260222 일시후원신청 == 유네스코한국위원회와 함께 만드는 미래.pdf

What is this year’s theme?

The theme of the 2026 birthday donation is: Reducing the Carbon Footprint of AI. This year’s project is not only about the speed of technological advancement, but also about asking a deeper question: How sustainable is the future that AI is creating?

1.mp4

All donations this year will be given as a one-time contribution to the UNESCO Korea “Future Response Fund.” The fund supports initiatives that address complex societal challenges such as climate change, social inequality, and the rapid advancement of science and technology.

The reason I decided to donate was a report published by UNESCO. The report, titled “Smarter, Smaller, Stronger: Resource Efficient AI and the Future of Digital Transformation,” presented experimental results on how to significantly reduce the energy consumption of artificial intelligence. It introduced three main approaches:

4.mp4

5.mp4

First, using Small Language Models (SLMs) instead of Large Language Models (LLMs)

The concept of Large Language Models (LLMs) is now familiar to us. However, not every situation requires a large and highly versatile model. There are also Small Language Models (SLMs), optimized for specific tasks. According to the report, using SLMs instead of LLMs reduced energy consumption by up to 90%. What’s remarkable is that despite this significant reduction in energy use, there was almost no loss in the accuracy of the results produced by the AI.

Second, compressing prompts and outputs

Many of us have experienced asking AI a question, only to receive an overly long answer— taking more time than a simple search would have. When AI generates unnecessarily long outputs, it requires more computation and therefore more energy. Simply compressing prompts and outputs to deliver only the necessary information reduced energy consumption by approximately 50%.

Third, applying quantization

Quantization is a method of adjusting the precision of AI outputs to only what is necessary. For example, the number pi can be expressed very precisely as 3.14159265358… but not every situation requires that level of precision. Reducing it to something like 3.14 when high precision is unnecessary can lower energy consumption by about 44%.

7.mp4

8.mp4

When I am caught up in daily life and simply trying to survive, I think I often lose my broader social perspective. After rushing through a busy year and beginning to prepare for my birthday donation project, I realize that it becomes a rare moment to think beyond myself — to think about the community again.

Over the past five years, my birthday has become an opportunity to ask myself: What can I do for the world? Each year, the messages from those who choose to participate have meant so much to me. Those messages have given me the strength to approach the project from new perspectives and to continue it in different ways.

I am equally grateful to those who participated through donations and to those who simply reached out with warm birthday wishes. My intention has always been to spend my birthday in a meaningful way — never to place any burden on those who celebrate it with me. And yet, each year I find myself wondering whether it might be misunderstood.