top of page
Industrial Abstract Object

Category Is All You Need
Unlocking Artificial General Intelligence through Category Theory



In the relentless pursuit of Artificial General Intelligence (AGI), New York General Group has made a groundbreaking discovery that could revolutionize the field of artificial intelligence. By incorporating category theory, a branch of mathematics that studies abstract structures and their relationships, into large-scale language models such as the Transformer architecture, our team has developed a novel approach that significantly enhances the model's ability to generalize and reason across diverse domains.


The Transformer architecture, which has already demonstrated remarkable performance in various natural language processing tasks, serves as the foundation for our research. However, despite its impressive capabilities, the Transformer model still falls short of achieving true AGI—the ability to understand and reason across a wide range of domains, akin to human cognition.


We recognized the untapped potential of category theory in addressing this challenge. By representing the input data, intermediate representations, and output structures as objects and morphisms in a category, we have enabled the model to capture complex relationships and abstractions. This categorical representation of language allows the model to grasp the intricate connections between linguistic entities, such as syntactic dependencies, semantic roles, and discourse relations.


Furthermore, we have introduced the concept of functorial learning, a groundbreaking technique that empowers the model to map between different categories seamlessly. This ability to transfer knowledge across tasks and modalities is a crucial step towards achieving AGI. By incorporating additional layers in the Transformer architecture that learn to map between categories, such as the LanguageCategory and the ImageCategory, our model can reason about the relationship between language and vision, among other domains.


To validate the effectiveness of our approach, we conducted a series of rigorous experiments on various language understanding tasks, including natural language inference, question answering, and text generation. The results were astounding. Our model, enhanced with category theory, consistently outperformed baseline Transformer models and other state-of-the-art language models. The improved performance can be attributed to the model's heightened level of generalization and reasoning capabilities, as evidenced by its ability to handle complex linguistic structures and transfer knowledge across different domains.


The implications of this breakthrough extend far beyond the realm of language modeling. By unlocking the power of category theory in artificial intelligence, we have opened up new avenues for research and development in AGI. The potential applications are vast, ranging from advanced virtual assistants that can seamlessly navigate multiple domains to intelligent systems that can reason and make decisions in complex, real-world scenarios.


At New York General Group, we are committed to pushing the boundaries of what is possible in artificial intelligence. Our team of experts, with their deep understanding of mathematics, computer science, and business, is uniquely positioned to drive this transformative approach forward. We believe that the incorporation of category theory into large-scale language models is a significant leap towards achieving AGI, and we are excited to continue exploring its potential.


As we embark on this journey, we invite collaborators from academia and industry to join us in shaping the future of artificial intelligence. Together, we can harness the power of category theory to create intelligent systems that not only understand and process language but also reason and adapt to the ever-changing world around us.


The quest for Artificial General Intelligence is not just a scientific endeavor; it is a transformative force that has the potential to reshape industries, economies, and societies. At New York General Group, we are at the forefront of this revolution, and we are committed to leveraging our expertise to drive meaningful change and create value for our clients and the world at large.

Yu Murakami, Founder

New York General Group

Abstract Background

Core Technology

CN (Categorical Network)


All of our AIs are based on category theory (we call it CN (Categorical Network)), which has higher performance and wider versatility than common AI models based on vector spaces. More details are available from the technical report below.

Technical Report: A Categorical Approach to Artificial Intelligence

bottom of page