AI solves decade-long 'Go First' dice puzzle in minutes

A free AI chatbot has reportedly solved a complex mathematical puzzle known as the 'Go First' dice problem for six players, achieving in minutes what had taken humans over a decade to approach. The discovery highlights the shifting boundary between human intuition and machine computation in recreational mathematics.

According to reporting by New Scientist, reporter Matthew Sparkes used a free AI chatbot to tackle the problem, with the AI producing a solution in 13 minutes. The 'Go First' challenge requires the creation of a specific set of dice used to determine the starting player in a board game. To be successful, the dice must ensure that every player has an exactly equal probability of winning and that ties are mathematically impossible.

The human struggle for symmetry

The difficulty of the problem lies in the strict constraints of probability and symmetry. For years, mathematicians have struggled to find sets that satisfy these conditions as the number of players increases. Eric Harshbarger of Auburn University, a key figure in the research of Go First dice, spent more than ten years working to find a viable solution for just five players. The jump to six players represents a significant increase in complexity, as the number of possible dice combinations grows exponentially.

Implications for mathematical discovery

This event demonstrates the increasing capability of Large Language Models (LLMs) to handle niche mathematical constraints that previously required years of dedicated human effort. While LLMs have historically struggled with precise arithmetic, this instance suggests they are becoming more adept at navigating the logical frameworks required for combinatorial problems. By rapidly iterating through constraints, AI may now act as an accelerator for discovery in both formal and recreational mathematics, solving 'bottleneck' problems that would otherwise remain stagnant.

What remains to be seen

While the New Scientist report claims the AI solved the six-player variation in 13 minutes, this specific result has not yet been independently verified by other mathematical repositories or external news outlets. The primary focus for the mathematical community will now be the formal verification of the AI's proposed dice values to ensure they truly eliminate ties and maintain perfect equity. If verified, the result marks a clear victory for AI-assisted problem solving in a field defined by human persistence.