The article explores machine vision for human-computer interaction. The exponential decrease in costs of computation and of communication is rapidly leading to convergence and ubiquity. At the same time, inexpensive computing power is enabling a quiet revolution in the machine perception of human action. In the near future, machine perception will converge with ubiquitous computing and communication. This convergence may lead to the widespread introduction of things and environments that see. However, reaping these benefits of ubiquitous perception will require consideration of human abilities and social needs as an integral part of system design. Research in human-computer interaction has developed cognitive theories, design methods and software tools for building useful and usable systems. Scientific results and empirical studies have led to sound principles such as that of direct manipulation. For nearly two decades, however, direct manipulation has been instantiated in the form of the "electronic desktop metaphor," jeopardizing the directness of the physical world.