Very interesting article, and the 14-minute version sounds fantastic! I think you are onto something with this line:
Perhaps computers are facilitating the construction of an additional visual language for music that could help provide imaginative leaps and provoke new musical ideas.
This is already happening in a way, as I believe the state-of-the-art generative AIs are relying mostly on spectrograms to train neural networks. Perhaps it is only a matter of time before they find new visual patterns that are pleasing to the ear.
I am also reminded of this previous thread exploring the idea of using SC code to generate visually appealing spectrograms.