Exploring the Intersection of AI and the Ancient Game of Go
Written on
The advent of algorithms in the timeless game of Go signifies more than just a technological achievement; it marks a shift in our understanding of intelligence and creativity. Before facing the AI program AlphaGo, Go Grandmaster Lee Sodol expressed his confidence, asserting that human intuition would prevail over artificial intelligence.
In the realm of ideas, the origin of concepts within society can often be elusive. Music producer Rick Rubin posits that creativity comes from an external source, suggesting that those attuned to these signals can harness them before they fade away. This phenomenon illustrates how similar ideas can emerge across different cultures simultaneously. Take, for example, the ancient game of Go, which some claim is over four thousand years old, while others estimate it to be around two thousand years.
John Fairbairn, a Go expert, notes that the game, referred to as "Weiqi" in Chinese, is mentioned in classic texts like the Analects of Confucius and Zuo Zhuan, dating back to the fifth century BC. Its precise origins remain shrouded in mystery, but it is believed to have traveled from Korea to Japan about fifteen hundred years ago, where it acquired the name "Igo."
Just as the term “internet” has become universally recognized, the concept of “AI” has permeated contemporary discourse. A casual inquiry about AI will likely spark a conversation, revealing a societal familiarity that underscores its significance. The recognition of AI reached a pivotal moment in March 2016 when computers officially achieved the status of “AI” by mastering the game of Go.
The Challenge of Go
While chess saw its first computer champion in 1997, Go presented a far greater challenge. The documentary AlphaGo illustrates why this ancient game was viewed as the “holy grail” of artificial intelligence. Frank Lantz from NYU describes Go as “putting your hand on the third rail of the universe,” highlighting the immense focus required.
The goal in Go is straightforward: control the most territory using a connected set of stones. However, the nearly limitless options available for each move create complex ripples throughout the game. Demis Hassabis, co-founder of Google’s DeepMind, explains that while chess offers about twenty potential moves at each turn, Go presents around two hundred. The configurations available in Go far exceed the total number of atoms in the universe, making it a daunting challenge for even the most advanced computing systems.
DeepMind's AlphaGo was developed to emulate human intuition in gameplay. After extensive research, a match was arranged against European Go Champion Fan Hui, who, despite his confidence in Go's complexity, lost five games to none against AlphaGo. The defeat left Hui visibly shaken, as losing to an AI in such a complex game was unfathomable.
How AlphaGo Plays and Wins
AlphaGo's architecture involved advanced neural networks that mimic human brain functions, allowing it to learn from experience. This capability enabled it to analyze thirty million moves from top players, mastering the fundamentals of Go. In addition, AlphaGo utilized reinforcement learning, allowing various versions of the program to learn from simulated matches, continuously refining its strategies.
The ultimate test for AlphaGo came when it faced Lee Sodol, a renowned player ranked “9th Dan,” compared to Hui's “2nd Dan.” Sodol had a legendary reputation, having turned professional at twelve and winning numerous championships. The anticipation surrounding their match in Seoul, South Korea, was palpable.
During the games, AlphaGo's aggressive tactics stunned spectators. In one pivotal moment, Sodol left the table in frustration after making a move, allowing AlphaGo to execute what was later termed “move 37.” Initially perceived as a mistake, this move turned out to be groundbreaking, leaving Sodol in a state of bewilderment. Ultimately, AlphaGo triumphed in four out of five games, leading to a collective realization of the potential of AI.
Reassessing Humanity's Relationship with AI
The aftermath of Sodol’s defeat prompted a reflection on the implications of AI's capabilities. Even the team at DeepMind felt a sense of loss, as if humanity's unique creative spark had been eclipsed by a machine. This moment was a turning point, reshaping our understanding of intelligence and the meaning of “artificial intelligence.”
Rick Rubin offers a different perspective, suggesting that AlphaGo's victory stemmed from its lack of preconceived notions, a state he refers to as the “beginner's mind.” This mindset is something that artists strive for to achieve genuine originality. Interestingly, AlphaGo's presence also pushed Sodol to elevate his game, leading to a remarkable move in Game Four known as “God’s Touch.”
The relationship between humans and AI need not be adversarial; history shows that tools like electricity and fire, once seen as threats, can become invaluable assets. AI can serve as a powerful ally that enhances human creativity rather than diminishes it, offering a fresh perspective that encourages growth.
In essence, just as Go has long been a vehicle for personal development, AI holds the potential to enrich our understanding of creativity in the right context.
For more insights like this, consider joining my mailing list. If you wish to connect, feel free to check my profile page.