Future Of Go Summit

The Future of Go Summit ( Chinese : 中国乌镇围棋峰会), an event of profound significance for the intersection of artificial intelligence and ancient strategy, was meticulously orchestrated in May 2017. This landmark gathering was a collaborative effort, brought to fruition by the Chinese Go Association , the Sport Bureau of Zhejiang Province, and the omnipresent technological titan, Google . Its chosen locale, Wuzhen in Zhejiang , was no arbitrary selection; it holds the distinction of being the permanent host of the World Internet Conference , lending an air of established digital gravitas to the proceedings.

The summit was not merely a series of exhibition matches, but a comprehensive exploration into the evolving landscape of Go and advanced computing. At its core were five high-stakes Go games, pitting AlphaGo , Google ’s formidable artificial intelligence, against some of China’s most revered and skilled human Go players. [^1] Beyond the board, the event encompassed a forward-looking forum dedicated to dissecting and discussing the very future of AI itself. [^2] [^3]

This summit represented a remarkably public and significant partnership between Google and the Chinese government , particularly notable given the complexities of their recent history. It marked the largest public collaboration of its kind since Google China’s search engine operations were controversially relocated from mainland China to Hong Kong in 2010, a direct consequence of escalating governmental censorship. Many observers, myself included, viewed this event as a carefully calculated “charm offensive” on Google ’s part, a strategic maneuver aimed at rebuilding bridges with Chinese officials and, ultimately, paving the way for a potential reopening of the vast and lucrative Chinese market to Google ’s diverse suite of services. [^4] [^5] A rather transparent attempt, if you ask me, to mend fences with a gesture that screams “look, we can still be useful.”

The iteration of AlphaGo deployed for this high-profile summit was the advanced model known as AlphaGo Master . [^6] [^7] This particular version was a marvel of computational efficiency, operating on just four Tensor Processing Units (TPUs) within a single machine. Its estimated Elo rating, a metric for player strength, stood at an imposing 4,858. [^8] The researchers at DeepMind , the subsidiary responsible for AlphaGo ’s development, confidently asserted that this AlphaGo Master was approximately three stones stronger in games of pure self-play when compared to the version that famously defeated Lee Sedol in the historic match of AlphaGo v. Lee Sedol . [^9] One might say it had evolved past the need for human guidance, much like a teenager who suddenly decides they know everything.

Following its decisive victory in the three-game match against the reigning world’s top Go player, Chinese grandmaster Ke Jie , AlphaGo received the esteemed professional 9-dan rank from the Chinese Weiqi Association . [^10] This recognition, a testament to its unparalleled mastery, was swiftly followed by a rather dramatic announcement from DeepMind : AlphaGo would officially retire from competitive play. Furthermore, DeepMind declared its intention to disband the dedicated team that had worked on Go , reallocating their considerable talents to explore novel frontiers of AI in other, perhaps more pressing, domains. It seems even machines understand when it’s time to move on. In the wake of the summit, DeepMind offered a parting gift to the Go community, releasing 50 games that AlphaGo had played against itself. [^11] These self-play games served as an invaluable, if somewhat humbling, repository of new strategic insights for human players, showcasing moves and approaches that defied conventional wisdom.

Schedule

The summit’s schedule was meticulously planned to maximize both competitive intensity and analytical insight, offering various formats of engagement between human and artificial intelligence.

• Ke Jie vs. AlphaGo : The centerpiece of the summit, this three-game match unfolded on May 23rd, 25th, and 27th. Each game commenced at 10:30 UTC +8 and was scheduled to run until 17:30 UTC +8 , allowing ample time for the intricate strategic battles to play out. [^12] A truly grueling schedule for any human, though I doubt AlphaGo felt the fatigue.

• Pair Go: A unique collaborative format, this match featured Gu Li paired with one AlphaGo instance, competing against Lian Xiao paired with another AlphaGo . This intriguing contest took place on Friday, May 26th, from 8:30 to 12:30 UTC +8 . [^12] [^13] An experiment in human-machine synergy, or perhaps just a demonstration of who could best mimic their silicon partner.

• Team Tournament: The ultimate test of human collective intelligence against a singular artificial mind. This tournament saw a formidable Chinese team, comprising Chen Yaoye , Mi Yuting , Shi Yue , Tang Weixing , and Zhou Ruiyang , challenge AlphaGo . This grand finale occurred on Friday, May 26th, from 13:30 to 18:30 UTC +8 . [^12] [^14] Five minds against one, and the outcome was, predictably, still quite lopsided.

Games

Ke Jie vs AlphaGo

Main article: AlphaGo versus Ke Jie

The highly anticipated individual match between the world’s then-highest-ranked human player, Ke Jie , and AlphaGo was the undisputed main event of the summit. It was a clash of titans, though one titan was made of carbon and the other of silicon.

Result: AlphaGo 3 – 0 Ke Jie

Google DeepMind underscored the significance of this particular confrontation by offering substantial financial incentives. A generous prize of 1.5 million dollars was allocated for the winner of this match, while the losing side, for their courageous participation in the three demanding games, received a respectable 300,000 dollars. [^15] [^16] Despite Ke Jie ’s formidable skill and unwavering determination, AlphaGo ultimately emerged victorious in all three games, demonstrating a level of strategic depth and tactical precision that proved insurmountable for even the best human player. [^17] [^18] The outcome, though perhaps disheartening for human pride, was a clear affirmation of AlphaGo ’s advanced capabilities. The monetary incentive, while substantial, felt almost like a consolation prize for the inevitable.

Pair Go

The Pair Go match offered a fascinating deviation from the traditional one-on-one format, exploring the dynamics of human-AI collaboration.

