The TensorGame program helped DeepMind to crack the problem of developing faster matrix multiplication. (Source: Adobe Stock)

AlphaZero Program Learns How to Solve Math Problems Faster, Saving Time and Energy Costs

A game-playing DeepMind AI program, AlphaZero, has figured out how to do multiplication faster to speed up machine learning and problem-solving, as well as in displaying images, according to a story on MIT’s By reducing the time for crunching numbers, it will save energy costs and cut costs.

The researchers at DeepMind developed the program and studied how to create “algorithms tailored to Nvidia V100 GPU and Google TPU processors, two of the most common chips used for training neural networks. The algorithms that they found were 10 to 20% faster at matrix multiplication than those typically used with those chips.”

“This is a really amazing result,” says François Le Gall, a mathematician at Nagoya University in Japan, who was not involved in the work. “Matrix multiplication is used everywhere in engineering,” he says. “Anything you want to solve numerically, you typically use matrices.”

LeGall said the challenge was always finding the best algorithm to solve a problem. A game program allowed researchers to turn the problem into a three-dimensional board game, called TensorGame.

“The researchers describe their work in a paper published in Nature today. The headline result is that AlphaTensor discovered a way to multiply together two four-by-four matrices that is faster than a method devised in 1969 by the German mathematician Volker Strassen, which nobody had been able to improve on since. The basic high school method takes 64 steps; Strassen’s takes 49 steps. AlphaTensor found a way to do it in 47 steps.”