Engineers at Alphabet's DeepMind AI division have developed the AlphaCode artificial intelligence system.

This system can generate code and solve problems from programming championships. The neural network was trained to understand the problem statement and find its solution.

For training, Alphabet used the GitHub archive with more than 700 GB of code, as well as comments on it in natural language. Then DeepMind organized an internal programming championship, and used its materials to teach AI. He was shown the full cycle: problem statement, working and non-working code, and test cases to verify it. The engineers noted that this approach is not new, but this time they allocated more resources for training.