-Introducing artificial intelligence semiconductors equipped with self-learning NPU IP
-Securing source technology to advance on-device AI semiconductor technology
▲ Nepes participated in the 2023 Artificial Intelligence Semiconductor Future Technology Conference
Nepes(CEO B.K. Lee), a cutting-edge semiconductor back-end process foundry company, presented interim results of the artificial intelligence (AI) semiconductor research and development national project at the ‘2023 Artificial Intelligence Semiconductor Future Technology Conference’ hosted by the Ministry of Science and ICT on the 19th
This event is held every year, hosted by the Ministry of Science and ICT and organized by the Artificial Intelligence Semiconductor Forum, the Next-Generation Intelligent Semiconductor Business Group, and the Information and Communication Planning and Evaluation Institute to develop the AI semiconductor industry. In addition to exhibiting the AI semiconductor development achievements of participating companies, the Ministry of Science and ICT and domestic and foreign experts provide industry advice. We provide an opportunity to announce and discuss technology trends and major policy trends.
At this event, which marks its 4th anniversary this year, Nepes exhibited 'JANUS', which has a mobile self-learning neural network processing unit (NPU) IP for object classification, and a distributed learning IP driving system. In particular, ‘Janus’ is a lightweight self-learning NPU IP-embedded artificial intelligence semiconductor developed to meet the needs of the market where the need for ‘on-device learning’ for optimal reconstruction (relearning) of artificial intelligence models in user development environments is emerging. This has a wide range of application fields, from intelligent automobile systems such as smart keys and driver monitoring systems through new user authentication to intelligent factory fields such as field learning systems to improve defect recognition rates.
In April 2020, Nepes was selected as the lead organization for the ‘Development of mobile self-learning recursive neural network processor technology’ project in the ‘Next-generation intelligent semiconductor technology development project’ hosted by the Ministry of Science and ICT and is carrying out five years of research. This study aims to secure core technologies that enable lightweight semi-supervised learning and distributed learning by equipping devices with learning functions to optimize the privacy of personal and security data and the user environment. Many institutions, including Chungbuk National University, Hanyang University, and Seoul National University of Science and Technology, are collaborating. Based on its AI design technology, Nepes has not only secured on-device specialized AI semiconductor source technologies such as ‘mobile self-learning’ and ‘AI semiconductor lightweight’, but also developed on-device AI integrated platform technology.
▲ AI Chip with built-in lightweight self-learning NPU IP
“The on-device AI semiconductor market is blooming, and in particular, AI semiconductors that can learn at the device level have endless applications in our daily lives,” said Park Yeon-sook, head of Nepes. “The result of this project is the development of on-device AI technology.” “It will secure fundamental technology for advancement,” she said.
In addition, Nepes is developing its own cutting-edge technologies, such as 'Development of chiplet heterogeneous integration ultra-high-performance artificial intelligence semiconductor' with a total project cost of KRW 44 billion hosted by the Ministry of Science and ICT, and 'Development of core materials and process technology for 3D IC manufacturing using FOWLP' hosted by the Ministry of Trade, Industry and Energy. Based on packaging manufacturing technology, we are carrying out various national projects to develop core technologies for system semiconductor and AI semiconductor packaging.