I looked at this interesting paper and I'll be first to admit I don't fully understand whats going on. The concepts some of which seem sort of familiar are a bit new and foreign. That being said a lot of researchers seem excited about the direction of this paper and according to benchmark scores the idea is looking like a valid enhancement to the neural network approach to AI.
For people more experienced and more focused on this line of AI research please feel free to evaluate this research result.
References
- Dynamic Routing Between Capsules
https://arxiv.org/abs/1710.09829