connections, which could be considerably harder to simulate.
The project paves the way for several other projects, partially when it comes to constructing more powerful supercomputers but also projects in neurology. The scientists behind the simulation will cooperate with several other projects in the future based on the results from this one.