Cherry Blossom predictions are in! Will AI bloom or bust?

Students, researchers, and citizen scientists from around the world submitted predictions to George Mason University’s annual International Cherry Blossom Prediction contest, with one new notable entrant….AI.  

Jonathan Auerbach, an assistant professor in the Department of Statistics and the competition co-organizer, said, “Obviously we see AI as the calculator of the 21st century—students should be able to use this tool, but that raises the baseline, so students really have to up their game.” 

The typical competitive entries use the concept of growing degree days, looking at heat accumulated by trees as measured by local temperatures. Although Auerbach said one trend he’s seeing is contestants increasingly modeling conditions in the fall preceding the bloom. 

“We know that springtime temperatures are important, and as climates warm, blooms are happening earlier and earlier. But fall is also important because trees create (or form) buds in the fall," said Auerbach. "You can actually saw off a branch of a cherry tree and put it in your closet and as long as it's warm enough and in water, it'll actually bloom.”  

Auerbach gets intrigued by what he called “zany” approaches, noting that people use unique data points for their predictions and that hotel prices have been used to predict bloom dates with some effectiveness. This year one entry used Google trends, and Auerbach noted there is some precedent for that—for some time Google searches for the flu predicted flu outbreaks, for example. 

What most entrants foresee is that the Washington, D.C., cherry blossoms will reach peak bloom in late March. Even ChatGPT, which you could say is really branching out—is in agreement.