Trouble at the Source: Errors and biases in artificial intelligence systems often reflect the data used to train them.

The article examines the difficulties introduced by errors and biases for the training of artificial intelligence (AI) and machine learning (ML) systems. It states that demands for increased fairness, accountability, and transparency (FAT) in AI is being sought by several groups. The article also discusses the benefits and drawbacks to training that relies upon Amazon's Mechanical Turk program.