Samsung Electronics Unveils Vision for “Copy and Paste” the Brain onto Neuromorphic Chips – Samsung Newsroom UK

0


With Harvard Researchers, Samsung presents a new approach to reverse engineering the brain on a memory chip in a Perspective paper published in Nature Electronics

Samsung Electronics, a global leader in advanced semiconductor technology, today shared a new insight that takes the world of realizing neuromorphic chips that can better mimic the brain one step closer.

The findings, drafted by leading engineers and scientists from Samsung and Harvard University, have been published by Nature Electronics as a Perspective Paper entitled “Neuromorphic Electronics Based on Copying and Pasting the Brain”. Donhee Ham, Fellow of Samsung Advanced Institute of Technology (SAIT) and Professor of Harvard University, Professor Hongkun Park of Harvard University, Sungwoo Hwang, President and CEO of Samsung SDS and former head of SAIT, and Kinam Kim, Vice Chairman and CEO of Samsung Electronics are the corresponding authors.

Image of rat neurons on CNEA (CMOS nanoelectrode array).

The essence of the authors’ vision can best be summed up in the two words “copy” and “paste”. The paper suggests a way to map the brain’s neural connection using a method developed by Dr. Ham and Dr. Park developed groundbreaking nanoelectrode arrays to copy and paste this card into a high-density, three-dimensional network of solid-state memories, the technology for which Samsung led the world.

Through this copy and paste approach, the authors envision developing a memory chip that approximates the unique computational properties of the brain – low power consumption, easy learning, adapting to the environment, and even autonomy and cognition – that cannot be achieved with current technology could .

The brain is made up of a large number of neurons, the wiring diagram of which is responsible for the functions of the brain. Hence, knowing the map is key to reverse engineering the brain.

While the original goal of the neuromorphic technique, which was introduced in the 1980s, was to mimic such structure and function of neural networks on a silicon chip, this proved difficult as little is known to this day, such as the large number Neurons are wired together to create the higher functions of the brain. This made it easier for neuromorphic technology to develop a chip “inspired” by the brain, rather than rigorously mimicking it.

This paper suggests a way to return to the original neuromorphic goal of reverse engineering the brain. The nanoelectrode array can effectively penetrate a large number of neurons so that it can record their electrical signals with high sensitivity. These massively parallel intracellular recordings inform the neural wiring map, indicating where neurons are connected to each other and how strong those connections are. Hence, the neural wiring map can be extracted or “copied” from these telltale records.

The copied neural map can then be “pasted” into a network of non-volatile memories – for example commercial flash memories, which are used in solid-state drives (SSD) in our everyday lives, or “new” memories such as resistive random access memories (RRAM ) – by programming each memory so that its conductivity represents the strength of each neural connection in the copied map.

(from left) Donhee Ham, Fellow of the Samsung Advanced Institute of Technology (SAIT) and Professor of Harvard University, Hongkun Park, Professor of Harvard University, Sungwoo Hwang, President and CEO of Samsung SDS (former head of SAIT) and Kinam Kim, Vice Chairman and CEO of Samsung Electronics, the corresponding authors.

The paper goes a step further and suggests a strategy to quickly insert the neural wiring card into a storage network. A network of specially designed non-volatile memories can learn and express the neural connection map when driven directly by the signals recorded intracellularly. This is a scheme that the brain’s neural connection map downloads directly to the memory chip.

Since the human brain is estimated to have around 100 billion neurons and around a thousand times more synaptic connections, the ultimate neuromorphic chip requires around 100 trillion memories. Integrating such a large number of memories on a single chip would be made possible by 3D memory integration, the Samsung-led technology that opened a new era for the memory industry.

Leveraging its leading edge in chip manufacturing, Samsung plans to continue its neuromorphic engineering research to develop Samsung’s leadership position in next-generation AI semiconductors.

“The vision we are presenting is very ambitious,” said Dr. Ham. “But working towards such a heroic goal will push the boundaries of machine intelligence, neuroscience, and semiconductor technology.”


Share.

Comments are closed.