Go Has A Cheating Problem

How AI turned the ancient sport of Go upside down

By Jake Kwon, CNN
Updated 9:42 PM EDT, Fri March 24, 2023

A young visitor plays Go with an AI robot during a special exhibition of the 5th Digital China Summit in Fuzhou, southeast China’s Fujian Province, in July 2022.

Hong Kong CNN —

In December, as AI chatbot ChatGPT awed the world with its human-like responses to questions, a major cheating scandal involving artificial intelligence was erupting in Asia.

The Chunlan Cup, an international tournament boasting $200,000 in prize money for winning the ancient Chinese board game of Go, was embroiled in controversy following a semifinal match.

In a David vs Goliath moment, a relative newcomer, Li Xuanhao of China, defeated the reigning world champion Shin Jin-seo of South Korea. On social media, Li’s own teammate accused him of cheating using AI, which is commonly used during training but banned during competition.

The controversy drew coverage from major newspapers, including Chinese state media. Players called for new measures to prevent AI-assisted cheating, saying it was an existential threat to the sport, which is known in China as weiqi and Korea as baduk.

A livestream of the semi-final match between Li Xuanhao (top) and Shin Jin-seo (bottom). Viewers can see the AI-calculated outcome probability in real time.

Though the Chinese Weiqi Association declared after weeks of investigation that it found no evidence of cheating, the scandal raised questions about the future of the 2,500-year old sport and offered a glimpse into the kind of disruption that technology like ChatGPT may bring into a world previously dominated by humans.

“AI destroyed all the existing orders of this [Go] community and rebuilt it,” Jiuheng He, an avid Go player who researches AI at Cornell University, told CNN. “Human experts used to dominate the whole realm. Now we have to accept a non-human actor who has expertise, maybe even has exceeded the human experts. So how are we going to deal with it?”

How AI changed the game

The scandal swirling over the Chunlan Cup wasn’t the first time AI disrupted the game of Go.

For thousands of years, it was considered the height of intellectual pursuit in East Asia. Even today, there are 40 million players in China studying in 200,000 schools, according to the Chinese Weiqi Association.

Unlike chess, which was dominated by computer programs starting in the 1990s, Go was considered too complex to be mechanized due to the near-infinite number of possible moves on its 19 by 19 grid.

Go is an ancient board game for two players that is noted for being rich in strategy despite its relatively simple rules.

Go masters, once household names in Asia, were held in reverence. Like gods, they appeared to “stand on mountains” and all knowledge of the game flowed from them, according to He. They were so famous they would publish books to advise players about life.

But Google’s superhuman AI arrived in 2016. The 18-time world champion Lee Se-dol of South Korea was soundly defeated by AlphaGo in a widely-publicized match. Lee announced his retirement three years later, citing the match as the reason.

“Humans had been playing Go for thousands of years, improving it, but AI within a year showed that they are better. That our level of Go was really beginner,” Ao Lixian, who teaches at Hong Kong Children’s Go College, told CNN.

At the school, which opened in 2003, Ao and another instructor, Ng Chee Man, teach children to play Go using AI, which has become an essential part of virtually every player’s journey.

Ao Lixian and her students discuss strategy at Hong Kong Children’s Go College (CNN blurred portions of the image to protect a child’s identity).

On an iPad, Ng demonstrated a practice game against AI to his students. Each time it was Ng’s turn, the AI program suggested the best moves in blue spots on the board.

On the corner of the screen, the program displayed which of his moves were considered “good” in green and “bad” in red, along with how close his moves were to that of the AI in percentages.

Cheating ‘made easy’?

While training with AI has become commonplace, competition is an entirely different matter.

Shin Jin-seo, the South Korean world champion, told CNN that cheating is a major problem during tournaments. There have been at least two known AI-assisted cheating scandals in his country alone since 2016.

A South Korean court sentenced two people to a year in prison in 2020 after they were caught using AI in an official competition, according to South Korean news agency Yonhap.

The player snuck a camera and an earphone into the match and received AI-calculated moves from an accomplice outside.

Later in the same year, the Korea Baduk Association investigated one of its professional players after allegations surfaced online. The association promised to prevent future AI-assisted cheating after the player admitted misconduct.

Though phones are banned in professional matches and there are cameras watching the players, games are still vulnerable, according to Shin.

“If I were to try to cheat, I can see teammates away from the camera, and when I go to the bathroom, there is no one there,” he said.

Shin says he doesn’t know whether cheating took place during his match against Li, but he fears the sport will lose its relevance if organizations can’t guarantee clean games.

On the online Go leagues that Jiuheng He frequents, top players are those who use AI, even if, strictly speaking, they’re not supposed to. The game which he has played since childhood has become less appealing for him.

The game used to be like having a conversation with your opponent, He said. Their thoughts and intents revealed themselves with each move. “(With AI), there’s no more dialogue because I really cannot understand [its] logic,” he said.
Living with AI

Shin spends more than 70% of his training hours using AI software called KataGo, developed in 2017 by American computer programmer David Wu. AI has succeeded in setting a new, higher bar for players, even as it disrupted the game, he said.

Professor Nam Chi-hyung, who had been teaching Go for more than 20 years, says AI became essential in her lessons. Rather than being replaced by technology, she found that her work simply changed.

“AI can pick the right moves but cannot explain why. People still need me to interpret AI,” she said.

For fans, AI has made the complex game more accessible. During matches, it’s not unusual for outcomes to be unclear to many viewers. But now with the help of AI, viewers can clearly see who is winning or losing during the match.

A man watches a television screen broadcasting live footage of the Google DeepMind Challenge Match at the Korea Baduk Association in Seoul in March 2016.

But AI isn’t perfect. The Financial Times reported last month that a human player had beaten KataGo by exploiting a vulnerability discovered using another program.

KataGo isn’t omnipotent, Wu told CNN. The programs make mistakes when they are given unfamiliar problems; the same problems a human may instinctively know how to solve.

Some players believe that AI is ruining the sport, for which the unpredictability and diversity of style was the charm. After all, we can’t go back, Nam said: “It’s done. Everyone is running their AI machines.”

Man beats machine at Go in human victory over AI

Financial Times 02/19/2023

Flickr user LNG0004

A human player has comprehensively defeated a top-ranked AI system at the board game Go, in a surprise reversal of the 2016 computer victory that was seen as a milestone in the rise of artificial intelligence.

Kellin Pelrine, an American player who is one level below the top amateur ranking, beat the machine by taking advantage of a previously unknown flaw that had been identified by another computer. But the head-to-head confrontation in which he won 14 of 15 games was undertaken without direct computer support.

The triumph, which has not previously been reported, highlighted a weakness in the best Go computer programs that is shared by most of today’s widely used AI systems, including the ChatGPT chatbot created by San Francisco-based OpenAI.

The tactics that put a human back on top on the Go board were suggested by a computer program that had probed the AI systems looking for weaknesses. The suggested plan was then ruthlessly delivered by Pelrine.

“It was surprisingly easy for us to exploit this system,” said Adam Gleave, chief executive of FAR AI, the Californian research firm that designed the program. The software played more than 1 million games against KataGo, one of the top Go-playing systems, to find a “blind spot” that a human player could take advantage of, he added.

The winning strategy revealed by the software “is not completely trivial, but it’s not super-difficult” for a human to learn and could be used by an intermediate-level player to beat the machines, said Pelrine. He also used the method to win against another top Go system, Leela Zero.

The decisive victory, albeit with the help of tactics suggested by a computer, comes seven years after AI appeared to have taken an unassailable lead over humans at what is often regarded as the most complex of all board games.

AlphaGo, a system devised by Google-owned research company DeepMind, defeated the world Go champion Lee Sedol by four games to one in 2016. Sedol attributed his retirement from Go three years later to the rise of AI, saying that it was “an entity that cannot be defeated.” AlphaGo is not publicly available, but the systems Pelrine prevailed against are considered on a par.

In a game of Go, two players alternately place black and white stones on a board marked out with a 19×19 grid, seeking to encircle their opponent’s stones and enclose the largest amount of space. The huge number of combinations means it is impossible for a computer to assess all potential future moves.

The tactics used by Pelrine involved slowly stringing together a large “loop” of stones to encircle one of his opponent’s own groups, while distracting the AI with moves in other corners of the board. The Go-playing bot did not notice its vulnerability, even when the encirclement was nearly complete, Pelrine said.

“As a human it would be quite easy to spot,” he added.

The discovery of a weakness in some of the most advanced Go-playing machines points to a fundamental flaw in the deep-learning systems that underpin today’s most advanced AI, said Stuart Russell, a computer science professor at the University of California, Berkeley.

The systems can “understand” only specific situations they have been exposed to in the past and are unable to generalize in a way that humans find easy, he added.

“It shows once again we’ve been far too hasty to ascribe superhuman levels of intelligence to machines,” Russell said.

The precise cause of the Go-playing systems’ failure is a matter of conjecture, according to the researchers. One likely reason is that the tactic exploited by Pelrine is rarely used, meaning the AI systems had not been trained on enough similar games to realize they were vulnerable, said Gleave.

It is common to find flaws in AI systems when they are exposed to the kind of “adversarial attack” used against the Go-playing computers, he added. Despite that, “we’re seeing very big [AI] systems being deployed at scale with little verification.”

Copyright The Financial Times Limited 2023 © 2023 The Financial Times Ltd. All rights reserved.



Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.