Mimicking the Human Brain With Next-Generation Custom-Built Computers

In this article
Custom-built computers inspired by human brain architecture are taking another step toward real-time simulation of the brain's neural network, with exciting performance.
What Is a Supercomputer?
A supercomputer can simply be defined as a system of massively parallel processors, high-performance vector processors and computer clusters (not necessarily all components). Supercomputers have enabled new scientific methodologies such as computational science, combining experiment and calculation.
Today's supercomputers perform, in 1 second, hundreds of trillions of operations. Our personal computers are called general-purpose computers.

Mimicking the Human Brain's Neural Network
Computers specially built to mimic neural networks are beginning to produce results similar to supercomputers simulating the brain. Frontiers in Neuroscience published research showing that SpiNNaker , a new computer, outperformed conventional supercomputers in speed and energy-efficiency tests, particularly energy efficiency. The study's main aim is to expand knowledge of neural activity and conditions such as epilepsy and Alzheimer's.
Dr. Sacha van Albada, the study's architect and head of the Theoretical Neuroanatomy Group at Germany's Jülich Research Center, said: "SpiNNaker can produce a detailed biological model of the cortex, the brain's outer layer processing sensory information. Its more extensive detailed results can contribute to advanced robotics research." as her statement.
The Human Brain
The human brain contains 100 billion neurons (brain cells) with many interconnections. Think of it as a vast forest: each tree is a neuron, and connections between cells number as many as the trees' leaves. It may therefore be considered the universe's most complex structure. We now know which brain regions govern which bodily functions from how cells interact and behave. But how neural activity becomes behavior, such as thought becoming muscle movement, or how information is stored is not fully understood.
Even the best simulation software and fastest supercomputers can simulate only 1% of the brain.
European Human Brain Project research shows current supercomputers take several minutes to perform what a human brain does in 1 second. Processes such as learning may take hours or weeks, making problem-specific computer architecture crucial.
Development of the Next-Generation Custom-Built SpiNNaker Computer
Developed over more than 15 years within the Human Brain Project, SpiNNaker is a custom-built computer with roughly half a million simple processors and 600 circuit boards, running its own software. Its principle, called digital neuromorphic computing, differs from conventional supercomputers. It achieved results close to NEST, one of the best supercomputers for neural calculations. Yet this analysis used only 6 boards, approximately 1% of SpiNNaker's total power. Researchers believe software and circuit-board improvements will substantially increase performance.
In conclusion,although computers perform certain operations very quickly, much development is needed to match even one human brain's complexity and skills. They remain weak in areas such as clinical decision-making (physicians' treatment decisions). However, if the SpiNNaker project succeeds, causes and treatments of mysterious neurological diseases may be uncovered. High-level brain simulation could also revolutionize energy efficiency and performance in robotic decision mechanisms.
Sources
Performance Comparison of the Digital Neuromorphic Hardware SpiNNaker and the Neural Network Simulation Software NEST for a Full-Scale Cortical Microcircuit Model.
Front. Neurosci., May 2018.
Was this article helpful?
Your feedback helps us improve our content.





Reader comments
Share your comment
Your name and comment are published only after administrator approval. Your email address is never shown to visitors.
Be the first to comment.