Advanced economies are competing to expand national research and development (R&D) investment amid the U.S.-China battle for technological supremacy, the artificial intelligence (AI) revolution, and the rise of national security and technological sovereignty. The United States is making large-scale investments in semiconductors, AI, and quantum technology, while China is pouring vast resources into achieving technological self-reliance. Korea, too, has continuously emphasized the need to expand its R&D budget to secure national innovation competitiveness.
Some raise a fundamental question: Can innovation competitiveness be secured through expanded R&D investment alone in the AI era? Because innovation competition in the AI era involves broad shifts and intense speed competition, how to rapidly create competitive innovation outcomes matters more than simply expanding investment. Yet the current research system faces a structural problem of continuously declining research productivity.
A research team led by Nicholas Bloom, an economist at Stanford University in the United States, demonstrated through decades of data analysis in 2020 that research productivity has been continuously declining. The team analyzed that today's innovation economy requires exponentially more cost and effort to produce the same level of innovation outcomes compared to the past. Put simply, more is being invested, but generating results is becoming increasingly difficult.
For example, in the semiconductor field, Moore's Law (the principle that microchip performance doubles every two years) is observed, but the research workforce required to maintain it has increased more than 18-fold compared to the early 1970s. This means that the innovation once accomplished by one researcher in the 1970s now barely takes 18 people. Similar patterns are found in agriculture and medicine. The reason is that discovering new ideas is becoming increasingly difficult compared to the past. In other words, the low-hanging fruit has already been picked, leaving only far more complex and difficult tasks that require advanced technology.
In fact, as technical difficulty has risen sharply in major fields such as new drugs, advanced materials, and robotics, the scale of research required for a single innovation has grown to a degree incomparable to the past. AI technology is no exception. Developing cutting-edge AI models requires astronomical computing resources and research personnel. Expanding investment is essential, but it has become an era in which investment alone cannot guarantee innovation.
Recently, as AI technology is being actively used in research activities, the possibility of dramatically improving productivity across the entire R&D process — from literature searches, data analysis, and experiment design to drug candidate discovery, materials exploration, and software development — is growing. This raises hopes that the use of AI could be a breakthrough that halts the decline in research productivity. However, there are still considerable limitations and constraints on productivity improvements using AI technology.
Although AI technology is advancing rapidly, it is still imperfect and relies on learning from existing data, leading to assessments that it has limits in creating disruptive innovation. On top of this, the institutional rigidity and inefficiency inherent in the existing research system act as major constraints. In particular, the experience of the information technology (IT) era shows that productivity innovation is possible only when the introduction of new technology is accompanied by organizational and institutional innovation.
Research productivity innovation in the AI era must not stop at improving the performance of individual research projects or individual researchers, but should be approached from the perspective of enhancing the efficiency of the entire national R&D system. In particular, the system must be redesigned to create greater innovation with the same resources, spread results faster, and reduce unreasonable institutional costs.
Recently, major advanced countries have begun to focus on policy innovation to improve research productivity, moving beyond competition over R&D investment scale. A representative example is the meta-science policy being developed mainly in the United States and the United Kingdom. It can be described as a new policy domain that scientifically studies the research system itself based on data. It scientifically analyzes which research funding methods produce higher performance, how research evaluation systems should be designed, and under what conditions collaborative research is effective. In other words, it treats R&D policy itself as a research subject, attempting to accumulate evidence on what is effective and reflect it in policy.
Korea's policy discussions remain relatively concentrated on expanding the scale of R&D investment. Korea's national R&D investment scale is already among the world's highest relative to gross domestic product (GDP), but systematic analysis of and policy approaches to research productivity are lacking. In particular, there is no attempt to ask the fundamental question of whether the current research system is effective. Researchers' administrative burden, a project-centered management system, and an evaluation system focused on short-term results are problems that have long been raised, yet fundamental improvements have not been made.
In a situation where low birth rates and fiscal constraints are deepening, the input-driven growth strategy of the past is not sustainable. Research personnel cannot be expanded indefinitely, and there are limits to continuously increasing government finances. Going forward, what determines a nation's innovation competitiveness will depend not on who invests more resources, but on who creates knowledge and realizes innovation more efficiently. It is time for Korea's science and technology policy, too, to shift from a paradigm emphasizing expanded investment to one emphasizing productivity innovation.
