Dina Katabi still encounters everyday problems with computing, even though she is the Class of 1947 Career Development Associate Professor of Electrical Engineering and Computer Science and a member of MIT’s Computer Science and Artificial Intelligence Laboratory.
“I admit that I had been sending my print jobs to a graduate student, just to get around a simple configuration problem. Then one day we realized that a hundred people in this building had already set up their printers. So why not automate the task?”
Along with Nate Kushman, her graduate student, Katabi recently developed a novel system called WikiDo, software that builds a database of automated solutions for every essential computer task. Described as “crowd-sourcing for IT,” WikiDo delegates troubleshooting to the Internet masses.
Katabi and her team recently received an award from the MIT Deshpande Center for Technical Innovation, which she hopes will help get WikiDo ready for the marketplace. The Center helps emerging technologies reach the marketplace by funding novel, early-stage research and connecting MIT’s innovators to the business community.
WikiDo works thanks to the contributions of ordinary, non-expert users. Whether users are attaching a file to an email message or setting up a virtual private network, WikiDo records their graphical user interface actions, or GUI actions, and translates them into automated solutions. These solutions are stored in an online database, should a fellow user wish to perform a similar task.
“We’re leveraging the collective intelligence of social media with machine learning,” Kushman says. “Average users answer how-to questions by performing the task, rather than documenting it.”
In addition to collecting and disseminating IT solutions, WikiDo is translating the existing corpus of Internet help boards into WikiDo tasks, all while filtering for mistakes and machine differences. “Non-expert users rarely complete tasks the same way each time. They may also stop and check their email or abort the task. WikiDo filters for the correct and most efficient steps necessary to get the task done on any given system,” she says.
WikiDo could put an end to user frustration, one of the computing industry’s most vexing problems. “No one likes to be on hold with tech support,” Katabi notes, “and for the companies, staffing those lines is tremendously expensive. There’s also a need among home users who don’t have access to expert help.”
This realization led the professor to explore WikiDo’s commercial potential, which has piqued considerable interest among the entrepreneurial community at MIT.
“We wrote a business plan and entered the $100K Entrepreneurship Competition in collaboration with Prof. Regina Barzilay and a grad student,” says Katabi, “[and] we were semi-finalists. This was a great opportunity and a tremendous learning process.
“WikiDo captures the best practices and solutions and automatically implements them on your computer. It’s really different, really exciting, and something that could only happen at MIT.”