Artificial intelligence (AI) is significantly impacting various fields, including mathematics . Notable advancements have been made, such as the ability of AI to generalize critical mathematical concepts. For instance, in October 2024, Meta AI achieved a remarkable milestone by generalizing the Lyapunov function , a concept introduced by the Russian mathematician Aleksander Lyapunov in 1892. This function is pivotal in the study of dynamic systems , yet mathematicians have long encountered challenges in identifying these functions broadly. However, with the emergence of Goal AI , significant progress is being made.
This breakthrough is just one among several recent successes in employing AI within advanced mathematics. Sergei Gukov, a professor at the California Institute of Technology (Caltech) , is leading efforts to utilize AI for tackling complex mathematical problems that require extensive computational resources—sometimes necessitating thousands, millions, or even billions of operational steps. Currently, Gukov’s team is engaged in exploring the Andrews-Curtis conjecture , a combinatorial theory problem that has puzzled mathematicians for over six decades.
Google and OpenAI AI Have Won Gold in the Mathematics Olympiad
Although Gukov and his team are yet to resolve the main conjecture, their collaboration with AI has yielded critical outcomes. They have successfully refuted several families of counterexamples related to the Andrews-Curtis conjecture, many of which have remained unresolved for more than 25 years. Gukov acknowledges existing limitations in current AI models when addressing intricate mathematical dilemmas. Nevertheless, he remains hopeful that as this technology evolves, it may eventually empower researchers to tackle and solve the millennium’s most pressing mathematical problems.
Reinforcement learning offers a promising approach for optimizing AI’s problem-solving skills in mathematics.
According to Gukov, one of the most promising tools available to researchers is the strategy of reinforcement learning to instruct AI in overcoming mathematical challenges. Importantly, a recent achievement has captured attention, as models developed by Google and OpenAI AI have collectively secured gold medals at the International Mathematics Olympiad . These AI systems succeeded in solving five out of the six problems presented, employing general-purpose reasoning models capable of understanding mathematical concepts framed in natural language. This innovative approach contrasts sharply with prior methodologies used by AI companies in mathematical assessments.
Experts predict that the rapid pace at which AI models are evolving suggests they might not be far from solving longstanding mathematical challenges. A specialist consulted by SCMP noted that these advancements indicate we could see AI tackle unresolved problems that mathematicians have wrestled with for decades. While Gukov supports this view, he refrains from committing to a specific timeline for when AI might achieve solutions for these enduring mathematical enigmas. Nevertheless, the possibility that we might be on the verge of breakthroughs related to the millennium problems is an exciting prospect for mathematicians and enthusiasts alike.
Image | Jesus Thomas
For further insights on this topic, more information can be found at SCMP .
In summary, the integration of AI into the realm of mathematics is transforming problem-solving capabilities and shed light on long-standing challenges. Developments like those made by Meta AI and the accomplishments of teams led by Gukov are just the beginning. The blending of computational technology with human intellect may soon lead to unprecedented advancements in mathematical understanding. Indeed, the coming years could unveil solutions to the most intricate problems that have eluded mathematicians, promising an exciting future for both the field of mathematics and artificial intelligence.