Working with scientists from Harvard University, Samsung is working on a future project involving the human brain.
Samsung Electronics has just shared a new perspective on the possibility of realizing a neural chip that mimics the human brain. Released as a Perspective paper, the document showcases the visions of the world’s leading engineers and scholars on cutting-edge semiconductor technology.
Name Neuroelectronics based on brain replication (Neuromorphic electronics based on copying and pasting the brain), the document is the scientific research of leading figures from Samsung and Harvard University.
The list of authors includes Professor, Samsung Advanced Institute of Technology (SAIT) member Donhee Ham, Harvard University Professor Hongkun Park, Samsung SDS President and CEO, former SAIT Director Sungwoo Hwang, Vice President President and CEO of Samsung Electronics Kinam Kim.
Image of mouse neurons on CNEA (CMOS nanoelectrode array). Photo: Samsung.
According to Samsung, the essence of this document is summed up in two words “copy” and “paste”. To replicate the nervous system of the brain, the thesis suggests using a groundbreaking nanoelectrode array developed by Professor Ham and Professor Park. The system is then glued onto a three-dimensional high-density lattice of semiconductor memory, technology in which Samsung is leading the world.
Through a copy-paste approach, the authors envision a memory chip that has unique computing characteristics of the human brain such as low energy use, easy learning, and environmental adaptability. The chip even has autonomy and awareness, which puts it beyond the reach of current technology.
In essence, the human brain is made up of a large number of neurons, connected by a nervous system to perform functions. So, gaining knowledge about the system will be the key to being able to reverse engineer the brain.
Introduced in the 1980s, the original goal of neuromorphic computing was to mimic the structure and function of a neuronal system on a silicon chip. However, the work immediately ran into difficulties because until now, how the neurons were reconnected, working together was still a question mark.
As a result, the goal of neural computing has been lowered to design a chip that is “inspired” by the brain rather than rigidly mimicked.
The Samsung document suggests looking back at the original goal of neural computing, which was to reverse engineer the brain. The nanoelectrode array can effectively penetrate a large number of neurons, recording their electrical signals with high sensitivity. These massive intracellular records show the nervous system, where neurons connect to each other, and the strength of these connections. From the transcript, a nervous system can be extracted or ‘copy’.
The system can then be “pasted” into a network of electrostatic memories, such as commercial flash memories (SSDs) or memories such as RRAM (resistive RAM). In it, the conductivity of each memory is programmed to represent the strength of each neuron connection in the system.
Samsung releases research papers with bold aim. Photo: Samsung.
The document goes a step further, proposing a method for fast gluing neural systems to memory networks. When driven by recorded signals from the intracellular matrix, special electrostatic memory networks can learn and represent the neuronal system. This diagram will load the brain’s system directly onto the memory chip.
The human brain is estimated to have about 100 billion neurons, adding about a thousand other neural connections. So the neural chip will eventually need 100 trillion memory. Integrating large amounts of memory on a single chip may be possible using integrated 3D memory. This is the technology that has opened a new era for the memory chip industry in which Samsung currently has the upper hand.
Leveraging on its experience in chip manufacturing, Samsung plans to continue its neural computing research, building a position in the field of next-generation AI semiconductors.
“The vision we set is very ambitious. But striving for such a courageous goal will push the boundaries of both machine intelligence, neuroscience and semiconductor technology,” Professor Ham said.
Shocking Elon Musk, monkeys play games with their thoughts after having a chip implanted in their brain
Neuralink, the ambitious project of tech billionaire Elon Musk to implant computer chips into human brains, has taken an important step forward when successfully implanting chips into monkey brains.