Artificial intelligence has long been hailed as a technology that could potentially revolutionize the way we live our lives, and a recent breakthrough in the field of mathematics might just bring us one step closer to that reality. Imagine being able to predict major life-altering events, such as stock market crashes, extreme weather events, or debilitating diseases, years in advance and taking steps to protect ourselves from them. While such accurate predictions may seem like a work of science fiction, experts believe that with the help of AI, it might just be possible.
A recent preprint paper describes a mathematical breakthrough that could pave the way for AI systems capable of making billions of connections in vast datasets to reveal patterns and outcomes that would otherwise be impossible to predict. To achieve this feat, a team of researchers at the California Institute of Technology turned to the Andrews-Curtis conjecture, a complex mathematical problem that has puzzled mathematicians for over 60 years.
The Andrews-Curtis conjecture, proposed by James Andrews and Morton Curtis in 1965, posits that any complicated mathematical configuration can be reduced to its most basic form through a finite sequence of three moves. To visualize this concept, imagine a vast maze where a player is trying to connect all points to a central “home” point. The player may need to take millions or even billions of steps in the maze to achieve this goal, highlighting the immense complexity of the problem.
By using the Andrews-Curtis conjecture as a model, the research team created a game that challenges players to navigate a chess-like board with millions or even billions of squares to reach a designated “home” square using a limited set of moves. Through reinforcement learning, an AI technique that allows agents to learn through trial and error, the team trained AI systems to tackle this game.
The AI system consists of two agents: a player and an observer. The player executes moves to reach the home square, while the observer watches and evaluates the player’s actions to develop strategic “supermoves” that can help the player make bigger leaps across the board. By combining basic moves into supermoves, the observer guides the player through the complex maze, enabling it to tackle increasingly difficult coordinates.
While the game may require thousands of moves to solve, the AI system has already made significant progress in solving long-standing counterexamples to the Andrews-Curtis conjecture. By breaking down complex configurations into simpler forms, the team has managed to debunk several potential counterexamples that have remained unresolved for decades.
The implications of this research are profound, suggesting that AI systems could one day help us navigate the complexities of our world with unprecedented accuracy. While predicting the future will always be a challenging task, advancements in AI bring us closer to a reality where we can anticipate and mitigate potential crises before they occur. A recent preprint study conducted at the University of Liverpool has confirmed the results of Gukov’s team, showcasing the power of artificial intelligence (AI) in experimental mathematics. Alexei Miasnikov, a mathematics professor at the Stevens Institute of Technology, praised the work done by Gukov’s team, stating that it exceeded his expectations for what AI could achieve with the Andrews-Curtis conjecture. Miasnikov, who has conducted his own research on the conjecture, emphasized the importance of machine reinforcement in generating novel and insightful results that would be impossible to obtain without the aid of a computer.
The team led by Gukov aims to develop AI tools that can be applied to a wide range of mathematical and real-world problems. While existing AI systems like AlphaGo and AlphaStar focus on solving known problems, Gukov’s team is pushing the boundaries by tackling problems where solutions are not yet known. Their ultimate goal is to create systems that can address complex and uncertain scenarios, such as predicting machine failures, identifying errors in automated systems, and understanding long-term health outcomes.
The potential applications of these AI tools extend beyond mathematics and into fields like medicine, finance, cryptography, and climate modeling. By training their AI models with mathematical problems, Gukov and his team are laying the groundwork for future advancements in predictive analytics and problem-solving. Gukov emphasized that their focus on mathematics serves as a cost-effective way to refine their AI systems before applying them to practical applications.
While the AI system developed by Gukov’s team is not yet capable of proving or disproving the Andrews-Curtis conjecture, their work has provided valuable insights and support for the conjecture. Initially thought to be false by many in the mathematics community, Gukov now believes there is a strong possibility that the conjecture may actually be true. Through their innovative use of AI in experimental mathematics, Gukov and his team are paving the way for exciting developments in the field, with the potential to revolutionize problem-solving and predictive modeling in diverse disciplines.
