Artificial intelligence apps today are built upon deep-learning algorithms that use of patterns found in data. The problem is that bias can find its way into the deep-learning process at various stages, including the period before data is even collected. Eliminating bias is key to the eventual success or failure of AI.