I think that many people are aware of how AlphaGo Zero managed to defeat Lee Sedol, the then-reigning world champion of Go, with 4 against 1 back in 2016 (Chaiyong, 2020). It was one of the moments where the possibilities and potential of AI became more visible to the public than ever, and Go players around the world were impressed by the unconventional way in which AlphaGo played (Chan, 2017). The best player had been defeated, meaning that AI finally surpassed human capabilities.
The idea behind AlphaGo is as follows: the AI used a rule-based learning process. Using these rules, AlphaGo played against itself millions of times, serving as its own teacher (Silver, Schrittwieser, & Simonyan, 2017). Every time a new game started, the learnings from the last iteration was used to perform better in the next game. Ever since, the company behind AlphaGo, the UK-based DeepMind has been working on AlphaZero. AlphaZero learned how to win games against grandmasters in chess, shogi, and even Starcraft, going beyond what any human player is capable of doing (Greene, 2020).
However, after having defeated the best human players, the team behind AlphaZero found a new use for it: it is now not only able to defeat other players, but also able to play more creatively (Simonite, 2020). For example, playing chess against AlphaZero does not have to be frustrating for players anymore. It can also help hone their skills, and bring creativity back into the game. It can do so in various ways, such as mimicking the style of the player in order to challenge them, or allowing players to choose styles of specific grandmasters would. Then, AlphaZero will mimic the style of that grandmaster and challenge the player as that grandmaster. This is a large step away from the inhuman style that AI were often credited to have (Strogatz, 2018). Instead, it shows that AI can also complement even the best human players and bring creativity and new strategies back into the game (Simonite, 2020).
What do you think about this? Do you mainly see the future of AI as a way to complement human skill, or as a way to go beyond human capabilities?
Sources:
Chaiyong, S. 2020. Facing the Future. Bangkok Post. https://www.bangkokpost.com/tech/1982947/facing-the-future.
Chan, D. 2017. The AI That Has Nothing to Learn From Humans. The Atlantic. https://www.theatlantic.com/technology/archive/2017/10/alphago-zero-the-ai-that-taught-itself-go/543450/.
Greene, T. 2020. AlphaZero Beat Humans at Chess and Starcraft, Now it’s Working With Quantum Computers. The Next Web. https://thenextweb.com/artificial-intelligence/2020/01/16/alphazero-beat-humans-at-chess-and-starcraft-now-its-working-with-quantum-computers/.
Silver, D., Schrittwieser, J., & Simonyan, K. 2017. Mastering the Game of Go without Human Knowledge. Nature, 550: 354–359.
Simonite, T. 2020. AI Ruined Chess. Now, it’s Making the Game Beautiful Again. Wired.
Strogatz, S. 2018. One Giant Step For a Chess-Playing Machine. New York Times. https://www.nytimes.com/2018/12/26/science/chess-artificial-intelligence.html .
