A branch of computer science known as genetic programming has been given a boost with the application of large language models that are trained on the combined intuition of the world’s programmers.

The researchers tested ChatGPT against QuixBugs 40 Python-only problems, and then manually checked whether the suggested solution was correct or not.