GPT-5.6 Sol disproves 30-year-old statistics conjecture in 90 minutes

A University of Pennsylvania statistics professor used OpenAI's GPT-5.6 Sol Pro to disprove a central open conjecture about the Benjamini-Hochberg method in roughly 90 minutes. The predecessor model, GPT-5.5, couldn't find a solution even after 20 hours. The answer combines known methods in a new way, keeping the bigger question alive: can AI produce genuinely new knowledge, or does it just recombine what it already learned?

The professor's success with GPT-5.6 Sol Pro raises questions about the capabilities of current AI systems. By disproving the conjecture, the model demonstrated an ability to combine existing knowledge in innovative ways. This achievement is particularly notable given the time frame in which it was accomplished - a task that had stumped humans for 30 years was resolved in under two hours.

The Benjamini-Hochberg method is a widely used statistical technique, and the conjecture in question had significant implications for its application. While GPT-5.6 Sol Pro was able to provide a solution, it remains unclear whether the model truly understood the underlying principles or simply identified a novel combination of existing methods. The achievement also highlights the rapid progress being made in AI development.

The predecessor model, GPT-5.5, was unable to solve the problem despite being given 20 hours to work on it. In contrast, GPT-5.6 Sol Pro was able to provide a solution in a fraction of the time. Why this matters: The ability of AI systems like GPT-5.6 Sol Pro to disprove long-standing conjectures has significant implications for various industries, including statistics, mathematics, and computer science.

For developers, this achievement demonstrates the potential for AI to accelerate discovery and innovation. However, it also raises questions about the role of human intuition and understanding in the discovery process. As AI systems continue to advance, it will be essential to consider the broader implications of their capabilities and limitations.

Can AI systems like GPT-5.6 Sol Pro truly produce new knowledge, or are they simply recombining existing ideas? As researchers and developers, we must consider these questions and explore the potential applications and consequences of AI-driven discovery.