“The machine learning research field would benefit from more scientific inquiry aimed at understanding phenomena, rather than focusing primarily on "hill climbing" benchmarks with new methods.”