We offer mathematical crash courses and valuable resources designed to help AI and ML researchers and engineers gain a mathematical edge.
This week, OpenAI announced a remarkable breakthrough: their new O3 model achieved a 25% success rate on the FrontierMath benchmark. This is an impressive feat given that the benchmark was introduced only a while ago and poses challenges typically reserved for PhD-level mathematicians. While 25% may seem modest at first glance, it represents a significant stride forward. In the spirit of Neil Armstrong’s iconic words, this could be considered “One small step for AI and a giant leap in humanity’s quest toward AGI.”
This progress has energised our team around a compelling question we’ve been pondering: should we shift our focus from AGI (Artificial General Intelligence) to a more expansive concept: Super AGI? By “Super AGI,” we mean a collective intelligence comprising multiple specialised AGI agents, each an expert in a specific domain such as mathematics, physics, biology, chemistry, social sciences, and more. All working in concert toward a unified goal.
Moreover, we might look to religious analogies for inspiration. In Christianity, the concept of the Trinity describes one being—God—existing as three distinct persons: the Father, the Son, and the Holy Spirit. Analogously, one could imagine multiple specialised AGIs where each possessing unique capabilities. Then merging into a single, more profound intelligence. This multiplicity might address the complexity of human knowledge in a way that a single, monolithic AI cannot achieve on its own.