Can artificial intelligence surpass not only human reasoning, but also human creativity? The emergence of AlphaGo has made us rethink the possibilities and limitations of AI beyond simple technological innovation. Let’s take a look at the changes brought about by deep learning and the future of AI.
In 2016, the Go program AlphaGo, developed by Google DeepMind, became a global sensation. AlphaGo set a record of 5 wins, 4 losses, and 1 draw in a game against legendary Go player Lee Sedol, 9th Dan. It is said that AlphaGo was the first artificial intelligence to beat a professional Go player in the history of Go. Go was considered a game where AI would have a hard time beating a professional compared to games like chess. The reason for this is that chess allows only a fixed number of pieces to be placed on a limited 8×8 chessboard, while Go requires the consideration of countless possibilities on a large 19×19 Go board. Calculating all the possible positions on these 361 squares requires an enormous amount of computation, which is impossible to achieve with a simple algorithm.
However, AlphaGo was able to overcome these difficulties and beat the world’s best Go player thanks to a new AI (artificial intelligence) algorithm called deep learning. Deep learning is the result of about 50 years of AI development and uses innovative methods that are unlike any existing algorithms. The history of AI dates back to the mid-20th century and began with the concept of artificial intelligence proposed by Alan Turing. Turing’s theory, famously known as the Turing Test, became the cornerstone of early AI research, including programs for mathematical proof and programs that played chess. However, computers at the time were not powerful enough, so algorithms that simply computed all possible cases were limited.
To overcome these limitations, researchers looked for different methods, and one of them was the algorithm of machine learning. Unlike traditional algorithms that simply deal with fixed relationships between inputs and outputs, machine learning adopts a method in which a program learns various data to find rules and patterns on its own. In particular, various learning methods such as supervised learning, unsupervised learning, and reinforcement learning have been developed to dramatically improve the performance of AI.
In the case of AlphaGo, it learned Go by analyzing about 30 million professional Go games. In the process, it learned the basic rules of Go from the records through supervised learning, and learned the standard moves and response methods for each situation on its own through unsupervised learning. Finally, it evaluated the win rate of each move through reinforcement learning and acquired the ability to select the move with the highest probability of winning. Thanks to this learning process, AlphaGo was able to reach a level where it could beat Lee Sedol, a nine-dan (9th level) professional Go player.
Deep learning is an evolution of the artificial neural network algorithm, which is based on machine learning. The artificial neural network algorithm mimics the way neurons are connected in living organisms, and it derives results through a series of processes. Although early artificial neural networks had limitations, deep learning has improved flexibility by stacking these processes in multiple layers. In addition, as modern computers have become capable of handling the vast amount of computation required by deep learning, this algorithm has become a key technology in AI research.
The revival of deep learning has brought AI’s heyday back. AI, once in the limelight but soon entering a dark age, is now showing new possibilities. The success of AlphaGo is an example of AI’s ability to solve complex problems, which has opened new horizons for AI research. There are predictions that the day may not be far off when AI will have thinking abilities similar to those of humans. The emergence of AlphaGo was an important milestone that showed that artificial intelligence had advanced to the point where it could actually compete with humans, beyond simply changing the world of Go.
As such, AlphaGo has remained a symbolic figure that has realized the possibility of AI that humanity has dreamed of, beyond a simple technological achievement. Considering that the development of artificial intelligence has only just begun, it is difficult to imagine the impact that AI will have on our society in the future. Expectations are still growing about how far the convergence of human creativity and technology will go and what changes it will bring to human life. The future of AI is being mapped right now, and we are at the center of that change.
