[Gettyimages Bank]
For the efficient development of AI-related technologies based on logic similar to that of a human brain, computer systems with powerful graphic processing units (GPUs) are required. GPU is a specialized electronic circuit that is designed to rapidly process massive amounts of memory to accelerate computing capabilities.
GPU is widely used by researchers to speed up the computational process of deep learning, an essential part of modern AI infrastructure. However, many SMEs and small research centers cannot afford such equipment. High-performance computers consume a lot of energy and many researchers have to rebuild their power supply system before installing a new cluster of supercomputers.
The Ministry of Science and ICT said in a statement on March 8 that Naver Cloud was selected to offer up to 80 teraflops a month for 1,000 SMEs, ventures, public organizations and graduate schools. A teraflop rating measures the performance of a GPU. A 1-teraflop computer system can perform one million floating operations per second.
"Through our high-performance computing resource-sharing project, we will create an environment where anyone with innovative ideas can easily and quickly develop AI technologies," said Song Kyung-hee, who heads the science ministry's AI policy team. Through an earlier project, 883 SMEs and organizations saved an average of 3.9 months in development time and a monthly average of five million won ($4,427) in operating costs.
reporter
Park Sae-jin
swatchsjp@ajunews.com
Copyright ⓒ Aju Press All rights reserved.
