Go to contents

Samsung to develop AI semiconductors that mimic brain neural networks

Samsung to develop AI semiconductors that mimic brain neural networks

Posted September. 27, 2021 07:26,   

Updated September. 27, 2021 07:26


A joint research team led by Samsung Electronics released a study in a world-renowned academic journal on the development of future semiconductors driven by artificial intelligence (AI) that mimic the human brain neural system. It is positively assessed that the South Korean electronics innovator has just come closer to the development of neuromorphic semiconductor systems that enable high-level brain functions such as reasoning and cognition.

Samsung Electronics announced on Sunday that British academic journal Nature Electronics posted a thesis on neuromorphic engineering written by Samsung SDS President and CEO Hwang Sung-woo, Samsung Electronics Vice Chairman Kim Ki-nam, Samsung Advanced Institute of Technology’s fellow researcher Ham Don-hee at Harvard University and Chemistry and Chemical Biology Professor Park Hong-geun at Harvard University.

Jointly studied by Samsung Electronics and Harvard University since 2019, neuromorphic semiconductors emulate the human brain neural networks and neurons. The technology designs neural networks just like a map by identifying fine signals at synapses - the site of transmission of electric nerve impulses between two nerve cells (neurons) – when electric signals occur. A map of neural networks is applied to semiconductors so that each memory can serve as a contact point between neurons.

“It takes a bunch of time to draw a neural map by allowing computers to analyze an enormous number of signals found in the neural networks,” said Samsung Electronics. “The recent study proposes an innovative technology to activate memory platforms via measurement signals to facilitate the application of a neural network map.”

Hyun-Seok Lim lhs@donga.com