Revolutionary research from the University of Limerick (UL) and the University of Twente brings us closer to a new computer that works like a brain. Researchers have developed a new type of molecular switch that behaves like a ‘synapse’ in the brain, able to learn from the past.
Electronic devices such as computers and data centres consume large amounts of energy. To meet that energy demand, we are now encouraging the installation of solar panels and building giant wind farms. But while it’s important to generate more energy, it’s also important that we look to make our electronics work more efficiently. Our brain is the most efficient computer. It uses 10,000 times less energy than the most economical computer.
This is because our brains process data in a completely different way. Whereas computers process binary 1’s and 0’s, our brains are analogues that use time-dependent pulses. Unlike conventional electronics, our brains can effortlessly process input from millions of neurons and five senses. This means that the brain only uses energy when transmitting, so it can process large amounts of data at once more efficiently.
Therefore, the research team researched to see if this efficient movement could be reproduced. The result is a 2-nanometer- thick molecular layer. It is 1/50,000th thinner than human hair, but it can memorize the history of passing electrons. This molecular material offers a disruptive new alternative to traditional silicon-based digital switches, either on or off, because the switching probabilities and on/off state values change continuously.
And the dynamic organic switches are created to display all the mathematical logic functions required for deep learning, successfully emulating synaptic-like brain operations.
It is possible to simulate the dynamic behaviour of synapses at the molecular level by combining fast electron transfer with slow diffusion-limited proton coupling, similar to the role of biological calcium ions and neurotransmitters. This molecule can change the strength and duration of this pulse. Thus, the molecule demonstrates a form of classical conditioning. It adapts its behaviour to previous stimuli. It is a kind of learning.
This breakthrough will enable the development of entirely new customizable and reconfigurable systems. And these could lead to new multifunctional adaptive systems that greatly simplify artificial neural networks. Then it will also be possible to reduce the energy consumption of electronic devices dramatically. On the other hand, multifunctional molecules that can detect photosensitive or other molecules may help create new types of neural networks and sensors.
