I am part of a 40+ author team that published this perspective article. In this work, we discuss and recommend bio-inspired approaches that we believe are crucial to in our quest to create machine learning techniques that are capable of learning throughout their operational lifetime. This desired approach is much the same way humans learn as opposed to most current ML algorithms which having training and inference phases.
It would be a thrill to hear the HN community's opinions and ideas as I continue this research.
Disclaimer: We came together as part of the DARPA Lifelong Learning Machines (L2M) program.