Hmm. The very simplest model you could have of this would be the German Tank Problem [1]. If discoveries of X (e.g. art, hunting tools, whatever) are made at random, i.e. evidence of X is not more likely to be destroyed as time passes, then you are sampling times from a distribution with a maximum of the first invention of X, and the best estimator for this is (m-1)(k-1)/(k-2) where m is the oldest discovery and k is the number of discoveries.
In particular, a new record for oldest art will almost always push your estimate up (as long as k is large so (k-1)/(k-2) is about 1). But you should also be taking into account all the discoveries of art which aren't records. This matters especially when k is not yet big. This page only lists 30-40 pieces of paleolithic art [2].
A better model would take into account that older stuff is less likely to be discovered because e.g. rocks erode. I wonder if anyone has done this.
[1] https://en.wikipedia.org/wiki/German_tank_problem [2] https://en.wikipedia.org/wiki/List_of_Stone_Age_art