Verified Functional Analysis for Safe AI

The objective of the VeriFunAI project is to create a mathematical foundation for developing verifiably correct ML frameworks that enable the construction of safe AI systems.

Artificial Intelligence (AI) is becoming an integral part of our daily lives, from self-driving cars to medical diagnosis systems. However, ensuring that these AI systems are safe and reliable is a significant challenge. The VeriFunAI project aims to address this challenge by developing a solid mathematical foundation for creating AI systems that are both safe and error-free.

Our approach combines advanced mathematics and computer science tools to verify the correctness of AI systems. Specifically, we will use Interactive Theorem Provers (ITPs) to formalize mathematical theories and develop new methods for training AI systems to meet safety requirements. This work will focus on a type of logic called Differentiable Logics, which can help ensure that AI systems behave as expected, even in uncertain or changing environments.

By collaborating with leading experts in Japan and the UK, this project will contribute to the development of safer AI technologies, benefiting society and advancing the field of AI research.