Researchers from the National Centre for Scientific Research (CNRS) in Thales, the University of Bordeaux in Paris-Sud, and Evry have created an artificial synapse called a memristor directly on a chip. This leads the way for intelligent systems that will use less energy and take less time to learn autonomously.
The Memristor Works Similar To Synapses In Human Brain
The human’s brain has trillions of synapses, which serve as connections between neurons. Person’s ability to learn directly depends on the ability of synapses to reconfigure the strength by which they connect neurons. The overall distribution of the synaptic strengths provides the neural network with memory. The more the synapse is stimulated, the more the neuronal connection is reinforced, resulting in improved learning.
This mechanism was an inspiration for researchers to develop an artificial synapse called memristor. It consists of a thin ferroelectric layer set between two electrodes. Layer’s resistance can be tuned using voltage pulses similar to those in neurons. The memristor’s capacity for learning is based on this adjustable resistance. Therefore, if the resistance is low the synaptic connection will be strong. Oppositely, if the resistance is high the connection will be weak.
The idea of the memristor is not new. It was first conceptualized in the 1970s by circuit theorist Leon Chua and subsequently built in 2008. Originally, the memristor was a hypothetical non-linear passive two-terminal electrical component relating electric charge and magnetic flux linkage. In 2008, researchers at HP Labs have solved a decades-old mystery by proving the existence of a fourth basic element in integrated circuits that could make it possible to develop computers that turn on and off like an electric light. Finally, the development of artificial synapse takes this matter to the next level.
Now, when researchers manage to create 45 memristors which were able to learn to detect simple patterns without any assistance, the next goal is to develop neural networks on a chip that contain hundreds of these ferroelectric memristors.
Doubtless, neural networks are responsible for some of the biggest breakthroughs in AI. Probably one day we will see AI systems that are able to learn just as well as the human brain.