Compared with biological computers—also known as brains—today’s computer chips are simplistic energy hogs. Which is why some computer scientists have been exploring neuromorphic computing, where they try to emulate neurons with silicon. Yesterday, researchers at IBM announced a new neuromorphic processor, dubbed TrueNorth, in an article published in the journal Science.
At one million “neurons,” TrueNorth is about as complex as a bee’s brain. Experts are saying this little device (about the size of a postage stamp) is the newest and most promising development in “neuromorphic” computing. Despite its 5.4 billion transistors, the entire system consumes only 70 milliwatts of power, a strikingly low amount. The clock speed on the chip is slow, measured in megahertz—today’s computer chips zip along at the gigahertz level—but its vast parallel circuitry allows it to perform 46 billion operations a second per watt of energy.
What sets TrueNorth apart is how well it seems to tackle the problem of pattern recognition, something that’s incredibly simple for a human brain but devilishly difficult for computers. The chip functions through its “neurons,” which are tightly connected; each is linked to 256 others. It’s a design that emulates the neural networks seen in animal and human brains, and it helps TrueNorth identify moving cars, pedestrians, buses, and more.
Here’s John Markoff, writing for The New York Times:
The chip’s electronic “neurons” are able to signal others when a type of data — light, for example — passes a certain threshold. Working in parallel, the neurons begin to organize the data into patterns suggesting the light is growing brighter, or changing color or shape.
The processor may thus be able to recognize that a woman in a video is picking up a purse, or control a robot that is reaching into a pocket and pulling out a quarter. Humans are able to recognize these acts without conscious thought, yet today’s computers and robots struggle to interpret them.
Despite the promise, some scientists are skeptical about TrueNorth’s potential, claiming that it’s not that much more impressive than what a cell phone camera can already do. Still others see it as overhyped or just one of many possible neuromorphic strategies.
Jonathan Webb, writing for BBC News:
Prof Steve Furber is a computer engineer at the University of Manchester who works on a similarly ambitious brain simulation project called SpiNNaker. That initiative uses a more flexible strategy, where the connections between neurons are not hard-wired.
He told BBC News that “time will tell” which strategy succeeds in different applications.
Proponents argue that the chip is endlessly scalable, meaning additional units can be assembled into bigger, more powerful machines. And if its processing potential improves, as traditional silicon chips did in the past, then TrueNorth’s neuromorphic successors could lead to cell phones powered by extremely high-power, energy-efficient processors, the sort that could make today’s smartphone CPUs look like those in early PCs.