A Korean research team developed a semiconductor capable of implementing artificial intelligence (AI) in smartphones and the like. AI semiconductors are semiconductors with AI processing functions such as recognition, reasoning, learning and judgment, and optimized technologies based on super intelligence, ultra low power and super-strong reliability.
The Ministry of Science and ICT announced on February 26 that the research team led by professor Yoo Hoi-jun of the Korea Advanced Institute of Science and Technology (KAIST) and venture start-up UX Factory jointly developed the AI semiconductor.
In order to realize AI in a mobile environment, high-speed computing must be processed with low power. But now it is necessary to develop an AI acceleration processor due to the use of software technology which is slow in operation and consumes much electric power. The newly developed AI semiconductor is loaded with variable artificial neural network technology that adjusts energy efficiency and accuracy so it can process deep learning more efficiently.
To top it off, the research team widened the scope of application by developing an AI semiconductor that can learn not only various image objects but data that change with the passage of time.
Based on this, the team developed an emotional recognition system that recognizes seven facial expressions such as happiness, sadness, surprise, fear, and expressionlessness by recognizing human facial expressions through a smartphone camera and automatically displays them on smartphones in real time.
"This research is meaningful in that we have developed a semiconductor that accelerates while consuming less power to realize AI in mobile devices," professor Yoo said. "In the future, it is expected to be applied to various sectors such as object recognition, emotion recognition, motion recognition and automatic translation among others."