DeepMind AI Learns the Art of Coding

Google’s Deepmind is using artificial intelligence (AI) to optimise coding by speeding up sorting algorithms in C++, a popular coding language. Known as AlphaDev, the system uses reinforcement learning — previously used by Deepmind to create AI that excels in games such as chess and Go — to learn by doing. The system is used to optimise simple but significant algorithms, such as sorting, hashing and basic tasks, which have been optimised by humans for decades. Since AlphaDev’s C++ sorting algorithm was introduced in April 2020, it is up to 1.7% more efficient than previous methods when sorting long sequences of data and can be up to 70% faster for shorter sequences.