AI Rewires Elite Go Players' Thinking, Dominating Training a Decade After Milestone Victory

South Korea - Ekhbary News Agency

AI Rewires Elite Go Players' Thinking, Dominating Training a Decade After Milestone Victory

A decade ago, Google DeepMind's AlphaGo program sent shockwaves through the world by defeating South Korean Go champion Lee Sedol. Today, the landscape of this ancient and revered strategy game has been fundamentally transformed. AI is no longer just a challenger; it has become the dominant force in professional Go training, prompting players to confront a new reality where machine intelligence dictates the path to mastery.

Within the hallowed halls of the Korea Baduk Association in Seoul, the traditional quietude of Go strategy has been replaced by the hum of computers and the clicking of mice. Players, once deeply engrossed in the tactile feel of stones and the strategic depth of the physical board, now spend their hours hunched over monitors, replaying matches analyzed by sophisticated AI programs. Coaches meticulously compare their students' moves against AI suggestions, while some players sit in contemplative silence, observing AI programs engage in high-level play against each other. The game, steeped in centuries of human tradition and intuition, is undergoing an unprecedented evolution.

AI has not merely influenced the game; it has overturned established principles and introduced entirely novel strategic concepts. The pursuit of mastery now often involves replicating AI's moves with uncanny precision, even when the underlying logic remains opaque to human understanding. This shift has rendered AI indispensable for professional competition. While some lament a perceived loss of creativity and the erosion of traditional Go artistry, others see potential for new forms of human-AI collaboration and innovation within the game's framework. Simultaneously, AI is democratizing access to high-level training resources, a development that has coincided with a notable rise in the participation and success of female players.

For Shin Jin-seo, currently the world's top-ranked Go player, AI, specifically the program KataGo, is an indispensable training partner. His daily routine involves meticulously studying KataGo's recommendations, tracing the on-screen indicators that denote optimal moves. Nicknamed "Shintelligence" for the remarkable alignment between his play and AI's, Shin dedicates significant time to deciphering the machine's strategic choices. "I constantly think about why AI chose a move," he states, highlighting his commitment to understanding the AI's reasoning rather than blindly following its directives.

Shin's rigorous training regimen, largely centered around KataGo, is described as an "ascetic practice." Data from a 2022 study by the Korean Baduk League reveals that Shin's moves align with AI recommendations an impressive 37.5% of the time, significantly exceeding the 28.5% average observed among other professional players. "My game has changed a lot," Shin admits, "because I have to follow the directions suggested by AI to some extent." This adaptation underscores the profound impact AI is having on the very definition of professional Go play.

In a symbolic nod to the game's transformative journey, the Korea Baduk Association has reportedly reached out to Google DeepMind regarding a potential exhibition match between Shin and AlphaGo, to mark the tenth anniversary of its historic victory over Lee Sedol. While Google DeepMind has declined to comment on specific information, the prospect of such a match, pitting a modern, AI-honed player against the AI that once revolutionized the game, is a testament to the enduring narrative of human-machine competition and evolution in Go.

Go, an abstract strategy board game originating in China over 2,500 years ago, involves two players placing black and white stones on a 19x19 grid to surround territory. Its complexity is staggering, with an estimated 10^170 possible board configurations, far exceeding the number of atoms in the observable universe. If chess is a battle, Go is often described as a war, demanding intricate positional play and deep strategic foresight.

The development of AI to master Go involved feeding vast datasets of human moves into neural networks. AlphaGo was trained on 30 million human moves and further refined by playing against itself millions of times. Its successor, AlphaGo Zero, pioneered a 'tabula rasa' approach in 2017, learning the game solely from its rules and self-play, unburdened by human biases or conventional wisdom. This blank-slate method proved exceptionally powerful, leading AlphaGo Zero to decisively defeat its predecessor, AlphaGo Lee, 100-0 within days.

Following AlphaGo's retirement, open-source AI models inspired by AlphaGo Zero's success, such as KataGo, emerged. KataGo is now the preferred tool for many professional players due to its speed, advanced predictive capabilities—including ownership of board points—and its ability to grasp long-term strategy by analyzing the entire board, rather than just localized sections. Its objective is not merely to win, but to maximize the score, reflecting a sophisticated understanding of game efficiency.

This AI integration has fundamentally reshaped Go strategy. For centuries, players relied on heuristics to navigate the game's immense complexity. Elegant opening sequences and established principles, like avoiding early corner invasions, guided players. However, as commentator Park Jeong-sang notes, "AI has changed everything." Traditional common-sense moves are now obsolete, replaced by novel techniques pioneered by AI. The opening phase, once a canvas for personal expression and creativity, is now largely standardized, with players memorizing efficient AI-suggested sequences. The strategic focus has shifted towards the mid-game, where computational prowess takes precedence over imaginative flair.

This reliance on AI has led to a homogenization of playing styles. Chinese player Ke Jie has publicly expressed his weariness with the repetitive nature of AI-driven openings, likening it to a "tiring and painful" experience for both players and spectators. While fans may appreciate occasional deviations from the AI script, such moments are increasingly rare. A 2023 study indicated that over a third of moves by top players mirror AI recommendations, with the initial 50 moves often being identical across games.

Lee Sedol, reflecting on his post-AI career, laments the shift: "Go has become a mind sport... Before AI, we sought something greater. I learned Go as an art. But if you copy your moves from an answer key, that's no longer art." The game, for some, has transformed from an exploration of frontiers into a disciplined adherence to an oracle's dictates. Lee expresses a loss of purpose, stating, "My reason for playing Go has vanished."

Yet, the human element persists as players strive to adapt and innovate. Kim Chae-young, a top female player, describes the challenging process of unlearning ingrained intuition and rebuilding her strategic framework in response to AI's influence. "The intuition I had built up over the years turned out to be wrong," she shares, illustrating the profound cognitive recalibration required. Even top players like Kim and Shin admit to not fully grasping all of AI's moves, perceiving its strategic thinking as operating on a "higher dimension." Learning from AI, for them, is less about rational deduction and more about cultivating a new form of intuition to comprehend its advanced play.

Ekhbary News Agency