Researchers open-source benchmarks measuring quality of AI-generated code

A new study finds that large AI language models are beginning to learn to code, albeit with a few mistakes.