On October 10, 2026, OpenAI published over 370 mathematical results across various fields, including algebra and theoretical computer science, generated in approximately three hours using a single prompt. Alex Kontorovich, a mathematics professor at Rutgers University, remarked that a human achieving this feat would likely receive a Fields Medal without question.
Recent months have seen a surge in AI-driven mathematical advancements. OpenAI previously announced that its AI solved the Navier-Stokes existence and smoothness problem, a Millennium Prize Problem, after 88 hours of computation. Concurrently, Google DeepMind introduced AlphaGenome Atlas, which predicts the molecular effects of all possible single-letter changes to human DNA.
This rapid innovation raises questions about the value of human contributions in the face of AI advancements. Economist Julian Simon, in his 1981 book, referred to people as “the ultimate resource,” emphasizing that human ideas drive innovation and economic growth. However, with frequent AI breakthroughs dominating the news, skepticism about Simon's assertion has grown.
Matt Ridley, in his 2010 book, argued that innovation arises from the collaboration of human minds rather than individual genius. He described this collective knowledge as the “collective brain,” suggesting that societies benefit from diverse expertise and collaboration.
AI breakthroughs, such as those from OpenAI and DeepMind, rely on existing human knowledge. OpenAI's agents utilized a cached version of the internet, while AlphaGenome was trained on extensive experimental data. This raises the question of whether AI can replace human collaboration in generating new ideas.
However, this perspective may be overly pessimistic. AI systems recombine existing human ideas at unprecedented speeds, acting as a medium for human knowledge rather than replacing it. OpenAI's approach to solving the Navier-Stokes problem involved 10,000 agents working in parallel, exchanging approximately 2.7 million messages to consolidate insights.
While AI can produce additional knowledge, the ultimate bottleneck for innovation remains human understanding. For instance, AlphaGenome Atlas can predict genetic variant effects, but it lacks comprehension of the underlying biology. Similarly, mathematicians are still evaluating the validity of OpenAI's proofs, with 25 Fields Medal winners emphasizing that solving problems is merely a means to achieve deeper conceptual understanding.
Economic historian Joel Mokyr, a 2025 Nobel Prize winner, distinguishes between two types of knowledge: prescriptive knowledge (knowing that something works) and propositional knowledge (understanding why it works). He argues that both types of knowledge must progress together to foster innovation.
Despite concerns about AI's impact on human roles in innovation, Mokyr's insights suggest that understanding the reasons behind discoveries is as crucial as the discoveries themselves. AIs can rapidly generate prescriptive knowledge, but human minds are essential for verifying and contextualizing this output. Thus, the ultimate resource remains human intellect, underscoring the need for more educated individuals in the face of advancing AI capabilities.