Samsung Electronics joined forces with leading Korean universities such as Seoul National University and the KAIST to develop a neuromorphic chip, a next-generation artificial intelligence (AI) semiconductor that resembles the human brain. The neuromorphic chip is a next-generation semiconductor that imitates human brain nerves. The chip can realize artificial intelligence functions such as deep running, and has an excellent non-standardized data processing capability such as images and sounds compared to conventional semiconductor. The chip consumes one 100 millionth of electric power used by current semiconductors. These facts make the chip a core technology that will determine the future of the semiconductor market.
According to related industries on January 16, the Engineering College of Seoul National University opened the Neural Processing Research Center (NPRC) last December with financial support from Samsung Electronics only. The center will carry out a mammoth-sized academic-industrial cooperation project involving 17 professors and 100 researchers from four universities -- Seoul National University, the KAIST, POSTECH, the UNIST (Ulsan Institute of Science and Technology). Choi Ki-young, a professor of Electrical Information Engineering at Seoul National University and a world-renowned authority in the field of low-power systems was appointed as director of the center and has been heading the center. The project attracts more attention because it is a new attempt which eluded Samsung’s practices that hardly disclose research and development projects.
The project aims to develop 17 core technologies to ultimately win global competition in neuromorphic chip market. The new neuromorphic chip is a semiconductor that realizes the way the human brain works. It is a future technology for which global information technology (IT) companies such as Apple, Intel and IBM are competing. Leading company IBM's neuromorphic chip, TrueNorth, has 260 million artificial neurons and evaluated to be close to the processing capacity of the bee brain.
Samsung Electronics will support 9 billion won for three years in the first phase. It is said that including the company’s financial support for some professors’ research in progress separately from Samsung, the amount of support will reach close to 10 billion won a year.
The human brain has about 10 billion neurons and about 10 trillion synapses connecting the neurons. This brain structure comes in handy for processing atypical data such as high-volume images, voices, and location information. Most big data which are the core material of the 4th industrial revolution such as AI, driverless cars and smart factories are none other than atypical data. It was pointed out that Samsung's artificial intelligence assistant, Bixby, should be 1000 times faster in processing data for perfect voice command processing. That is why the new neuromorphic chip is emerging as a next-generation semiconductor that processes more data with less energy.
Market researcher IDC predicted that the global AI market would surpass US$ 100 billion by 2022 although the size of the market stood at US$ 8 billion last year. The new neuromorphic chip is a key technology that will lead the growth of the AI market. Global companies such as Apple, Intel, Qualcomm and Huawei are rushing into the market of the new neuromorphic chip to preempt it. Apple loaded the AP A11 Bionic mixed with neural network technology into the iPhone 8 and the iPhone X launched late last year.
In addition to Korean universities such as Seoul National University, Samsung is stepping up its collaboration with the world's top research groups. The Korean IT giant is working with Canadian universities including the University of Montreal, McGill University, and the University of Toronto on fundamental AI technology and with American Universities including Harvard University and the University of Chicago on graphene and other next-generation materials. "It is an unprecedented project where companies and universities have banded together to develop next-generation semiconductors," an SNU official said.