Apparently they are planning to create their own machine learning model with all the font data they have gathered.