Good point, time only moves in one direction so nobody can really be sure whether some piece of history is worth studying.

This came to my mind as I was thinking about data collection in AI and whether there is value in collecting more historical data if it does not improve the performance of an AI model