AI in pesticide development?

Via the Wall Street Journal, this article about pesticide development and "AI": 

Bayer and rivals like Corteva and Syngenta say new AI systems are helping speed up the lengthy, complicated and costly process of bringing new chemicals to market. Their efforts include not only new herbicides for weeds but novel fungicides and insecticides as well. Syngenta estimates that AI will reduce the average time from discovery to commercialization by a third—to 10 years from 15 years—and likely decrease the number of lab and field tests by 30%.

Bayer is using an AI system, internally named “CropKey,” to help match the protein structure of a weed with a chemical molecule that targets that structure by sifting through data faster than humans. Molecules selected with CropKey’s help could have a higher success rate during field testing than conventional research, Reiter says. It gives the company an edge—like counting cards in a game of blackjack—and is similar to how drug companies are using AI to speed up the search for molecules that target a particular disease.

Companies say an advantage of molecules selected with AI is that they can be screened during the process for toxicity to humans—a critical point for pesticides sprayed on crops people will eat—as well as environmental safety and cost. 

It'd be interesting to know how different this AI program is from whatever computational chemistry/molecular modeling program they had ten years ago was.