If an article about machine learning begins with dismissing sentences like "data is the new oil", I am interested. Benedict Evans, Partner at Andreessen Horowitz, offers his perspective, and as always, it's valuable. You should read it, but if I try to summarize, Ben makes a case that ML is currently a disruptor to what we used to call "data mining". While data analysis is the goal-oriented examination of data, data mining describes the analysis of data sets without asking a question - trying to find correlations, anomalies and potentially significant insights in data. So ML can help in both areas, give you better anwers to questions you already have, and deliver answers to questions that you didn't ask. And it can broaden your data set as ML makes it possible to include new data types in an analysis, like audio, video or images.
Worth the time to read through it: