Many years ago, I blogged about programming evolving into a partnership between computers and programmers. These days, GitHub copilot seems to be attracting a lot of attention.
But copilot has many issues, particularly with respect to security and attribution. It reminds me how I tried to learn some complex content planning programs for my thesis (which didn't work) when probabilistic content plans where much readily available to learn from data.
Instead, this idea advocates for creating a new programming language that can benefit from a human/computer partnership. This language can include features from probabilistic programming and possible also from task oriented programming.
And instead of using a large amount of code authored by other people, this partnership can start slow, with the human programmer teaching its AI counterpart how to solve problems in a way that makes sense to them.