I believe the correct configuration for a robot right now is a mobile device with no onboard intelligence, connected using Wi-Fi, to the processing power of a home computer.

At a stroke, this paradigm unleashes the unused processing power of home computers and solves the practical impossibility of putting processing power on a small robot. Robots with little onboard processing will only ever lead to robots with little intelligence. Wi-Fi is the key to robots today because Wi-Fi allows 'splitbots' to exist easily.

Robotics has been held back until cheap wireless existed. Urgently -- very urgently -- we must unleash the power of independent programmers on robotics problems. Only in this way will there be an explosion of machine intelligence development, as there was an explosion of software development from 1980. When the commonplace home computer is one half of the robot, this becomes immediately possible.

The future of products such as small household robots is to be a splitbot: there will be a stationary processing box, probably running a Unix variant, connected by Wi-Fi to the mobile unit. Indeed, the future of home computers is to be the stationary half of a splitbot.

Sketches:

Early Design Renderings:

Prototype: