Microchips designed more like brains than those used in traditional computers may allow researchers to develop machines that come close to thinking the way biological organisms do.
Today's computer circuitry relies on the von Neumann paradigm to transmit data between a central processor and memory chips for linear operations, but it's a bad system for powering machinery that adapts to complex environments or applies situational awareness the way that brains allow living creatures to.
The situation has prompted companies to develop neuromorphic chips, a class of microchips that compute more dynamically and may one day lead to computers and mobile devices that understand their surroundings and operate more agilely. They're modeled on biological brains and can respond to data changes without needing specific programming. Here are three advances in neuromorphic chips to keep an eye on:
- Researchers in Qualcomm's Zeroth program are currently testing neuromorphics as the company prepares to commercially release brain-mimicking technology in 2015. They held a demo this year where a neuromorphic chip-enabled robot's camera learned where to place an action figure after "seeing" an engineer mimic the same action;
- Stanford researchers have developed the Neurogrid, a circuit board comprising 16 Neurocore chips, and each chip simulates 65,536 neurons. They designed the board to simulate 1 million neurons and billions of synaptic connections, and plan to test its ability to power robotic limbs; and
- The Defense Advanced Research Projects Agency is sponsoring a $100 million project for IBM and HRL Laboratories to develop neuromorphic chips in the Systems of Neuromorphic Adaptive Plastic Scalable Electronics program (SyNAPSE). They intend to create systems that can emulate the brain's ability to grow synaptic connections and solve problems on the fly.
Scientists and designers hope that this technology, when combined with existing processors in current gadgets, will make common machinery smarter, like smartphones that can see, speak, listen and drive in scenarios that can't always be programmed for ahead of time.