You don't even have the beginnings of any meaningful predictive curve with just a few data points. It's just selective function fitting if there's not a rich data set. For example, you can draw many curves through 2 data points, but you can also draw a straight line. The Our World in Data "Primary vs. planted forest" data that he cites for the "Old Growth Forest Cover" chart has just 2 data points (1990 and 2019).