New Artificial Visual System with record low energy consumption performs close to human brain

In a new development, an artificial visual system that works on ultra-low energy has been built by a joint effort of a team of researchers at the City University of Hong Kong. Mimicking the human brain, the visual system performed data-intensive cognitive tasks impeccably. The outcome of the experiment could provide a promising device system for next-generation artificial intelligence applications.

The development of the visual system is published in the scientific journal Science Advances.

In fact, advances in semiconductor technologies used in digital computing is nearing stagnation, hence, neuromorphic computing systems have been regarded as an alternative. Scientists have been making attempts to develop next-generation advanced AI computers, which could be energy-efficient, lightweight, and adaptable as the human brain.

Drawbacks of Existing Artificial Synapses challenges performance of cognitive tasks

“Meanwhile, performance of existing artificial synapses to emulate the human brain is challenging. They have drawbacks of effective emulation of the brain’s neuroplasticity, or to re-wire itself using ultra-low power,” said the lead author of the study.

However, with advancement, artificial synapses emulate biological ones, meaning maintaining the gap over which the two neurons transmit electrical signals for transfer of information with each other in the brain. As a result, artificial synapses mimic the efficient neural signal transmission of the brain and memory formation process.

In order to increase the energy efficiency of artificial synapses, for the first time, the research team introduced quasi-two dimensional electron gases into artificial neuromorphic systems. To do this, researchers created oxide superlattice nanowires, and designed quasi-2DEG photonic synaptic gadgets that attained a record energy consumption of as low as 0.7fJ per synaptic event. This implies 93% reduction in energy consumption when compared to synapses in the human brain.