What is a neural net processor?
A neural net processor is a central processing unit (CPU) that holds the modeled workings of how a human brain operates on a single chip.
Neural net processors reduce the requirements for brainlike computer processing from whole networks of computers that excel in complex applications -- such as artificial intelligence (AI), machine learning or computer vision -- down to one multicore chip.
Neural network processors for AI
Many implementations of convolutional neural networks are currently software-modeled and can span many computers. These current implementations might use many CPUs, graphics processing units for faster parallel processing or even purpose-designed field-programmable gate array cards.
Much as computers were reduced from room-sized monstrosities, it is now possible to make processors that function like the human brain in a single package. For example, IBM's 4,096-core TrueNorth is a single chip that mimics 1 million human neurons and 256 million synapses. That design can reduce the power requirements for neural net processing down to one-tenth of what was formerly required.
Other designs, such as those of silicon chip design company Synopsys, implement embedded vision coprocessors along with neural net processors for greater efficiency in computer vision tasks.