Google AI Cracks Longstanding Math Challenge

October 6, 2022 - 5 minutes read


DeepMind, Google’s innovative AI solution is reapplying its board-game playing AI known as AlphaZero to discover a faster way to solve a fundamental math problem in computer science. And it’s doing such a good job that it just beat a standing world record in mathematical problem-solving, a record that has stood for over 50 years. The mathematical challenge known as solving for matrix multiplication is a needed calculation that is at the core of many different computer applications today. Matrix multiplication is used in an array of computer science applications and AI-based applications such as displaying images on a screen and simulating complex physics models. And as a San Francisco-based app developer squarely focused on artificial intelligence app development

Further, matrix multiplication lays the foundation for machine learning itself, so cracking the code on improving these fundamental operations is key to enhancing the longevity and usefulness that machine learning, AI, and an array of computer science-related functions rely on. François Le Gall, a mathematician at Nagoya University in Japan, noted how incredibly valuable this AI-based breakthrough is for tech development and society as a whole.  “Matrix multiplication is used everywhere in engineering,” he says. “Anything you want to solve numerically, you typically use matrices.” Although these calculations are so important, Nobody knows the best algorithm for solving it,” says Le Gall. “It’s one of the biggest open problems in computer science.”

Despite the calculation’s use in so many areas of technology and science, there is still a layer of mystery associated these this mathematical approach and fortunately, DeepMind is making tremendous headway in solving these sophisticated mathematical operations.

What is a matrix and why is it important?

A mathematical matrix is a grid of numbers that represents anything you want. Sounds simple right? Well, it gets complicated quickly. Matrix multiplication involves multiplying two matrices together by multiplying the rows and the columns of the matrix together to produce a multiplied value of the matrices. Now, once this matrix is multiplied, the basic technique for solving the problem is taught in high school, known as “ the ABC of computing,” says Pushmeet Kohli, head of DeepMind’s AI for Science team. However, when you try to find a faster method of solving the multiplied matrix outside the standard procedural prescribed operations, it gets complicated. 

This is because there are more ways to multiply two matrices together than there are atoms in the universe (10 to the power of 33, for some of the cases the researchers looked at). “The number of possible actions is almost infinite,” says Thomas Hubert, an engineer at DeepMind. To solve this, DeepMind turned this mathematical problem into a game called TensorGame. The board of this TensorGame represents the multiplication problem to be solved, and each move represents the next step in solving that problem. This allowed Google’s DeepMind to consider this challenge under its well-understood design parameters of game theory.

This approach resulted in Google researchers training a new version of AlphaZero, called AlphaTensor, to play this game. AlphaTensor was trained to make the best series of steps to make when multiplying matrices. In doing so, AlphaTensor was rewarded for winning the game in as few moves as possible.

“We transformed this into a game, our favorite kind of framework,” says Hubert, who was one of the lead researchers on AlphaZero. 

What does this mean for artificial intelligence?

This is yet another fascinating data point that artificial intelligence is here to stay and will continue to become more immersive in our daily lives each and every day. For those looking to bring an artificial intelligence-based application into their portfolio of solutions, working with a third party may truly be the best option.

By working with a dedicated development partner like Dogtown Media, you’ll gain access to an industry-leading team of developers that are dedicated to developing an ironclad, security-first AI application that is based on your unique application concept.

Contact us today to learn more about how Dogtown Media can be a trusted advisor as you look to bring your next artificial intelligence app to the marketplace.

Tags: , , , , , , , , , , , , , , , ,