DeepMind AI Cracks the Code of a Long-Standing Math Problem

When AI \”played\” math, it cracked an Internet chokepoint

A new study reveals that an artificial intelligence system at Google’s sister company DeepMind found a way to solve the fundamental math problem at heart of modern computing. Researchers say that a modification of AlphaZero, the game engine used by Google to beat chess legends and chess masters in Go, outperformed a 50-year-old algorithm.

The research focused on multiplying matrices, which are grids of numbers. The matrix multiplication operation is essential for many computations, including processing images, recognizing voice commands, training neural network, running simulations, and compressing information to share on the Internet.