When I was building robots in the early 1990s, the problems of voice recognition, image understanding, VOIP, even touchscreen technologies - these were robotics problems.

Around the late 1990s, I'd become convinced that one of the killer applications of robotics came from connecting robots to the Internet. The idea of solving generalized artificial intelligence was still far away, but heck, I could rent brains by hiring operators. iRobot was the name of the company and one of our most ambitious projects, iRobot LE.

The ideal vacuum cleaner would be one you never see. It needs to not just be a cool gadget, but a product that cleans your floor correctly. I can imagine people having a cupboard full of robots that only come out when you need them to fulfil a specific purpose.

The reason it has taken so long for the robotics industry to move forward is because people keep trying to make something that is cool but difficult to achieve rather than trying to find solutions to actual human problems. Technology can be extremely expensive if you don't focus.

Building robot versions of people is very expensive.

When I was building robots in the early 1990s, the problems of voice recognition, image understanding, VOIP, even touchscreen technologies - these were robotics problems.

In the end, robots do things that people can do. So there is a cost above which you can hire somebody to do it, and that bounds the opportunity.