It seems there is no stopping the development of artificial intelligence (AI). The advancement in technologies that mimic how the human brain works led to AI systems called neural networks.
Today's AI has the capability to imitate how the brain recognizes speech and images, which has algorithms that can be trained. However, training algorithms running on an artificial neural network is time-consuming, not to mention it does take a significant amount of energy.
Researchers from the Universities of Bordeaux, Paris-Sud, Evry, and French National Center for Scientific Research (CNRS) in Thales, have developed "memristor," an artificial synapse directly embedded on a chip that is capable of learning autonomously.
The researchers report, they have developed a physical model explaining this capacity and to help predict how it functions, as published in the journal Nature Communications. Additionally, the discovery makes it possible to create a network of synapses and intelligent systems that consume less time and requiring less energy.
According to IFLScience, the idea of a memristor is not new, it was first conceptualized in the 1970s and was reportedly built in 2008, the new study, however, takes it to a new level. The general idea of a memristor is to mimic the brain's neurons and synapses by creating an electronic equivalent. Like our own biological neurons and synapses, the memristor is able to process and store information.
Human Brain synapses function as connectors between neurons. We learn when these connections are reinforced, and improved when connections are reinforced, improved and when synapses are stimulated.
Memristors emulate the behavior of these synapses, by way of variable resistance depending on how these artificial synapses receive electronic excitations. Memristor is made up of a thin ferroelectric layer that is enclosed between two electrodes.
Voltage pulses allow the chip's resistance to be adjusted, similar to that of a biological neuron. Accordingly, memristor's synaptic connection will be strong when resistance is low, and vice-versa. Its capacity to learn is dependent on adjustable resistance, according to Futurism.
Though the research is focused on the development of these artificial synapses, the extent of the functions of these devices remains unknown. However, a full understanding of the process will pave the way for the creation of more complex systems, such as a series of artificial neurons interconnected by these memristors.
It might not be long before artificial neural networks are incorporated as a standard part of processors, should the researchers work proceed according to their plans.
