Are there any good human experiences in reviewing AI-generated code?

There was quite a long discussion on a similar topic earlier this year.