On Thursday, IBM’s Watson supercomputer had a practice round of Jeopardy against against Jeopardy champions Ken Jennings and Brad Rutter. Both Engadget and Wired have good write ups of the events (including video), but one thing which hasn’t really been covered is that the face / avatar for Watson was created with Adobe Flash.
Joshua Davis and Branden Hall of Automata Studios worked on the project, creating a visual avatar for the supercomputer that expresses emotion and reacts to the environment. The avatar was created in Flash Professional CS5 using the ActionScript HYPE visual framework, and deployed via Adobe Flash Player 10.1. The avatar connects to the Watson supercomputer via an XML socket server which sends information on the computer’s current mood and state (such as “I know the answer”, “I want to buzz in”, “I won the buzz”, “I got the question right”, etc…). In addition, the avatar receives audio input from Watson’s physical voice by analyzing audio directly from the Microphone (a feature added in Flash Player 10.1). In essence, the avatar is the face of the Watson supercomputer.
On the development side, Branden and Joshua used Flash Professional CS5, along with Illustrator CS5 to develop and design the avatar. They even built a mock Watson using the XML server socket support in Adobe AIR 2.0, so they could test the UI, without being hooked up to the actual supercomputer (the real Watson supercomputer fills up an entire room).
The actual episodes are being filmed tomorrow, and will air on TV in February. Branden and Josh have posted more information along with pictures of the avatar on the HYPE framework blog.