An Opinionated Guide to ML Research (joschu.net)

In a recently revisited essay from 2017, the author shares invaluable insights for aspiring machine learning (ML) researchers, emphasizing the importance of choosing impactful research problems and balancing time management. The guide outlines a three-part strategy to success: selecting the right challenges, making consistent progress, and fostering personal growth within the research community. The author argues that developing a keen sense for significant problems is even more crucial than raw technical abilities, and individuals should cultivate this intuition by observing which ideas thrive or fade in the ML landscape. The essay also contrasts two approaches to research: idea-driven and goal-driven. The author advocates for the latter, as it encourages unique perspectives and can lead to greater breakthroughs. By providing case studies from their own PhD work, particularly in robotic manipulation and locomotion using reinforcement learning, the author illustrates how focused goals can guide researchers towards innovative solutions that advance the field. This perspective not only fosters collaborative efforts but also highlights the potential for ambitious, non-incremental improvements in ML, ultimately inspiring the next generation of researchers to push the boundaries of what is achievable in the domain.