Haha. I remember a similar story from the Gamma Iota Sigma conference.
I spoke to a carrier data scientist who said in a few years their auto insurance pricing system would have many hundreds, if not thousands, of rating factors that accessed "big data."
For example, one factor might be whether someone is in a Wednesday night mixed bowling league because there is often a lot of beer drinking at these events. This was just a hypothetical, but it illustrates the wide variety of factors that might be used and, of course, calls into question the reliability and relevance of the data and the predictability of the algorithms.