Today’s AI uses the brain as inspiration for software that runs on traditional computers (or supercomputers). But scientists are working on another option to model the brain’s hardware by building a neural network on a microchip.
Called neuromorphic computing (‘in the form of a nervous system’), in theory, the chips offer a completely different and more energy-efficient way to build a computer.
Traditional computers are power-hungry because information is shuttled back and forth between the central processing unit (CPU) and the memory storage.
In neuromorphic chips, units process and store information as an integrated part of the same operation, just like neurons do via synapses (the connection points between neurons). And the strength of those connections can grow or fade, as they do in our brains.
This would let neuromorphic chips undertake parallel processing, similar to the brain, and would make some computing tasks like image processing (e.g. detecting cats) much faster and less energy intensive, by about 1000-fold.
For pure calculation, though, a normal computer would still outperform a neuromorphic chip – just like it outperforms a brain.
As it is, neuromorphic chips aren’t so useful…yet.
“Just because we know how to build hardware that simulates brain components doesn’t mean we know how to make use of it,” says Dr Peter Stratton, a computer scientist working at the Queensland Brain Institute.
“That’s because we don’t really know how the brain uses its own hardware! Until we understand more about how a brain computes things, the full potential of neuromorphic chips isn’t likely to be realised.”
