The Korean Information Society Development Institute (KISDI, President Kwon Ho-yeol) published the KISDI Premium Report (21-03), titled An Analysis of Factors Hindering the Introduction and Spread of AI and the Policy Implications.

As the digital transformation and innovation competition is accelerating rapidly due to the COVID-19 pandemic, securing data and AI capabilities is becoming increasingly important as a major factor in determining the pace of development of countries and companies. However, there is a widespread perception that the introduction and spread of AI in Korea is still poor. Accordingly, in this report, KISDI analyzes the factors hindering the successful introduction and operation of AI, and the performance and limitations of the current government policy, and presents policy tasks designed to promote the introduction and spread of AI in the private sector.

First, it emphasizes that the low adoption and spread of AI is not a problem unique to Korea. According to an analysis of overseas cases by research fellows Lee Kyoung-sun and Kim Sung-ok, many companies that have actually introduced AI have either failed to introduce AI or are having difficulties in creating practical results despite investing considerable time and money. They point out that the problem is largely because of uncertainties caused by data dependency and Black box learning process of AIs. Companies that adopt AI must continuously upgrade and verify their AI models according to changes in the business environment due to data dependence and uncertainty. But this is a difficult task not only for SMEs but also for large enterprises, so many companies end up going through a protracted process of trial and error. In addition, the results of the analysis of domestic AI companies show that there are still many cases of companies experiencing difficulties in creating outcomes after introducing AI due to a lack of internal operating personnel, difficulties in verifying results, difficulties in upgrading models, and a high dependence on suppliers.

In this regard, the report suggests that it is necessary to gradually expand the government projects, which are currently focused on companies in the early stages of introduction, in a direction that supports staged growth in order to improve the performance of government projects aimed at the introduction and spread of AI. This is because the current government support projects have limitations as regards resolving the difficulties of companies that are experiencing trial and error after introducing AI, even though such projects - when carried out in the form of matching between supply and demand companies and providing funding support for each AI element - do contribute to resolving the difficulties (cost burden, data acquisition and pre-processing, model development) of companies in the early stages of AI introduction, as well as to the formation of the domestic AI market. In particular, this report suggests the establishment of a step-by-step level diagnosis system and the promotion of advancement support projects as more and more companies adopt AI in the future with benefit from government projects and the need for the implementation of policies that will lead their step-by-step growth will increase. The mission is to create success cases of AI introduction that can trigger the introduction and spread of AI in various fields by building a level diagnosis system for companies that adopt AI based on the evolutionary stage of AI introduction, and by supporting companies with potential in their data and AI capabilities internalization and staged sophistication.

In addition, it is suggested that the government should nurture various layers of the AI service industry and present guidelines for AI technology trading. Currently, as AI technology is being made into a platform and generalized, and as segmentation of the service market is progressing, it is expected that the barriers to the introduction of AI technology will gradually be lowered in the future. However, if the introduction of AI by companies with a low technological level is expanded as such, the possible risks from data/algorithm bias/discrimination, AI platform lock-in, solidification of the gaps between companies depending on the level of service used etc., as well as the risk of disputes related to responsibility for the quality of AI results between participants in the ecosystem, entitlement, fair trade, etc., could intensify considerably.

KISDI Research Fellows Lee Kyoung-sun and Kim Sung-ok emphasized that although the domestic legislative environment is being created to prevent these side effects of AI, support for technology development and services is required in order to overcome the limitations of AI in Korea in consideration of the overseas situation, where the technology service industry is being established in order to resolve the uncertainties and dysfunctions surrounding AI. Furthermore, they argue that it is necessary to prepare guidelines for AI technology transactions to allow companies considering the introduction of AI to recognize the possible risks in advance and refer to them in resolving disputes, in order to create a safe and fair AI technology trading environment. This is because the role of the government in supporting the smooth and sound operation of the domestic AI ecosystem and the associated introduction and proliferation of AI in the private sector is extremely important at the present time.