The government is risking everything to secure GPUs, which are considered the core of artificial intelligence (AI). Acting President Choi Sang-mok held the '3rd AI Computing Infrastructure Special Committee' on the 17th and announced plans to secure 10,000 state-of-the-art graphics processing units (GPUs) within this year and open the National AI Computing Center early. This decision seems to reflect a sense of crisis that if the early establishment of AI infrastructure is not successful, the country may fall behind in the AI ​​competition.

However, field assessments have shown that it is dangerous to think that large-scale language models (LLMs) such as DeepSec or Chat GPT can be produced just by securing GPUs. It is explained that it will be difficult to leap to AI G3 without a solid foundation in various fields such as cloud computing, coding, nurturing talent in mathematics, and basic science.

On the 17th, the National Life Science Advisory Group, the Federation of Korean Science and Technology Societies, and the Korean Academy of Science and Technology held an emergency joint forum on the topic of 'DeepSec Wave and Future Prospects.' 

(...)

Jaesik Choi, CEO of AI prediction technology solution company INEEJI, explained that the background behind China's creation of LLMs such as DeepSec was the fostering of programmers who understand hardware and application companies. He said, "In 2016, during the AlphaGo match, engineers at Microsoft Beijing Research Institute created a deep learning model called ResNet. It is evaluated to be more accurate than the model created by Google," and "We gave up on training technology for nearly 10 years, but in China, the technology and foundation for training were created, and I think that led to the current LLM model."

Regarding the recent budget direction focusing on securing GPUs, he said, "Tens of thousands of GPUs are needed for AI research, but the foundation of the AI ​​industry is cloud computing." He continued, "80% of the cloud service market in Korea is imported, and cloud infrastructure is insufficient. In addition to securing GPUs, support should be evenly provided in various fields such as the foundation for domestic cloud companies to grow and increasing cloud usage. 

"All news articles shared on our website are excerpts provided solely for informational purposes and remain the intellectual property of their original publishers.

Reproducing, copying, or redistributing this content without prior consent from the original publishers is strictly prohibited."