In this blog post, we’ll examine how close artificial intelligence can come to human intuition through the match between AlphaGo and Lee Sedol.

Lee Sedol vs. AlphaGo: The Showdown Between Human and AI

One of the biggest stories of 2016 was the match between 9-dan Lee Sedol and Google DeepMind’s AlphaGo. Lee Sedol 9-dan was hailed as the strongest player in the Go world, while Google DeepMind was a subsidiary of Google, one of the world’s most innovative IT companies. The clash between these two entities went beyond a simple Go match, sparking widespread interest by creating the traditional yet fascinating narrative of “computer versus human.” This match was seen as an opportunity to test the limits of existing artificial intelligence and sparked curiosity about the capabilities of both sides’ technologies and the limits of human ability.
The match consisted of five games, and ultimately, AlphaGo defeated Lee Sedol 9-dan with an overwhelming score of 4-1. After the match concluded, people believed it was virtually impossible for a human to defeat a computer capable of rapidly calculating every possible move. Although Lee Sedol 9-dan won only one game, many hailed him as a monumental challenger and showered him with praise. If AlphaGo had relied solely on calculating every possible move, this victory would have been nothing more than the result of a computer with improved computational power. However, AlphaGo’s victory did not stem from improved hardware performance but from innovative advancements in its internal algorithms, demonstrating that AlphaGo’s understanding of Go was on a completely different level from that of existing AI. AlphaGo’s algorithms have since been applied to various fields, opening up the possibility of profoundly impacting our lives.

Existing Algorithms: The Minimax Algorithm and Its Limitations

To understand why AlphaGo is special, we must first understand the Minimax algorithm used by existing board game AIs. In games like Connect Four and chess, AI has been able to defeat humans for a long time; in chess, it has been decades since a computer first defeated a world champion. The Minimax algorithm used at the time is based on the concept of “considering all possible moves to select the best one.” True to its name, the Minimax algorithm aims to derive the best possible outcome for itself by preparing for the worst-case scenario that the opponent might play. In the case of a chessboard, the number of possibilities is relatively small when considering the movements of specific pieces within the limited 8×8 grid; therefore, as long as computational power is sufficient, moves can be made while looking as far ahead as possible.
However, the number of possibilities in Go is on a completely different scale. The Go board spans a vast 19×19 grid, and since there are almost no restrictions on where stones can be placed, hundreds of millions or even trillions of possible scenarios arise from the very first move. If one were to calculate all possible moves using the Minimax algorithm, analyzing just six moves ahead would require evaluating approximately 22 trillion possibilities. Assuming one calculation per second, it would take 700,000 years to complete all calculations. Even in chess, analyzing all possibilities through simple computation alone was impractical, so various shortcut techniques were employed; however, this approach was no longer effective for complex games like Go.

AlphaGo’s Innovation: Intuitive Judgment Through Deep Learning

The reason AlphaGo was able to defeat humans at Go was that it introduced a new approach that overcame the limitations of the Minimax algorithm.
Since selecting the best move in Go is extremely difficult, AlphaGo used a method that drastically reduced the computational process by employing judgment similar to human intuition. The core of this new approach stemmed from a technique called deep learning. Deep learning is an artificial intelligence methodology that mimics the neural networks of the human brain, enabling it to solve complex problems through learning and intuition.
In deep learning, learning occurs as input data is processed through multiple layers of neurons via artificial neural networks. In this process, rather than simply calculating every possible move, AlphaGo predicted the move with the highest winning probability at a specific position based on learned patterns and played accordingly. Through gradient descent—one of the key learning techniques in deep learning—the AI reduces errors through repeated learning, enabling it to make more accurate judgments. These deep learning techniques have the advantage of performing all operations on a matrix basis, allowing for parallel processing; this enables the use of GPUs to process calculations at tremendous speeds.

The Introduction of CNNs: Recognizing the Go Board as an Image

Among the deep learning techniques used by AlphaGo, CNNs (Convolutional Neural Networks) played a pivotal role. CNNs are algorithms that originally demonstrated significant success in image recognition and classification tasks, excelling at recognizing patterns and extracting features from images. CNNs operate by analyzing each pixel of an image to classify the shapes and features of the objects contained within it. AlphaGo converted the Go board into pixel data for the CNN, visualizing the placement of the stones and the overall board configuration to facilitate learning. Based on the characteristics of the CNN, AlphaGo analyzed each move on the Go board as if it were a color pattern in an image, enabling it to calculate the probability of winning for the next move. This enabled it to make highly intuitive moves through learning and pattern recognition rather than simple computation, allowing it to make decisions on a different level from existing AI.
AlphaGo was equipped with self-learning capabilities that set it apart from existing AI, developing judgment skills close to human intuition by repeatedly training itself using game records. This signifies that deep learning technology has endowed AI with self-learning capabilities, enabling it to perform tasks requiring complex thought processes—going beyond merely solving given problems. This self-learning capability serves as an example demonstrating that AI is increasingly mimicking human thought processes and is approaching a stage where it can independently generate new strategies.

The Future Unlocked by Deep Learning: Artificial Intelligence Entering Our Lives

The impact of deep learning techniques on daily life is beyond imagination. Even now, AI is providing significant assistance by mimicking human thought and judgment in various fields, such as image classification, language translation, and speech recognition. In the future, as deep learning becomes more widespread and AI gains the ability to learn an individual’s behavioral patterns, it will evolve beyond a simple assistant to play a role in predicting a person’s actions to reduce mistakes and making necessary decisions independently. For example, if we enter an era where AI can learn a person’s health status and daily habits to detect risk factors in advance, or predict traffic conditions in real time to guide users along the optimal route, our lives will become more convenient and safer.
Furthermore, as AI becomes deeply involved in people’s lives and gradually takes on a greater role, we will be able to experience new ways of living that we could not have even imagined in the past. In this sense, the 2016 match between AlphaGo and Lee Sedol went beyond a simple game of Go; it marked the beginning of the breakdown of the boundary between humans and AI, and can be seen as a preview of our future, in which we will live alongside AI.