.svg){kind=logo}
Thinking New
Mathematics
Behind Intelligence
Do the mathematical structures used by today's large language models and state-space models have what it takes to lead to general intelligence?
The Kompact AI Math Lab, initiated by Ziroh Labs in collaboration with academicians from leading institutions including IIT Guwahati, is an initiative to investigate that question from first principles.
Rather than focusing only on making existing AI models bigger or faster, the Math Lab looks deeper, at the mathematics that makes these systems work.
From Scaling AI to Understanding Intelligence
Large language models and state-space models have demonstrated remarkable capabilities. Yet their underlying mathematical structures were not necessarily designed with general intelligence as the end goal.
The Math Lab seeks to examine these foundations rigorously:
- What numerical structures are currently used to represent, transform and reason over information?
- What are their fundamental mathematical properties and limitations?
- Can these structures support the kind of abstraction, reasoning and generalization expected from general intelligence?
- If not, what mathematical structures might be needed instead?
A Research Question for the Next
Generation of AI
The future of AI may not be determined solely by larger models, more data, or greater compute.
It may depend on discovering which mathematical foundations lie beneath intelligence itself.
The objective is not to build another model.
It is to understand what mathematical substrate the next generation of AI requires.