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.

source https://www.engadget.com/mit-datasets-ai-machine-learning-label-errors-040042574.html