Korea Intelligent Information Society Agency AI Research Report Released
"Gap with Big Tech Exists... Policy Support Needed"
A research report has emerged stating that domestic artificial intelligence (AI) companies need to build data for 'Vertical AI,' an AI specialized in specific industrial sectors, in order to compete with capital-rich overseas big tech companies.
On the 24th, the National Information Society Agency (NIA) of Korea stated in its report titled "Changes and Challenges toward Vertical AI" that "to secure global superiority in the domestic AI industry, a shift in the direction of AI data construction projects and customized strategies are necessary," emphasizing the need to build Vertical AI data.
Vertical AI refers to AI that specializes in learning information from specific industries or fields, unlike the existing general-purpose large language models (LLMs) that learn from broad data. It advances beyond small language models (sLLMs) specialized in particular domains by utilizing deep data such as regulations and cultural norms to improve performance and reduce hallucination phenomena. It is used across various industries including law, healthcare, beauty, media, and content.
Vertical AI is expanding globally. According to recent data from the global market research firm CB Insights, $3.2 billion (10.2%) is invested in the Vertical AI sector within the global AI market, while $2.7 billion (8.5%) is invested in the horizontal AI sector. NIA analyzed that "as the industry enters a phase of technological advancement and expansion, it is moving toward maximizing AI utilization within existing vertical industry structures," adding, "this means that Vertical AI creates actual business value beyond technological innovation."
However, companies are facing difficulties in securing data. There are even cases where AI development is abandoned. The NIA report cited an example of an accounting firm. The firm developed an accounting AI service by training its data on an open-source AI model but failed due to insufficient performance. Improving performance required learning data in the relevant field, such as accounting standards, but the cost burden was significant.
The NIA stated that since the domestic AI industry has a gap compared to global big tech companies with capital and economies of scale, policy support is necessary. An NIA official said, "It is necessary to relax sandbox regulations to increase data utilization, such as allowing data use within data safe zones or expanding these zones." The official added, "Issues such as how to resolve copyright and cost problems when building data also need to be addressed."
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
