SEOUL, Feb. 26 (Korea Bizwire) — A group of South Korean scientists have developed an artificial intelligence-based processor that can be utilized for various deep learning technology, the ICT ministry said Monday.
According to the Ministry of Science and ICT, the team led by Yoo Hoi-jun of the Korea Advanced Institute of Science and Technology (KAIST) developed the neural network recognition processor that is more energy efficient compared with those in use today.
Noticeably, the team’s processor delivered fourfold higher efficiency compared with Google Inc.’s Tensor Processing Unit (TPU), considered a front runner in neural network computations. The new processor can also deliver both convolutional and recurrent neural networks simultaneously.
“It is meaningful that the research team developed the processor that can be activated at low power to realize the AI technology,” said Yoo, who is a professor of electrical engineering.
The research was funded by the ministry and was announced during the annual International Solid-State Circuits Conference.