Is the Government Looking to Reorganize Its Sovereign AI Program Around a “Frontier AI” Model? SPC Formation Also Mentioned
IT DAILY ·
✦ AI Summary
It was reported that the government is reviewing the operating direction of its independent artificial intelligence (AI) foundation model development project.
So far, the project has operated by providing GPU and data to multiple elite teams and selecting competitive models through step-by-step evaluation.
The Ministry of Science and ICT said it recognizes the need to concentrate funding on developing top-tier AI models, but that it has not yet decided whether to discontinue the independent AI foundation model project or establish an SPC.
According to IT Dailly, the government’s planned direction for the development project for an independent artificial intelligence (AI) foundation model, or independent foundation model, may be changing. As of the 29th, industry sources said the government was reviewing how to operate the project.
The purpose of pushing ahead with the independent foundation model project is to secure a global-level AI foundation model. So far, the project has been carried out under an operating structure that provides GPU and data to multiple elite teams.
Selecting competitive models through step-by-step evaluation has also been considered a core part of the project’s operating structure. In other words, the government has been supporting and evaluating multiple teams to identify promising models.
Recently, however, AI model performance has been advancing rapidly, mainly in the United States and China. That has raised questions about whether the existing competition- and evaluation-centered approach alone can narrow the gap with global frontier models.
Amid this trend, a plan to establish a special purpose company (SPC) for developing frontier AI models is also being discussed alongside the government’s review of the project’s operating structure. As a result, expectations are growing that the direction of the project could change.
A frontier model refers to an AI model at the top end of current performance and applicability. That benchmark is continually raised as technology advances. Major AI companies such as OpenAI, Anthropic, and Google DeepMind use the term 'frontier model' when describing their latest top-tier models.
Among domestic models, Motif Technologies' 'Motif 3' is being cited as a relatively high-performing case. In last month's evaluation by Artificial Analysis' 'Artificial Analysis Intelligence Index (AAII),' Motif 3 scored 47 points. Motif 3 received the highest score among the four models participating in the project, but the gap with global top-tier models remained.
Against this backdrop, the government has expressed concern over the existing evaluation-and-competition approach for the project. The second Science, Technology and AI Future Strategy Meeting was held on the 27th of last month. At the meeting, Deputy Prime Minister and Science and ICT Minister Bae Kyung-hoon said that, in relation to the results of the project's second evaluation, they were still disappointing when compared with global standards.
The Ministry of Science and ICT distributed a briefing document the previous day and said the government recognizes the need to concentrate funding on the development of top-tier AI models. However, the ministry said it has not yet been decided whether to discontinue the independent AI foundation model project tied to this effort, or whether to establish an SPC for the development of top-tier AI models. In other words, the ministry drew a distinction between recognizing the need for support and deciding on individual projects and organizational plans.
Source: IT DAILY · Yang Seung-gap
Original: https://www.itdaily.kr/news/articleView.html?idxno=241865
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.