On 9 March 2016, Lee Sedol began a match he was sure he would win. His opponent, a Go-playing software system called AlphaGo, had convincingly beaten European professional player Fan Hui the previous year1. However, Lee Sedol’s Go ranking was several rungs higher than Hui’s. Widely admired for his creative play, he was seen as having a much stronger chance. Go is a highly complex game in which creative and intuitive play is essential, and the expectation was that the best human players, like Lee Sedol, would continue to defeat machines for at least another decade.

AlphaGo was developed by researchers at DeepMind, building on their earlier game-playing artificial intelligence (AI) approach involving deep neural networks and reinforcement learning2. A key component of AlphaGo was its capability to learn from millions of games through self-play; as a result, the program could simulate the kind of ‘creativity’ needed to play Go at the highest level. DeepMind challenged Lee Sedol to a five-game match, held from 9 to 15 March 2016 in Seoul. AlphaGo won four out of the five games. It was a transformative moment for the Go community, and Lee Sedol retired in 2019, feeling that the competition had become pointless once AI could not be defeated.

It is certainly a sobering thought that AI systems can so easily outplay humans in complex board games such as chess and Go, which are often viewed as ultimate tests of human intellect. But playing games is about far more than intellectual prowess or calculating winning moves; it is also a deeply social activity. In the widely broadcast match between AlphaGo and Lee Sedol, the reporting repeatedly highlighted the human player’s state of mind — and the impossibility of sensing any emotion in his machine opponent. There is no gauging of AlphaGo’s ‘state of mind’ before a match, no way to assess what it is ‘thinking’ when it makes a move that could be either a mistake or a stroke of genius, no sense of emotion when a game is won or lost.

More than 15 years earlier, in another famous six-game match in 1997, IBM’s Deep Blue defeated the world chess champion Garry Kasparov. As was the case for Lee Sedol, the loss was hard to take for Kasparov, at least at first. But in response, Kasparov advocated new ways to play chess, moving beyond the human versus machine perspective and instead forming teams of human and AI players in Advanced Chess tournaments. Kasparov describes these ideas, and the Advanced Chess tournaments, which began in 1998, in his 2017 book Deep Thinking. He writes about making peace with the fast developments of AI, which have changed his beloved game of chess forever, and actively promotes human–AI collaboration, believing it is possible to combine human creativity and AI’s computational strengths in productive ways3.

Despite AI systems having cracked chess and Go, both games are still highly popular, even more than before. As mentioned above, new formats for tournaments have emerged and a new generation is playing these games on user-friendly mobile platforms, with many options to compete against others across the world in different game formats and to engage with online community building. With the rise of AI, many tools have become available for practice and game analysis.

The famous match between Lee Sedol and AlphaGo has not only changed the world of Go but also became a pivotal moment in AI development. The event almost certainly had a catalysing role in China’s ambitious decision to invest billions of dollars in AI research, aiming to become the global leader in AI by 20304. Elsewhere, Elon Musk and others expressed concern about DeepMind’s achievements, especially following Google’s acquisition of the company. OpenAI was founded in 2015 with the aim of preventing a single company, particularly Google, from gaining too much control over AI developments5. In 2022, OpenAI released ChatGPT, starting a race in the development of large language models (LLMs) and agentic AI.

In the past decade, AI systems have repeatedly surpassed human benchmarks in cognitive skills, such as passing graduate-level medical exams6 or winning a real-time strategy game like StarCraft7, prompting debate about what these milestone achievements mean for human society. However, how close we are to any particular end‑point does not seem a meaningful question to ask; instead, it is of interest to examine how each advance deepens our understanding of intelligence. The long game in AI is this ongoing, open‑ended scientific exploration.