What are you working on?

I work on developing robots. I began by working on mobile robots, then spent about two decades working on humanoid robots, and now I’m really interested in personal robots. Personal robots will empower people to increase productivity and, ultimately, become their own automation engineers. Basically, any person in a home or small business will be able to use a personal robot to help them do their work. Right now, robots like the ones we see in factories are dangerous. They can kill you unintentionally. So rather than developing
more robots for these same types of big industrial tasks, I’m working on ones that can interact safely with people in ordinary spaces.

If you look back 25 or 30 years, computers were in backrooms. Regular people didn’t interact with them. The people who oversaw these machines functioned as mysterious and mandatory go-betweens, akin to the way the priesthood has served as the medium between the congregation and the divine. Then came a tremendous liberation that allowed everyday people to use computers for all sorts of things––like e-mail and the Web––that even the original designers had never envisioned. I would like to empower people with mechanized physical action in the same way that we’re currently being empowered with new actions for communication and computation.

A big breakthrough with personal computers occurred when offi ce workers were suddenly able to program spreadsheets. They didn’t consciously think about programming, they just looked at the screen and entered the numbers. Now I want people to be able to show a robot what they want done, and have it execute the task for them. I want them to talk to it, but not through a user’s menu. The machines have to be completely interactive and give direct and useful feedback.

The major technical diffi culty in this fi eld is making the robots dexterous enough to touch and feel things, and also understand what those objects are. My goal is to build these machines with the visual recognition capabilities of a two-year old child and the manual dexterity of a six-year old child. Two-year olds can name all the objects in a room regardless or whether or not they were previously familiar with that environment. Six-year olds can tie their laces and accomplish all the manual tasks we require of our robots.

[ ... ]
for full text see VISIONS - MIT Interviews book