Computer scientists have discovered a new way to multiply large matrices faster by eliminating a previously unknown inefficiency, leading to the largest improvement in matrix multiplication efficiency in a decade. This could eventually speed up artificial intelligence models that rely heavily on matrix multiplication.