There are many well-defined tasks that AI can do, but we still need people to solve undefined problems in unexpected, yet practical ways. In other words, creative work. As all designers know, creativity is not a magic trick; it is a result of the iterative design process: brainstorming, prototyping, testing, and improving the result. I present two systems, VisiBlends and VisiFit, that use AI tools to support the iterative design process for one creative task: visual blends - a graphic design challenge to convey messages visually like “Football is dangerous” or “Starbucks summer drinks are here”. These systems support the two key aspects of iterative design: 1) flare and focus to generate prototypes and 2) iterative improvement to turn prototypes into publishable blends. In these system, I show that when AI does low-level procedural tasks, we can augment people’s ability to do the high level tasks of recombining and modifying existing pieces into new creations. In cognitive psychology, this is called augmenting fluid intelligence - a key characteristic of creative work.
Lydia Chilton is an Assistant Professor in the Computer Science Department at Columbia University. She is an early pioneer in crowdsourcing complex tasks on Mechanical Turk. Her current work is in computational design - coordinating people and computers to complete complex and creative tasks that neither computers or individuals can do alone. These tasks involve conveying messages implicitly through text and images, translating and adapting research and ideas to new areas, and finding actionable insights from data.