1.1 Mailstone of AI, AlphaGo, 2016

Machine learning is not a new concept, but the latest version of the machine learning legend has indeed just happened recently. Have you heard about AlphaGo? AlphaGo is an artificial intelligence program developed by DeepMind that made history in 2016 by defeating a world champion Go player.

L: Go is an ancient board game originating from China over 2,500 years ago, in which two players compete to capture territory using black and white stones on a grid. The game’s simple rules allow for deep strategic complexity, making it one of the most intellectually challenging games in the world. R: AlphaGo is an artificial intelligence program developed by DeepMind that became the first to defeat a professional human player, and eventually world champions, in the complex board game Go, showcasing a major milestone in AI’s capabilities in strategic thinking and decision-making. Source: Google search.

Back in the year 2000, the computer program Deep Blue played to a draw against the world chess champion Garry Kasparov. Under this milestone, human couldn’t even fathom how to defeat top Go players on a Go board, and some believed it to be an impossible task forever. Indeed, the \(19 \times 19\) Go board has 361 intersections, and the number of possible combinations is astronomical, making it a task that even the most powerful computers couldn’t handle. Furthermore, unlike chess, every Go piece (stone) has equal values and there is no difference between them, which makes it very difficult to evaluate the situation on the board and make decisions accordingly.

Unlike in chess, each piece in Go starts with the same value, but its worth on the board isn’t fixed; it changes continuously with the evolving situation. This makes assessing the situation extraordinarily difficult, let alone writing a program to evaluate it. This resonates with life—though, unlike Go pieces, we have the power to determine our own path, at least to some extent.

However, in just 16 years, this last bastion of human intelligence was breached by computer programs. In 2016, a computer program, AlphaGo, defeated top Go players from South Korea for the first time in a Go competition. In one year, the new version AlphaGo, Go Master, defeated the current world No. 1 ranking player from China. After that, the strongest version of AlphaZero could give three handicaps to top professional Go players. I consider this event to be a significant milestone in the history of AI. Among the many technologies behind AlphaGo, machine learning played a significant role. Following that, various applications of machine learning blossomed, and people started applying this computer technology to a wide range of fields.

Machine learning applications: 1) AlphaGo, 2) Fraud detection, 3) Application in finance, 4) Pseudo CT image, 5) Face detection, 6) Spam filter, 7) Breast cancer automatical detection based on medical image.

For example, people use machine learning to generate CT scan images with potential side effects from harmless MRI scans; Machine learning trains programs to assist doctors in extracting essential information from medical images; More powerful spam filter is also trained by machine learning algorithm; When you take photos with your cellphone, automatic portrait recognition and focus are also achievements of machine learning; Up until now, you can communicate with ChatGPT seamlessly and obtain reliable information, all thanks to the dividends from machine learning.

Previous page | Lecture 1 Homepage | Next page