10+ years ago companies were hoovering up data for ML - trying to find correlations in high-dimensionality data. Mostly the results were garbage but occasionally you hit on a real, unexpected phenomenon.

Nowadays you just throw all the data into a black box and believe whatever it says blindly.