A miniature, massively parallel computer, powered by a million ARM processors, could produce the best brain simulations yet
For all their progress, computers are still pretty unimpressive. Sure, they can pilot aircraft and simulate nuclear reactors. But even our best machines struggle with tasks that we humans find easy, like controlling limbs and parsing the meaning of this paragraph.It’s a little sobering, actually. The average human brain packs a hundred billion or so neurons—connected by a quadrillion (1015) constantly changing synapses—into a space the size of a cantaloupe. It consumes a paltry 20 watts, much less than a typical incandescent lightbulb. But simulating this mess of wetware with traditional digital circuits would require a supercomputer that’s a good 1000 times as powerful as the best ones we have available today. And we’d need the output of an entire nuclear power plant to run it.
Closing this computational gap is important for a couple of reasons. First, it can help us understand how the brain works and how it breaks down. There is only so much to learn on the coarse level, from imagers that show how the brain lights up when we remember a joke or tell a lie, and on the fine level, from laboratory studies of the basic biology of neurons and their wirelike dendrites and axons. All the real action happens at the intermediate level, where millions of networked neurons work in concert to produce behaviors you couldn’t possibly predict by watching a handful of neurons fire. To make progress in this area you need computational muscle.