Though computers are fast and getting faster, processors are still far from competing with the human brain. Computer chips usually only send or receive a signal from one place at a time. This is radically different from neurons in the brain that are connected to hundreds of others, and send and receive many messages simultaneously. They also send and receive different kinds of messages simultaneously, as opposed to the binary “on” or “off” of computer transistors. In the past, scientists have tried to build networked brain models out of regular transistors, but researchers at IBM took a different approach, creating a new chip that functions like real brain cells. It’s called TrueNorth, and ArsTechnica explains what it’s made of:
Its 5.4 billion transistors include over 4,000 individual cores, each of which contain a collection of circuitry that behaves like a set of neurons. Each core has over 100,000 bits of memory, which store things like the neuron’s state, the addresses of the neurons it receives signals from, and the addresses of the neurons it sends signals to. The memory also holds a value that reflects the strength of various connections, something seen in real neurons. Each core can receive input from 256 different “neurons” and can send spikes to a further 256.
The chip is shockingly energy efficient. Because the transistors are only active when receiving or sending a message, they need much less energy than regular computer chips that must constantly use energy to function. TrueNorth only uses about 20 milliwats per square centimeter, while regular processors use 50 watts or more.
There is no software built to work with TrueNorth, but researchers designed it to match with the already existing Compass software package. This means programs can work with the chip if they are pre-programmed for Compass, and those results can be compared to the results of running the same operation on a normal chip. These experiments proved that TrueNorth uses exponentially less energy than other networked chips, “[cutting] energy use by 176,000-fold compared to a traditional processor and by a factor of over 700 compared to specialized hardware designed to host neural networks.”
This is just the beginning: TrueNorth eventually will be able to work as one tiny piece of a larger brain-like machine, “by tiling multiple TrueNorth chips, creating systems with hundreds of thousands of cores, hundreds of millions of neurons, and hundreds of billion of synapses.” (Images: ArsTechnica)