What is novelty?

And can a GenAI system generate it? I think that a lot of folks have a knee-jerk answer to this question, in both directions (“Of course!”, “Definitely not!”). However, it’s not as simple as it seems on the surface. Novelty in language is easy. The likelihood of a sentence having been said, written, etc. before in exactly the same sequence is very low, because of all of the options available as you go word by word. But novelty in science is hard. A new idea that builds on a foundation of validated results is a totally different type of novelty.