A team of researchers from the Universities of Oxford and Exeter in the UK and the University of Münster in Germany have made a significant breakthrough in the field of artificial intelligence by developing a microchip with a circuitry structure mimicking that of a synapse in the brain. By engineering chip structure to emulate the physical structure of the brain, these researchers intend to grant computing systems abilities that humans take for granted, such as parallel processing and built-in learning.
In the human brain, neurons communicate with each other by transmitting electrochemical signals back and forth across gaps between nerve endings. These gaps are known as synapses, and the signals sent across them are the origin of everything we think, feel, and do.
According to Professor Wolfram Pernice, a co-author of the paper from the University of Münster, there are approximately 10,000 times more synapses than neurons in the brain. This means that in order to create a computer that “thinks” like the human brain, the synaptic mechanism must be replicated in the structure of the computer’s circuitry — which is exactly what has been accomplished.
As if this neuronal desgin didn’t massively increase potential computing power already, there’s another important feature of this microchip — it is photonic, meaning it operates using light instead of electricity. Photonic circuitry exponentially increases a computer’s speed relative to electric circuitry, due to the inherent speed of light. Photonic chips also use a fraction of the energy that traditional chips use, making them more efficient and more powerful.
“We demonstrated the world’s first integrated photonic synapse, which… is the [most] crucial step for photonic neuromorphic computing,” said Harish Bhaskaran, Professor of Applied Nanomaterials at Oxford University and leader of this research team.
But this is just the first step for the researchers.
“The next obvious step would be a photonic firing neuron and a complete architecture for the photonic neuromorphic computing,” Bhaskaran said. “In the long term, our research collaboration would like to… build entire computing platforms using such devices.”
In other words, the researchers plan on building an entire processing unit made up of these photonic chips, transmitting signals and data in the same manner as synapses in the brain, but in optic rather than electrical form. To put it more simply, they plan on building an entire computer that mimics the brain and operates at the speed of light.
“We can imagine plenty of applications in artificial intelligence, big data and internet of things, but we must emphasize that [this technology is still in its] early days – we need many more innovations in terms of being able to compute on a chip, and build true analogues of brain-inspired computers such as on-chip neurons, etc.,” said Bhaskaran.
While this technology may still be in its infancy, it is a crucial step towards unlocking computational power, the likes of which scientists have dreamt of and theorized about for decades.
The full paper on this research can be found here.
Zengguang Cheng and Carlos Ríos of Oxford University, as well as Professor C. David Wright of the University of Exeter, contributed to this research.
