Settings

ⓕ font-size

  • -2
  • -1
  • 0
  • +1
  • +2

Samsung presents vision for brain-like neuromorphic chips

  • Facebook share button
  • Twitter share button
  • Kakao share button
  • Mail share button
  • Link share button
Image of rat neurons on CMOS nanoelectrode array / Courtesy of Samsung Electronics
Image of rat neurons on CMOS nanoelectrode array / Courtesy of Samsung Electronics

By Kim Bo-eun

Samsung Electronics introduced a new approach to reverse engineer the brain on a memory chip in a paper co-authored with Harvard University researchers that was published in the journal Nature Electronics.

The paper titled, "Neuromorphic electronics based on copying and pasting the brain," was authored by Samsung Advanced Institute of Technology fellow and Harvard University professor Ham Don-hee, Harvard University professor Park Hong-kun, Samsung SDS CEO Hwang Sung-woo and Samsung Electronics Vice Chairman Kim Ki-nam.

The paper suggests a way to copy the brain's neuronal connection map using a nanoelectrode array developed by Ham and Park, and to paste this map onto a high-density three-dimensional network of solid-state memory chips.

Through this copy and paste approach, the authors envision creating a memory chip that approximates the computing traits of the brain, such as low power, facile learning, adaption to the environment and autonomy and cognition, which have been beyond the reach of existing technology.

The brain is comprised of a large number of neurons and their wiring map is responsible for the brain's functions. Knowledge of this map is therefore key to reverse engineering the brain.

Neuromorphic engineering was launched in the 1980s. The original goal was to mimic the structure and functions of the brain's neuronal networks on a silicon chip. But this proved to be difficult because, even until now, little is known of how the large number of neurons are wired together to create the brain's higher functions.

Given this barrier, the goal of neuromorphic engineering had been eased to designing a chip inspired by the brain rather than something mimicking it.

The latest paper, however, suggests a way to return to the original neuromorphic goal of brain reverse engineering. The nanoelectrode array can effectively enter a large number of neurons so it can record their electrical signals with high sensitivity. These massively parallel intracellular recordings inform the neuronal wiring map, indicating where neurons connect with one another and how strong these connections are. From these recordings, the neuronal wiring map can be extracted.

The copied neuronal map can then be pasted onto a network of non-volatile memory chips, such as commercial flash memories used in daily life in solid-state drives or new memory technology, such as resistive random access memories (RRAM), by programming each memory chip so that its conductance represents the strength of each neuronal connection in the copied map.

The paper goes a step further and suggests a strategy to rapidly paste the neuronal wiring map onto a memory network. A network of specially-engineered non-volatile memory can learn and express the neuronal connection map when directly driven by the intracellularly recorded signals. This is a scheme that directly downloads the brain's neuronal connection map onto the memory chip.

Since the human brain has an estimated 100 billion or so neurons, and a thousand or so times more synaptic connections, the ultimate neuromorphic chip will require a capacity that can accommodate 100 trillion virtual neurons and synapses. Integrating such a vast number of memory addresses on a single chip would be made possible by 3D memory integration technology. Samsung is the leader in the field of 3D memory integration technology.

"The vision we present is highly ambitious," Ham was cited as saying in a press release. "But working toward such a heroic goal will push the boundaries of machine intelligence, neuroscience, and semiconductor technology."


Kim Bo-eun bkim@koreatimes.co.kr


X
CLOSE

Top 10 Stories

go top LETTER