Supercomputer May Unlock Secrets of The Brain

A supercomputer that can process data in the same way and at the same speed as a human will be operational next year. It's a claim that's far more impressive than it might sound.

Unlike most supercomputers, DeepSouth's main purpose is to help scientists understand more about how the brain operates as a "low power device".

While even the most basic computers can carry out calculations extremely quickly (and far faster than a human), they can't operate in the same way. In simple terms, that's because a single computer processor can only do one thing at a time.

A brain uses a network of neurons which send electrical charges. Using the network means the brain can carry out tasks and solve problems by sending multiple signals at once. (That's pretty handy as, for example, you wouldn't want an idle thought to take priority and delay a signal telling your heart to beat).

This approach, known as a "spiking neural network" also makes it easier to solve problems by analyzing multiple possibilities at once. While we do this for many actions, an easy example to understand is a chess player considering how a particular move might lead to different situations several moves ahead.

228 Trillion Operations Per Second

DeepSouth is the work of the International Centre for Neuromorphic Systems in Sydney, assisted by Dell and Intel. It's the largest "neuromorphic" computer ever made and will be able to carry out 228 trillion synaptic operations (one "neuron" firing to send a signal) every second. That's in the same ballpark as the estimated performance of a human brain. (Source: businessinsider.com)

While most supercomputers are designed to carry out complex tasks such as modeling weather patterns, the people behind DeepSouth are much more interested in how it operates than what it does.

Brain Is Energy Efficient

The big question they want to learn is why the brain is able to work so quickly while using almost no power compared to a computer. New Scientist notes the energy used for the brain to operate is "barely more power than [used by] a light bulb." (Source: newscientist.com)

One explanation is that the way the brain processes data through the network is more efficient. Another is that traditional computers are effectively running continuously in the background, while the brain's activity varies depending on the specific tasks at hand.

Among other uses, researchers will be able to create hypothetical models of a brain and simulate them on the computer. It's also possible they could learn about how the brain can (and can't) reconfigure itself to cope with injuries such as concussion and head trauma.

What's Your Opinion?

Did you realize the brain was so efficient? How beneficial would it be to have powerful computers with much less energy demand? Do you think this project will prove worthwhile?