Google DeepMind‘s AlphaDev, a game-playing AI, has discovered new algorithms that can speed up code by up to 70%. The algorithms were discovered by training AlphaDev to play a game where it had to write assembly code that was correct and faster than existing algorithms.
Assembly code is a programming language that is used to give specific instructions to a computer’s hardware. It is typically written by experts as it can be difficult to write and debug. However, it can also be very efficient as it gives the programmer complete control over how the code is executed.
AlphaDev was able to discover new algorithms for sorting and hashing that are significantly faster than existing algorithms. These algorithms could be used to speed up a wide variety of applications including web browsers, databases, and scientific computing software.
Daniel Mankowitz, a research scientist at Google DeepMind says, “Moore’s Law is coming to an end, where chips are approaching their fundamental physical limits. We need to find new and innovative ways of optimising computing.”
New algorithms for code optimisation
DeepMind published its results of these new algorithms in Nature, which showcase a step forward in the development of AI-powered code optimisation. It could lead to improvements in the performance of software applications as well as open up new possibilities for the development of new software.
In addition to the speed improvements, the new algorithms discovered by AlphaDev are also more efficient in terms of memory usage. This could lead to further improvements in the performance of software applications, as it could free up memory for other tasks.
“We honestly didn’t expect to achieve anything better,” says Mankowitz. “But to our surprise, we managed to make it faster. We initially thought this was a mistake or a bug or something, but when we analysed the program we realised that AlphaDev had actually discovered something.”
The development of AlphaDev is a significant achievement for Google DeepMind. It demonstrates the potential of AI to be used for code optimisation and it could lead to major advances in the field of software development.