David M. W. Powers
The world of science fiction has long known computers, robots and spaceships that can converse like humans, pass the Turing Test with ease, and interact with their environments with superhuman intelligence. There is an entire spectrum from a Data to a HAL to the ubiquitous Star Trek computer that will dim your lights, serve you tea - or commence the autodestruct sequence. Many of these, like Data, HAL and Astroboy have learned most of their knowledge, including their language capabilities, rather than being programmed [Clark, 1972; Ishiguru 1962], paralleling both early and recent research into the acquisition of language and ontology [McCarthy, Earnest, Reddy and Vicens, 1968; Block, Moulton and Robinson, 1975; Feldman, Lakoff, Stolcke and Hollback Weber, 1990; Steels and Brooks, 1995]. This paper describes a program of research into language acquisition using robot babies and intelligent rooms, and summarizes some preliminary results and applications.