Prediction: State Space Models (SSMs, such as Mamba & co) with their ability to efficiently model ultra long sequences, will be applied to the music/audio modeling tasks very soon and completely transform the domain. The incoming next gen models will be mind blowing.