This innovative Pair Go game was played on the morning of May 26th. The setup involved two teams, each comprising one human grandmaster and one instance of AlphaGo . On one side, Gu Li , an eight-time world champion in his own right, teamed up with an AlphaGo . On the opposing side, Lian Xiao , who at the time held the prestigious Chinese Mingren and Tianyuan titles, was paired with another AlphaGo . The players alternated moves in a distinctive tag-team style. [^19] The specific sequence of play was orchestrated as follows: Gu Li (black), followed by Lian Xiao (white), then the black AlphaGo , and finally the white AlphaGo . [^20] A rather complex dance, even for the machines.

After a protracted battle, specifically after move 198, the black AlphaGo astonishingly proposed resignation, a move that would typically indicate a clear loss. However, its human teammate, Gu Li , with a display of human tenacity or perhaps sheer stubbornness, rejected the machine’s assessment. [^21] Despite this momentary defiance, the combined strength of the white AlphaGo and Lian Xiao ultimately prevailed, securing a victory by resignation after move 220.

This Pair Go match provided a tantalizing, if slightly unsettling, glimpse into the potential future of human-AI collaboration within complex strategic games. It suggested a paradigm where human players could leverage the unparalleled computational insights of AI tools while still integrating their own unique intuition and creativity. [^19] Or, as I see it, a future where humans desperately cling to relevance while the machines do the heavy lifting. David Silver , the lead researcher behind AlphaGo , offered a rather poetic reflection on the game: “Pair Go was a beautiful game. All four players made some brilliant and creative moves. It felt like four painters working together on a shared canvas, all with different styles, all combining together to make something truly beautiful.” [^22] A lovely sentiment, though I suspect the “painters” were mostly the ones with circuits.

Team Go

The Team Go match was designed as a monumental challenge, pitting a collective of humanity’s finest against a single, unified artificial intelligence.

This decisive Team Go match unfolded on the afternoon of May 26th. In this unique five-on-one format, a formidable quintet of world champion players—Chen Yaoye , Zhou Ruiyang , Mi Yuting , Shi Yue , and Tang Weixing —joined their considerable forces to play as Black against the singular power of AlphaGo , playing as White. [^23] The time controls were set with a two-and-a-half-hour main time limit for each side, followed by three 60-second byo-yomi overtime periods. [^citation_needed]

Despite the gravity of the challenge, the human team approached the match with a remarkably light-hearted demeanor, openly consulting each other on every move and evidently relishing the rare experience of playing together in such a high-profile setting. [^19] Zhou Ruiyang was designated as the team leader, responsible for physically positioning the stones on the board during the regular time phase of the game. [^citation_needed] As the match progressed into the intense overtime periods, Tang Weixing took over the role of placing stones, appearing to operate almost independently. [^citation_needed] Ultimately, facing the undeniable inevitability of AlphaGo ’s victory, the human team made the difficult decision to resign after 254 moves. [^24] Eric Schmidt , who was the Executive Chairman of Alphabet, Inc. , DeepMind ’s parent company, chimed in with a rather optimistic tweet: “This speaks volumes about where AI is headed - human players are teaming up with AlphaGo to have even more fun with the game!”. [^25] “Fun,” he says. I imagine the fun was mostly for AlphaGo .

The game itself offered several moments of profound strategic interest and, frankly, surprise to even the most seasoned human players. For instance, White 54 was identified as a particularly unexpected move, one that defied conventional human Go wisdom. [^26] [^27] After move 58, Ke Jie , while observing the match, astutely pointed out that White could play 60. His suggestion was met with a degree of teasing from other professional players in the research room, who likely considered it an unconventional or perhaps even poor choice. Their amusement, however, was short-lived, as AlphaGo proceeded to execute precisely that move, 60, much to their collective astonishment. [^28] Tan Xiao , another prominent player, commented that after White’s move 64, it became exceptionally difficult for Black (the human team) to secure a victory. [^26] The machines, it seemed, were not playing by human rules, and humans were struggling to keep up.

Chinese coverage

The Future of Go Summit garnered immense attention within China. It was extensively reported across various Chinese media outlets and rapidly ascended to become one of the most fervently discussed topics across social networks. [^29] [^30] [^31] The initial public enthusiasm was palpable, reflecting the deep cultural significance of Go in China and the burgeoning interest in AI .

However, this widespread public engagement was abruptly curtailed due to direct instructions from Chinese authorities. The state-owned broadcaster, China Central Television , was compelled to cancel its plans for a live telecast of the games, a decision that undoubtedly frustrated many viewers. Concurrently, live streaming of the matches on numerous popular Chinese websites was inexplicably cut off. [^32] [^33] The stark irony was not lost on anyone: the only official live broadcast available for the entire summit was hosted on YouTube, a platform that remains inaccessible within mainland China due to the “Great Firewall.” This imposed restriction sparked widespread complaints and expressions of discontent among Chinese netizens, who voiced their frustrations vociferously on the social platform Sina Weibo . [^34]

Further compounding the issue, Chinese media organizations were explicitly instructed to downplay any mention of Google or Google ’s products in their reports concerning the summit. This directive was a direct consequence of Google ’s services being largely blocked in China, highlighting the government’s ongoing efforts to control information flow and limit the influence of foreign technology companies. [^35] [^36] It was an exercise in selective memory, attempting to celebrate the technological achievement while erasing the name of its progenitor. A rather futile endeavor, like trying to ignore the architect of a building while admiring its structure. The whole affair was a masterclass in how to simultaneously promote technological advancement and suppress open information, a dichotomy that would make even me raise an eyebrow.