This post was originally published on here
Researchers are setting their sights on an ambitious goal: creating a simulation at the scale of the entire human brain using one of the world’s most powerful supercomputers.
Leveraging a supercomputer known as JUPITER — short for Joint Undertaking Pioneer for Innovative and Transformative Exascale Research — the team believes it can overcome the hurdles that stalled previous attempts, such as the decade-old Human Brain Project.
The JUPITER system, currently ranked as the fourth most powerful supercomputer globally, possesses the computational firepower necessary to run simulations of billions of firing neurons.
“We have never been able to bring them all together into one place, into one larger brain model where we can check whether these ideas are at all consistent,” Markus Diesmann, a neurophysics professor leading the team at the Jülich Research Centre in Germany, told New Scientist. “This is now changing.”
Last month, the team demonstrated that a “spiking neural network”—a model of the brain’s neurons and synapses—could be scaled up to run on JUPITER. This simulation effectively matched the human cerebral cortex’s 20 billion neurons and 100 trillion connections.
The model is grounded in real anatomical data from smaller experiments on human brain neurons, utilizing JUPITER’s thousands of graphical processing units. The hope is that by running simulations at this massive scale, researchers can reveal secrets of brain function previously hidden in smaller models.
“We know now that large networks can do qualitatively different things than small ones,” Mr. Diesmann said. “It’s clear the large networks are different.”
According to Thomas Nowotny, a mathematical physics professor at the University of Sussex, full-scale simulations are crucial because “downscaling is not just simplifying it a little bit, or making it a bit coarser, it means actually giving up certain properties altogether.”
Mr. Nowotny told New Scientist that these simulations could allow researchers to test theories impossible to observe in real brains, such as how memories are formed. Furthermore, the technology could revolutionize medicine by testing how conditions like epilepsy respond to certain drugs within a simulated environment.
Additionally, Johanna Senk from the University of Sussex, who is collaborating with Mr. Diesmann, points out that the increased computational power allows simulations to run faster. This speed could provide insights into relatively slow biological processes, like learning, while incorporating greater biological detail.
Despite the optimism, experts acknowledge that researchers are still scratching the surface of an organ that remains largely mysterious. Even with exascale computing, simulations lack essential functionalities of real brains, such as sensory input from real-world environments.
“We can’t actually build brains,” Mr. Nowotny said. “Even if we can make simulations of the size of a brain, we can’t make simulations of the brain.”
