Op-Ed
[Op-Ed] The Next Law of AI Competition: Wisdom, Not Size
Eunho Yang, Professor at the Kim Jaechul Graduate School of AI at KAIST and lead researcher for NAIRL’s sub-project “Beyond Neural Scaling Laws,” writes that the scaling law still holds, but the cost of following it has grown unsustainable — turning AI into a race only a few companies and countries can run.
The question, he argues, must change: not “how big can we build,” but “how much intelligence can we deliver with how few resources, and how reliably.” Researchers across NAIRL’s universities pursue this through model compression, efficient learning algorithms, new architectures, and data curation, while his own lab contributes quantization, ensembles, inference acceleration, and explainability research.
This arena is not stacked against Korea, Professor Yang concludes — it is decided by ideas and people, and the country that first sees beyond the scaling law will write the rules of the next game.
This column is part of NAIRL’s biweekly series “AI Talent Powerhouse,” published in ZDNet Korea.