Professor Namkyu Park, “shedding light on a new direction to develop low power, high-speed devices”

The human brain requires around 20W of power to operate. Google DeepMind’s artificial intelligence ‘AlphaGo’ uses 170kW of power to play the board game Go. Our brain only uses 85000th of its power to do everything, including computation and memorizing. Therefore, there has been active research on the ‘Neuromorphic Chip’, which mimics the brain to process a massive amount of data at high speed with a little amount of power.

The research team of professor Namkyu Park from Department of Electrical and Computer Engineering of SNU revealed on the 4th that they successfully designed a neuromorphic chip which operates with light rather than with the movement of electrons as in conventional semiconductor chips. The work was published in the international journal ‘Advanced Science’ on the 3rd.

Neuromorphic chip does not follow how conventional CPUs work and adopts a new computational method developed by mimicking the human brain. The brain processes information through exchanging electrical and chemical signals by adjusting the connection strength between neurons. Signals are controlled by the ion channels on the neuron cell surface.

The team engineered a metamaterial, which acts differently based on the light intensity, and proposed to adapt this as the channel in the neurons. A metamaterial is created to have a property that cannot be found in natural materials. Their work differs from other works in the sense that they focused on adopting the structural characteristic of the neurons instead of simply mimicking the way they operate.

They finished the designing but not yet implemented the neuromorphic chip. They proved theoretically that the power of electric signals can be maintained regardless of additional noise.

Professor Namkyu Park emphasized the significance of the work saying, “We transferred the operation behind biological structure into the physics domain to design a new optical device. We initiated a new direction in the development of high-speed neuromorphic chip and artificial intelligence. “


An optical neuron that works with photons on the channel [Photo credits, Ministry of Science and Technology]

Source: http://ee.snu.ac.kr/community/news?bm=v&bbsidx=48894
Translated by Kyungjin Lee, English Editor of Department of Electrical and Computer Engineering, jin11542@snu.ac.kr