What if we donated our spare AI subscriptions to math? (blog.kolen.dev)

A proposal has emerged suggesting that individuals could donate their unused AI subscription resources to collaboratively tackle complex mathematical problems, similar to existing projects like Prime95. This initiative aims to mobilize spare AI capacity—such as underutilized hours in subscriptions or idle computing power—allowing non-mathematicians to contribute to mathematical research through running open models designed for this purpose. Terence Tao's recently launched Open Math OS exemplifies this approach, aiming to create a platform for post-training AI models focused on mathematical challenges, which may yield significant insights despite being a year behind cutting-edge capabilities. The significance of this initiative lies in its potential to democratize access to mathematical research, providing opportunities for broader participation beyond traditional mathematicians. However, it also raises concerns regarding the verification of contributions, the danger of network poisoning by malicious inputs, and ethical implications regarding prioritization of problems. As the system encourages collaborative problem-solving among AI agents, a robust framework will be necessary to ensure quality and trustworthiness in contributions. If successfully implemented, this approach could radically shift how mathematical challenges are addressed, allowing diverse voices and ideas to drive progress in the field.