Microsoft is seeking to patent a deep learning-based code generation platform that uses reinforcement learning to fine-tune AI models for producing more relevant and useful source code. The system rewards or penalizes the model based on its performance in several code-quality factors, including syntax, errors, readability, and comprehensiveness. This approach aims to address bias and quality issues often found in conventional code generation systems.
