Scientists have succeeded in simulating a mouse brain using an IBM supercomputer

However, the simulation still only shows the technical characteristics of the brain's operation, and did not produce thoughts.

 Researchers in the US managed to simulate half of the brain of a virtual mouse on a supercomputer. The scientists ran on IBM's BlueGen L computer, a simulator that simulates the cerebral cortex, which managed to have a processing capacity as large and complex as half the brain of a mouse.

In smaller simulations, the researchers were able to see characteristics of thought patterns similar to those observed in the brains of real mice.

signs of life

Brain tissues represent a huge problem for those trying to model them due to their complexity and the sheer number of potential connections between the components involved. Three researchers - James Frye (Frye), Ragagupal Ananthanarayanan (Ananthanarayanan), and Darmanda Modha (Modha) - recently published a study under the title "Towards Real-Time, Mouse-scale Simulation of the Cerebral Cortex." -Scale Cortical Simulations".

Half the brain of a real mouse contains eight million neurons, each with up to 8,000 synapses, or connections, with other nerve fibers. In order to establish such a system, the three write, they had to make full use of the computing capacity, communication and memory of any computer system." The team, from IBM's Almaden Research Center and the University of Nevada, ran the simulation on the Blue Gene L supercomputer with 4,096 processors, each with 256 megabytes of memory.

Using this computer, the researchers were able to imitate only about half of the mouse's brain, meaning 8,000 neurons each with up to 6,300 synapses. The low complexity of the simulation means that the system can only be run for ten seconds at a speed 10 times slower, that is, in effect, to simulate one second of the operation of a mouse's brain.

In smaller simulations, the researchers were able to see consistent biological dynamic features that occurred when nerve impulses flowed along the virtual cerebral cortex. In these experiments, the researchers observed groups of neurons that formed groups spontaneously. They also saw neurons firing at virtual synapses in something that looked similar to the coordinated oscillatory patterns seen in nature.

The researchers said that although the simulation shares similar events with the mouse's mental structure in terms of nerves and connections, it does not have the structure seen in the brains of real mice. Enforcing such structures and getting the simulation to do useful work can be a much more difficult task than just installing the (nerve) plumbing.

In future experiments, the team intends to speed up the simulation, become more faithful to neurobiological reality, add structures visible in the brains of real mice and make the responses of neurons and synapses more detailed.