Machine-learning model fed web content makes racist and sexist associations

Human biases exposed by Implicit Association Tests can be replicated in machine learning using GloVe word embedding, according to a new study where GloVe was trained on "a corpus of text from the Web." (more")