Greetings from a world where…
does anyone else go back and read the Grantland archives from time to time
…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).
Feature Translation: Go player uses AI to cheat
Context: On March 15, 2022, the Chinese Weiqi (Go) Association announced that Liu Ruizhi had used AI to cheat in the preliminaries of the Changqi Cup competition. Liu’s victory was revoked and he was disqualified from participating in professional competitions for the next year. According to the xinzhiyuan report (link to original Chinese), this marked the first time that the Chinese Go community had officially punished a professional player for using AI to cheat.
Key Takeaways: How do you catch Go cheaters?
Sometimes it’s obvious. The day before during the same competition, another Go player named Zhou Weiping also incurred a rule violation for not being in the same location as his monitor (someone who makes sure players don’t use electronic devices and AI systems). Someone compared Zhou Weiping’s moves to a well-known AI engine, and the “match rate” was almost 94% by one calculation, which is absurdly high. Netizens joked about Zhou’s moves: “Walking a dog without a leash is like a dog walking another dog.” Walking the dog is Internet slang for using AI to cheat in Go: most Go AIs originate from the famous AlphaGo, and “dog” and “Go” in Chinese are homophonic.
Other times it’s less clear. The day after, Liu Ruizhi disguised his methods better. Across the whole match, his moves only matched the first-choice move of the AI system 42 percent of the time, which was even worse than his opponent. Still, his mistakes were very limited, and he only made one mistake that lowered his win probability by more than 10% (and this was only made when his victory was all but guaranteed). It’s possible he only used the engine in key moments.
The cheating affects more than just low-ranked players
The piece calls for further investigation of the local Go associations that hosted the online qualifiers for the competition. The extent to which these local organizations concealed or directly assisted these cheating incidents deserves further investigation, the Xinzhiyuan report argues.
Even the best of the best players can’t escape scrutiny. Ke Jie, the top Chinese Go player who lost to Alpha Go in May 2017, recently insinuated that the world’s #1-ranked Go player, a Korean player named Shin Jin Seo, used AI to cheat in their match.
Small note about how details can get lost in translation: When I was putting together this issue, I struggled with how to translate a key fact: the name of the Go competition (the Changqi Cup). Interestingly, there was some English-language coverage of the cheating incident by the American Go Association, which relied on an English-language article by The Global Times. Both sources stated the competition was “The Chinese Professional Go Championship,” which isn’t really a thing. A Google search turns up zero hits for that term before the last few weeks. Ultimately, getting the name of the competition wrong is a small thing, but it’s a good reminder for all of us to double-check our translations.
FULL TRANSLATION: Go player uses AI to cheat in a match! Association penalty: Cancellation of results, one-year suspension
ChinAI Links (Four to Forward)
Must-read: Sharing Powerful AI Models
By Toby Shevlane, published on the GovAI research blog, this piece describes how “structured access” to AI models (e.g., OpenAI’s API for GPT-3) could open up new ways of governing AI. It summarizes his chapter in the forthcoming Oxford Handbook on AI Governance, which expands more on how cloud-based interfaces allow AI developers greater control over end-use applications.
Should-read: Will China Set Global Tech Standards (A ChinaFile Conversation)
I contributed some thoughts on this ChinaFile conversation about the implications of China’s growing influence in global tech standards. Conversation brought a lot of good viewpoints; I benefited a lot from Helen Toner’s piece, especially. She writes:
A recent study run by the National Institute of Standards and Technology gave U.S. companies and industry organizations the opportunity to weigh in on how concerned the United States should be about Chinese participation in standards development. As the Carnegie Endowment summarized, only a small minority of private sector respondents expressed concerns. Far more prevalent than alarm about undue Chinese influence was consensus that the U.S. government could do more to support U.S. engagement. Relevant measures could include subsidizing the participation of companies or industry organizations in standards development processes—an expensive undertaking for which the Chinese government provides significant support.
Should-read: Chess’s cheating crisis: ‘paranoia has become the culture’
For those wanting to explore this week’s topic in more detail, Archie Bland’s 2020 article for The Guardian on the state of “computer doping” in chess is a good start.
Should-read: A Reading List for the Extremely Offline
For Longreads, librarian Lisa Bubert put together a reading list for those wanting to go from Extremely Online to Extremely Offline.
Thank you for reading and engaging.
These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is a postdoctoral fellow at Stanford's Center for International Security and Cooperation, sponsored by Stanford's Institute for Human-Centered Artificial Intelligence.
Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).
Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99
Ke Jie did not insinuate Shin cheated.