A recent study in London by researchers at DeepMind has resulted that Artificial Intelligence (AI) is capable of finding shortcuts in a fundamental type of mathematical calculation. It does so by turning the equation into a game and leveraging the machine learning techniques that other companies’ AIs used to defeat human players in several games.
Branch of Knowledge
AlphaTensor – the AI developed by DeepMind was designed to perform a way of calculation known as matrix multiplication. In it, the multiplying numbers are arranged in grids or matrices that may represent sets of pixels in images or any other internal workings of an artificial neural network.
It also can optimize matrix multiplication for specific hardware. The agents were trained on two different processors, one when it took fewer actions and the other when it reduced runtime. In comparison with previous algorithms, the AI sped up matrix multiplications by several percent.
This generic approach could apply to various types of mathematical operations, say the researchers, like decomposing complex waves or decomposing mathematical objects into simpler ones.
“This development would be very exciting if it can be used in practice,” says Virginia Vassilevska Williams, a computer scientist at Massachusetts Institute of Technology in Cambridge. “A boost in performance would improve a lot of applications.”
Grey Ballard, a computer scientist at Wake Forest University in Winston-Salem, North Carolina, sees potential for future human–computer collaborations. “While we may be able to push the boundaries a little further with this computational approach,” he says, “I’m excited for theoretical researchers to start analyzing the new algorithms they’ve found to find clues for where to search for the next breakthrough.”