zlacker

This comment is downvoted, but it makes an important point. AI systems that produce an outcome that can be easily verified by non-experts are far more practical. If my mom can get an illustration out of the AI that she wants, she is done. Not so for software, where she cannot really verify it‘s that going to reliably do what was specified.

This is especially true for complex pieces.

If an AI could produce a world-class totally amazing illustration or even a book I will afterwards easily see or read it.

On the other hand real-world software systems consist of hundreds of thousands or lines in distributed services. How would a layman really judge if they work?

Nevertheless I also expect AI to have a big impact since less engineers can do much more.