The Machine Intelligence Research Institute (MIRI) is a U.S. nonprofit research institute that studies AI alignment and existential risk from artificial intelligence. It was founded in 2000 by Eliezer Yudkowsky and others as the Singularity Institute for Artificial Intelligence and adopted its current name in 2013. It is based in Berkeley, California.
The institute initially concentrated on the mathematical foundations of alignment, including agent foundations and decision theory, and influenced the formation of the AI safety field. More recently it has shifted its emphasis toward policy proposals and public communication aimed at slowing or halting the development of AI systems that surpass human capability.