Submitted by chandan on Tue, 09/24/2019 - 17:42

Many concerns have been raised regarding biases, dogmas, and prejudices that AI can learn. It can happen owing to bias present in training data fed into machine learning algorithms. It can also happen owing to biased and bigoted AI designers. A recent news about ImageNet, one of the most wisely used annotated image data sets for machine learning, removing 600K images from its database to remove alleged racist bias, is worth a read and some thought: