What I find more promising is methods where hierarchical perception and control emerge naturally from information bottlenecks.<p>Eg <a href="https://www.frontiersin.org/articles/10.3389/frobt.2015.00027/full" rel="nofollow">https://www.frontiersin.org/articles/10.3389/frobt.2015.0002...</a>
I think I read about three articles like this per day.<p>Still haven't seen anyone actually using machine learning in any practical application yet, anywhere. Lots of neat toy demos, no real work seems to be getting done.