A 2016 Go AI placed a puzzling stone on the fifth line, sparking debate on whether large language models truly reason or merely mimic patterns.

In March 2016, an AI program placed a stone on the fifth line of a Go board during game two of a five‑game match in Seoul, a move that highlighted the limits of machine reasoning.

The Seoul Match

The match was part of a high‑profile series that pitted the AI against a top human player. On the afternoon of the game, the program’s creator watched the AI put a stone at move 37, a position that appeared to give the human opponent a free gift.

Why the Move Stood Out

Commentators described the placement as "absurd" and questioned whether the AI had erred. In a later essay, the program’s developer wrote that the move was intentional, noting that the AI evaluated the position as strategically sound despite its unconventional appearance.

Implications for AI Reasoning

The incident has been cited as evidence that current large language models and game‑playing AIs do not reason in the way humans do. Researchers argue that the AI’s decision‑making relies on pattern recognition and statistical evaluation rather than logical deduction.

One analyst said the episode underscores the need for clearer explanations of how AI systems arrive at unexpected actions. The developer confirmed that the AI’s evaluation function assigned a positive win probability to the fifth‑line stone, even though human experts initially perceived it as a mistake.

The episode continues to inform discussions about AI transparency and the gap between human intuition and algorithmic prediction. As AI systems become more integrated into strategic domains, understanding the basis for seemingly irrational moves remains a priority for both developers and policymakers.