South Korea's government, which has declared itself one of the top three nations in artificial intelligence (AI), is pursuing numerous related initiatives. However, more than half of these projects are being conducted in a survival format, leading to increased fatigue among companies. While the government aims to draw public interest, critics argue that the focus on survival-style competitions detracts from meaningful engagement in AI development.
According to industry sources, since the start of Yoon Suk Yeol's administration, the Ministry of Science and ICT has initiated at least nine competitive AI R&D, service, and contest projects.
Among these, five projects are explicitly designed to reduce the number of participating teams through a phased evaluation process after initial support. These include the Independent AI Foundation Model (IndepAI), AX Innovative Company Creative Technology Development, AI Global Big Tech Cultivation, AI Humanoid, and K-Cloud Technology Development. Three projects are conditional support types, depending on whether they meet specific criteria, and one is a contest format.
A notable compressed project is the 'IndepAI,' first announced in July of last year. Five elite teams, including Naver Cloud, SK Telecom, LG AI Research, Upstage, and NC AI, participated. Naver Cloud and NC AI were eliminated, while Motif Technologies joined the second phase evaluation through an additional application. The final results of the second phase evaluation will be announced next month.
In September of last year, the 'Specialized Foundation Model Project' for the medical and bio sectors and the 'Public AX Project' supporting AI in the public sector were launched. In January, the 'AX Innovative Company Creative Technology Development' and 'AI Global Big Tech Cultivation' projects began, and recently, a call for proposals for 'Everyone's AI,' aimed at creating AI services for all citizens, is underway.
The goals of these projects vary, focusing on general models, specialized models, and public service development. However, the support for graphics processing units (GPUs), consortium formation, and phased evaluation processes are similar. As a result, while the content of the projects differs slightly, the same companies are often involved. Companies participating in multiple projects must prepare similar project proposals and respond to evaluation performance verification.
Small AI companies, which often lack sufficient personnel and funding, face increased pressure as they must manage both technology development and evaluation responses. The Ministry of Science and ICT maintains that a competitive approach is optimal given the limited budget and the need to concentrate GPU resources on competitive firms.
A ministry official stated, "In a situation where resources such as budget, GPUs, and research personnel are limited, selecting capable project operators through competition is unavoidable. The government's role is to bear the risks until a market is established, after which private entities can grow through competition."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.