Options for an unknown future

Lee Jeong-dong 

The author is a professor of the Technology Management, Economics and Policy Program at Seoul National University’s College of Engineering. 

In fall 2012, University of Toronto professor Geoffrey Hinton and two graduate students entered a program in an image-recognition competition, trained on two gaming graphics cards. AI hardware shifted toward GPUs. Chips designed for games became core AI infrastructure, and Nvidia became one of the world's most valuable companies.

In 2017, eight Google researchers published “Attention Is All You Need” (2017). It began with machine translation, but its transformer architecture became the foundation of today's large language models.

AI is not unique. Reusable rocket boosters were long dismissed as reckless but have become a rule of the space industry. In batteries, solid-state technology without liquid electrolytes and sodium-ion cells replacing lithium are competing to define what comes next. Today's throne was yesterday's fringe.

The problem begins now. Transformers and GPUs have themselves become conventional wisdom. Once an idea becomes conventional, an ecosystem forms around it. Semiconductor designs, data centers, power grids, investment plans and university curricula are built on that assumption. Conventional wisdom opens one path while erasing others.

The natural-ice industry that dominated the 19th century responded to artificial ice by developing better steam-powered saws and storage. It worked harder toward a future it already knew. Dominant paradigms can hold even the most capable people until the end.

Today's assumptions about AI cannot continue indefinitely. Global data center electricity consumption is projected to rise from 415 terawatt-hours in 2024 to 945 terrawatt-hours in 2030 and exceed 1,200 terrawatt-hours by 2035. That is more electricity than Japan uses annually. Given such costs, today's road map cannot be sustainable. Yet everyone races down the same road.

Irreplaceable technologies are difficult to develop on a playing field someone else created. When others write the rules, we can only become excellent students who follow them. Irreplaceability emerges when the field itself changes.

But nobody knows what that game will be. Rather than predicting the future using keywords supplied by others, we must test possibilities and expand our options.

I recently attended a “Grand Quest” presentation at a science high school. For more than two years, the program had encouraged students to imagine beyond textbooks and conventional wisdom. They posed problems without answers and refused to retreat. Their voices carried conviction.

I left encouraged. A generation free from the memory of catching up was imagining the next playing field.

This cannot remain the story of one science high school. Research laboratories and technological competition need the same space. Researchers should be encouraged to ask what comes after textbooks and conventional wisdom. Solving problems already on a road map does not easily produce new questions. Countries that have never practiced asking questions fall behind when the rules change.

The government is pursuing a future response fund using additional tax revenue from the semiconductor boom. It plans investment in young people, growth engines, regions, education and talent. All are necessary, but one more element should be added.

Even when discussing ambitious research, we tend to make known problems harder — climbing a familiar mountain higher and faster. What we need instead are challenges that change the terrain and redraw the road map. The character of the funding matters more than its size. The question is not whether such projects receive 5 percent or 20 percent, but what kind of 5 percent we create.

Why not establish a window within the fund for such efforts? Being absent from the road map should become a reason for support rather than rejection. The money should test questions whose chances of success are difficult to demonstrate, while ensuring even failures pass experience and data to the next challenger.

A country with no door for supporting what was never planned is unprepared for an unplanned future. The fund should not try to pick tomorrow's winners. It should expand tomorrow's options. Corporate and market money flows toward known futures. Public money is needed most for futures that do not yet have names.

Technology does not leap. But the playing field changes. Behind every change lies a question asked long before. Gaming chips, transformers and reusable rockets all began that way.

I still remember the students' voices: “I can't see the answer, but I'll try.” The fund should nurture that voice and attitude. Expanding our options for reaching futures still unknown is what preparing for the future means.