IBM Research Develops The World's First Artificial Neurons

A team of researchers from IBM's research laboratory in Zurich has developed an artificial version of the neuron. It is the world's first artificial nanoscale stochastic phase-change neurons.

The study was printed in the journal Nature nanotechnology. The invention comprises of a small square of germanium antimony telluride, which is a common ingredient in optical disks, held between two electrodes. It is also known as phase-change material, wherein it can change its phase from an amorphous insulator to a crystalline conductor when struck with a strong enough electric pulse. These will act as a resistor and capacitor and simulate the behavior of biological neurons' lipid bilayer membrane, according to International Business Times.

IBM explained that the researchers applied a series of electrical pulses to the artificial neurons. This resulted in the progressive crystallization of the phase-change material, which causes the neuron to fire. They further explained that this is known as the integrate-and-fire property of biological neurons, which is the foundation for event-based computation. It is similar on how the brain triggers a response when you touch something hot.

ArsTechnica stated that IBM has already developed a population of 500 of artificial neurons. They used them to process a signal in a brain-like (neuromorphic) way. The IBM's artificial neuron has a neuronal membrane (lipid bilayer) around the spike generator (nucleus), inputs (dendrites) and an output (axon) juts like the biological neuron. It also has a back-propagation link from the spike generator back to the inputs. This is to strengthen of some of the input spikes. The difference is that the neuronal membrane is replaced with the germanium-antimony-tellurium.

An artificial neuron may also refer to as binary neuron, Nv neuron, or linear threshold function depending on the specific model. It has a mathematical function that conceives as a model of biological neurons. It receives one or more inputs (dendrites) and sums to create an output (axon).