Source |Fortune .com | BY: ANNE FISHER
About three years ago, Adobe research scientist Hailin Jin saw a need for a new technology that would make it easy for graphic designers to identify, compare, and choose typefaces. So he created the prototype of a product he dubbed DeepFont. The idea was to replace the often tedious and time-consuming old methods of selecting typefaces with a far quicker and more accurate point-and-click app.
Just one small problem: It didn’t work. “The different machine-learning algorithms we used weren’t accurate enough,” Jin says now. DeepFont was only accurate about 50% of the time, and customers expected at least 90%, he says. So Jin and his team went back to the drawing board, trying different technologies that eventually increased the tool’s speed and accuracy. “We’re over 90% now,” says Jin. “[And] customers like it.”