The Future That Came First

Written while heading to Goseong Ironman 70.3 in June 2026.

A book that surveys how the go world has changed since AlphaGo in 2016. LLMs in programming still can’t solve problems as well as an exceptionally skilled human can. But AlphaGo, in go, always wins against even the best human players in the world. When the day comes that AI surpasses humans in every respect, and humans can no longer even understand the moves the AI makes, how will the industry change?

According to this book, after AlphaGo, the definition of a “strong player” shifted to mean “a player who plays like AI,” and some people welcomed this change while others disliked it. However, players who “didn’t play like AI” were quickly pushed out of the rankings, and eventually players who “played like AI” came to form the mainstream. In other words, the freedom to “not play like AI” no longer exists.

For this reason, a person’s skill is now judged by how similarly they play to AI. Since the game of go inherently requires the premise that “a human is the one playing,” it seems things flowed in the direction of humans imitating AI. If that’s the case, what will happen to programming going forward, where there’s no particular need for a human to do it directly?

Of course, since MCTS-based AlphaGo and LLMs have different architectures, the point at which AI’s programming capability surpasses that of an exceptional human will come somewhat later. But since we never know when the situation might flip, it’s a question worth thinking about in advance, just in case.

Passages I Found Impressive

What’s the problem with AI being better at go than humans? Cars run faster than humans, but doesn’t track and field still have various running events? Have runners’ incomes or pride been damaged because of cars? …

First, even before cars appeared, there were many animals that ran faster than humans. Second, many people believe that “being human” has more to do with the capacity for thought than physical ability. Third, cars did not change the way humans run. Fourth, humans do not learn how to run from cars.

In that it’s a cross-border problem that nation-states cannot solve on their own, the challenge of public control over technology resembles the challenge of responding to the climate crisis. … Technology control and climate crisis response are also alike in that we realized the need for them far too late. Both technology control and climate crisis response struggle to win people’s support easily, because both hinder economic growth.

… Even after new drugs are developed or large construction projects are completed, they are monitored by government agencies for a certain period to see what actual effects they have on people. … Many people take this for granted. Isn’t this what “normal” should look like?