THE CRUNCH
A new institute dedicated to AI safety has been announced, aiming to use mathematical proof techniques to establish safety guarantees for artificial intelligence systems. The Mathematical AI Safety Institute (MAISI) was founded by Canadian mathematician Jacob Tsimerman, a recent recipient of the Fields Medal. The organisation intends to apply the same rigour used to prove the security of cryptographic codes to the un
The institute plans to prove that AI systems behave safely, rather than relying on empirical testing alone. This approach mirrors the methods used in cryptography, where codes are proven unbreakable through mathematical certainty. By focusing on formal verification, MAISI hopes to address the inherent unpredictability of large language models.
The initiative comes from a mathematician with a strong background in number theory, suggesting a focus on foundational and structural safety properties. This methodological shift could provide a more robust framework for ensuring AI reliability in critical applications.