It was always thought that artificial intelligence (AI) would outsmart humans in any strategic game, but a recent news story may prove otherwise. Kellin Pelrine, an amateur Go player, has beaten an AI system using tactics that distracted the algorithm, making it lose focus on the game’s attack.
Go: The Game
Go is an ancient board game that involves two players placing either a black or white stone on a 19 x 19 board. The goal of the game is to surround a larger area than your opponent. Stones can be removed from the board if they are surrounded by opposing stones.
Pelrine’s Tactics Against AI System
Pelrine used a strategy that involved diverting the algorithm’s attention with moves in other corners of the board while he worked on surrounding groups of his opponent’s stones. Although the AI system did not notice its predicament, a human player would have easily identified Pelrine’s strategy.
Beating AI Systems
Pelrine played against several Go-playing AI systems, including KataGo, based on techniques used by DeepMind in the creation of AlphaGo Zero.
Interestingly, the approach for defeating these AI systems was discovered by a computer program created by a team of researchers (including Pelrine) to probe for weaknesses in AI strategy.
The program played over a million games against KataGo to analyze its behavior, revealing a strategy that a human player could use to successfully beat the Go-playing AI models.
A Victory for Human Intelligence
In a surprising turn of events, an amateur Go player in the US has beaten several Go-playing AI systems, including KataGo, which was based on DeepMind’s AlphaGo Zero techniques.
Kellin Pelrine used a tactic that involved distracting the algorithm with moves in other corners of the board while he worked to surround groups of his opponent’s stones.
The Future of AI and Human Strategy
Pelrine’s victory against AI systems may suggest that human players are not obsolete in strategic games. With the use of AI, human players can discover new tactics to defeat increasingly sophisticated machine-learning models.
According to Pelrine, the strategy found by the software is not too difficult for human players to understand and use, and an intermediate player can use it to beat the Go-playing AI models.
The Blind Spots of AI
The latest move highlights that AI systems may appear to be experts at whatever processes their model has been trained to perform, but there can still be surprising holes in their capabilities.
Software engineer and professional chess and mind sports player Alain Dekker said any highly trained AI is likely to have these blind spots, and that adding more and more complexity to cover the blind spots is partly why it is so hard to get it working well.
The Future of AI Strategies
The research team told the FT that the exact cause of the blind spot in the AI Go players’ strategies is a matter of conjecture, but it is likely the approach used by Pelrine is so rare that the algorithm does not recognize it.
If so, it seems likely that updated models trained to recognize this strategy may in future not be so easily fooled.
Conclusion
In conclusion, the victory of Kellin Pelrine against Go-playing AI systems is a remarkable achievement. It shows that even the most sophisticated AI models can have blind spots and be vulnerable to unusual tactics.
The research team’s findings suggest that updated models trained to recognize this strategy may not be so easily fooled in the future.
