Very well written and informative article. I greatly enjoyed it. Congratulations.
Here's another (not so) future direction that while nascent I would argue we are starting to see some real connections: The relevant mathematical machinery that is being used to describe and understand certain classes of ML models, and GNN's in particular given their inherent structures, are guiding and giving us deep insights into the computational processes and algorithms of the human brain - not in some hand-wavy abstract way, but grounded in the known neurophysiology. It's as much a fascinating time for neuroscience as it is for ML and AI. What were once abstract topological and algebraic constructions are finding very real-world practical applications towards understanding the brain.