New institute tackles AI safety with mathematical rigor
mathematical ai safety institute founded to develop safety standards for ai, aiming to prove ai systems are safe like cryptographers prove codes are unbreakable.
the mathematical ai safety institute (maisi) was announced by jacob tsimerman, a recent fields medal recipient, with the goal of developing rigorous safety standards for ai systems. maisi plans to start operations in january 2027 with a team of 10-30 mathematicians in the san francisco bay area. the institute aims to prove ai systems are safe and responsible, similar to how cryptographers prove encryption schemes are unbreakable. tsimerman also joins openai's safety team, emphasizing the need for a much higher level of safety standards in ai.