Functional computer model of the brain

Neuroscientists at the University of Waterloo claim to have created the most complex and large-scale model of the human brain. Using the open source Nengo neurosimulator on a supercomputer, they emulated the work of 2.5 million neurons, divided by functionality, in accordance with the real parts of the human brain.

Unlike IBM Watson and other systems, a virtual model called SPAUN (Semantic Pointer Architecture Unified Network) was created not for solving practical problems, but for the most realistic modeling of the human brain. For example, in the Blue Brain project, IBm specialists focused on creating an accurate spatial model, but did not take into account the functional differences in brain regions.

Data entry into SPAUN is carried out through a digital “eye” with a resolution of 28x28 (784) pixels. No other input methods are provided. He can be shown a series of numbers and characters that are transmitted in memory. The system perceives some signs as commands that need to be executed. SPAUN calculates the result of the calculations with a mechanical “hand”.

Interestingly, information is processed by a computer in much the same way as the human brain does, as far as neuroscientists know today. The system has a prefrontal region, subcortical nuclei, basal ganglia, thalamus, etc. Neuroscientists tried to program the processing of information as close to natural as possible. Visual signals enter the visual cortex, then into the thalamus. The thalamus is responsible for the redistribution of information to different areas of the cerebral cortex. The basal ganglia control the flow of information through the prefrontal region, updating it in accordance with the current need.

For example, if you currently need to cook food, the necessary sequence of actions is loaded into the prefrontal area. In case of emergency, this information is erased - and information about driving is downloaded. A person is able to quickly switch from one task to another - has extremely high cognitive flexibility - thanks to the basal ganglia and long-term memory.

In the SPAUN simulator, scientists also laid down the limitations inherent in the human brain. For example, the system cannot store too long a sequence of characters in RAM.

Now, neuroscientists are working on that mechanism of self-training and re-recording of neurons in the “brain” so that SPAUN can reprogram itself and learn new tasks.

The Nengo simulator and the SPAUN neuromodel itself can be downloaded here : spaun.zip .

The scientific work of specialists from the University of Waterloo is published in the journal Science (behind a paid firewall). Free copies: 1 , 2 , 3 .







Sources: ExtremeTech , PopSci