CORALLIS, OR -- Researchers at Oregon State University may have found a way to use artificial intelligence to help protect bees during the development of new pesticides. 

Typically, pesticide developers expose a group of bees to a new chemical and if more than 50% survive, it’s deemed nontoxic for these critical pollinators. But, OSU Assistant Professor of Chemical Engineering Cory Simon says this new AI can predict bee toxicity in a fraction of a second, "Instead of conducting 1,000 bee toxicity experiments, which is time consuming, expensive, and yeah it kills a lot of bees, we could narrow down to a few pesticides that our model predicts not to be toxic for the experiments."

He says the system uses data from past experiments conducted on bees, "It learns to find patterns in the molecular structure of the pesticides. And then, when a new pesticide molecule comes along, it looks for the patterns in that pesticide to see if it’s more similar to the toxic or nontoxic ones that we observe in the experiments."

Simon believes the machine could be expanded to be useful in an infestation emergency, "Where there is a pest invading a field and they need to make a quick decision ‘which pesticide should we use among these new ones?’ And they haven’t had experiments yet, our model could be used to predict which pesticide to apply with the lowest risk to bees."

While herbicides, fungicides and insecticides are widely used on Oregon crops, those same plants rely on bee pollination. OSU estimates almost 100 commercial crops in the U.S. would vanish without bees to transfer pollen needed for reproduction, so protecting them from becoming collateral damage to pesticides, researchers say, is critical to the protection of our food supply.