AlphaDev, an artificial intelligence developed by Google’s DeepMind, has created a new way to sort data that is 70% faster than algorithms created by humans. This is important because it could make millions of software programs run faster.
Data sorting is important for computers to arrange data in a specific order. There are many sorting algorithms, but humans have a hard time creating new ones because they are already so optimized.
AlphaDev’s Unique Approach
AlphaDev is different because it creates new sorting algorithms instead of improving existing ones. It uses a language called assembly code, which is easier for AlphaDev to create more efficient algorithms.
AlphaDev creates one instruction at a time and tests its output for correctness while also trying to generate the shortest algorithm. This is difficult because of how many different instruction combinations there are.
When AlphaDev was asked to create a sorting algorithm, it found a method that was 70% faster than the best algorithm for sorting five pieces of data. Even when sorting 250,000 pieces of data, it could be 1.7% faster than the best algorithm.
DeepMind has open-sourced AlphaDev’s algorithms and added a common code base called Libc++, which means anyone can use them immediately.
AlphaDev’s Impressive Results
Daniel Mankowitz from DeepMind said AlphaDev may be able to make up for the end of Moore’s Law by increasing efficiency.
Mark Lee from the University of Birmingham said AlphaDev is interesting, and even a 1.7% speed increase would be useful. But he’s skeptical that this approach can make up for the end of Moore’s Law because it’s unlikely to make the same kind of progress in more complex software.