 Our devices collect so much personal data each day that people, even large teams of people, can't possibly make sense of it all. So companies have turned to machines to help. Episode 3 – Machine Learning Machine learning is when computers sift through big data sets looking for patterns or correlations that can be used to make predictions or draw conclusions. Machine learning can sometimes turn up strange, ultimately meaningless connections, such as sales going up when the moon is full, or divorce rates in Maine going down when margin consumption decreases. Still, machine learning is being used for increasingly important tasks. Google is using it for almost all of its services, everything from search engine results to driving directions on Google Maps. And it can make connections that aren't just strange, but actually discriminatory. Consider the case of the Google search results leading up to the 2012 presidential election, in which people who typed Obama were shown Obama results in their subsequent searches. But those who queried Romney were not shown any Romney results in subsequent searches. Google said that machine learning had found that people who searched for Obama wanted more Obama results, and those who searched for Romney didn't. For a traditional media company, that kind of disparity would be viewed as providing biased coverage, but it's more difficult to blame a machine for bias. In the era of algorithms, we need to examine the hidden biases of the machines we rely on.