MIT study finds labelling errors in datasets used to test AI

Over three percent of data in the most-cited datasets was deemed inaccurate or mislabeled.

from Engadget https://ift.tt/3lYuT7L