In 2016, the world watched with both astonishment and unease as AlphaGo, an artificial intelligence developed by DeepMind, defeated Lee Sedol, one of the greatest Go players of our time; for many, this announced the arrival of a new era in AI research. Go, a strategy game that is more than 4,000 years old and long considered too complex for machines to master, had finally fallen to a program trained on data. More than a clash between human intelligence and machine computation, the 2016 match in Seoul marked a shift in the relationship between humans and machines and revealed an early glimpse of human-AI collaboration.
Go is an abstract strategy board game where players place black and white stones with the goal of fencing off more territory than the opponent. The game originated in China and has long been admired for the elegance of its simple rules set against the profound depth of its strategy. Unlike chess, where brute-force search allowed machines to defeat humans decades earlier, Go’s complexity resisted such methods. The number of legal positions on a Go board (estimated at 2.1 x 10 ^170) surpasses the number of atoms in the observable universe (roughly 10^80), making exhaustive calculation impossible, yet AlphaGo circumvented this barrier by combining neural networks with reinforcement learning. It was first trained on millions of human games, both amateur and professional, to capture intuition, and then moved on to millions of self-play games. The result was a machine capable not only of mimicking human play but also of transcending it. The now legendary Move 37 in the second game of the five-game match, initially dismissed as a blunder by experts, later revealed itself to be a move of extraordinary insight – one that no human had yet imagined.
Yet AlphaGo’s triumph came with complications. Many of its moves were derived from calculations that humans could not readily explain. In the context of a game, this opacity is tolerable, but it becomes deeply problematic in high-stakes domains like medicine or criminal justice, where the consequences of a wrong decision are too severe to accept uninterpretable models. This underscores why keeping humans in the loop is essential: to monitor AI systems, guard against bias, and ensure accountability.
While AlphaGo won four out of five games, Lee Sedol’s sole victory in Game 4 – secured by Move 78, later celebrated as the “Godly Hand” – reminds us that human creativity and intuition still have a place even against the overwhelming computational power of machines.
Instead of discouraging players, AI has reinvigorated Go. Both amateurs and professionals now study AI-generated strategies once thought unthinkable, demonstrating what AI can achieve across fields: not replacing human expertise, but augmenting it. As students of analytics, we find ourselves in a similar position. Just as Go players learn to refine their skills with AI, we must learn to collaborate with these tools to produce better results. Machine learning libraries and cloud platforms make advanced methods more accessible, but they cannot replace the uniquely human contributions of problem framing, interpretation, diverse perspectives, and ethical reasoning. These capabilities form the core of our curriculum at the Institute for Advanced Analytics, where technical rigor is matched by teamwork, storytelling, and ethical awareness.
Columnist: Maggie Lin