When I was growing up, I always wanted to be a writer like Tolstoy or Agatha Christie—until I got my first period. I realized then that women must endure much more pain than men to fulfill the human reproduction responsibilities. So, I wanted to become a geneticist who could make men pregnant to share the pain with women. My "ambition" was shattered when I was rejected by the biology department of the university I applied to. Instead, I became a computer science student without ever having seen a computer before.
Initially, I wasn't fond of computer science. The turning point came when I was studying for a master's degree at Michigan State University. Lionel Ni sponsored me to develop a graphical user interface (GUI) for power management, while the late Carl Page taught me artificial intelligence (AI) and advised me on two AI projects. I loved the dynamic nature and instant gratification of building graphical user interfaces and was attracted to the magic of AI. After completing these projects, I was determined to pursue a Ph.D. to develop AI to create GUIs in support of better human-computer interaction (HCI).
Steve Feiner at Columbia University "bravely" admitted me as his Ph.D. student despite my miserable GRE score (I credit my enthusiasm and persistence). I loved working at the intersection of AI and HCI and wrote a thesis on an automated visual storyteller called IMPROVISE, which takes data, user tasks, and user visual preferences as input and automatically generates an animated series of visual illustrations to explain the data to users based on their tasks and visual preferences.
IMPROVISE was part of an intelligent system used at New York Presbyterian Hospital to automatically generate multimedia patient briefings for nurses. This real-world application gave me a glimpse into the future of a hybrid world: humans augmented by AI beings that would do what humans can't or don't want to do (for example, a study shows that the automated patient briefings were better than those created by human physicians). With my colleagues at IBM Research, I co-created another AI being, RealHunter, the world's first conversational AI information assistant, which can aid users, such as data analysts at the Department of Homeland Security, in exploring and analyzing large datasets through a natural language conversation. To extend machines' abilities of deeply understanding their users, we also built IBM Watson Personality Insights, the first commercial system that can automatically infer a user's Big 5 personality traits, values, and needs from the user's text communication.
Creating one capable AI being is non-trivial; enabling anyone to create his or her own custom AI beings is astronomically harder. My current startup Juji aims at democratizing the creation of AI beings. We want to empower everyone to customize and adopt their own AI beings without writing any code.
Imagine a world with AI beings: at a personal level, everyone could have AI beings that understand their unique interests, needs, and personality, and aid them in various aspects of life, from providing career guidance to healthcare tips. At a societal level, AI beings could serve as human aids to carry out tasks that humans don't like to do (for example, talking to strangers 24/7 and giving them instructions) and scale out human time infinitely.
Although I didn't get a chance to change the human race as a biologist, I'm fortunate to become a computer scientist who could augment the human race by making AI beings accessible to advance humanity. Looking back at my career, I'd like to thank the three Ps that have guided me thus far:
The Digital Library is published by the Association for Computing Machinery. Copyright © 2023 ACM, Inc.
No entries found