Revisiting Unreasonable Effectiveness of Data in Deep Learning Era

research.googleblog.com