Until recently, the only way to detect a flu epidemic was by accumulating information submitted by doctors about patient visits. This process took about two weeks to reach the CDC - much too long to be effective. In an effort to improve speed, researchers flipped the process and tried using big data. By using data from billions of Google searches over a period of five years, they were able to make amazing connections. They found that the number of flu related searches correlated with the number of people who had the flu. Most impressively, they could accurately predict flu outbreaks up to two weeks before the CDC by identifying specific search terms.