A leading researcher aims to develop an electronic brain with the linguistic ability of a six-year-old child by 2003.
The system has been designed to acquire, understand and use language in the same way that humans do.
John Taylor, emeritus professor of mathematics and director of the King's College, London, Centre for Neural Networks, said the approach might ultimately be capable of generating a limited form of consciousness.
While most artificial-intelligence technologies in the field were based on rules and a set vocabulary, Professor Taylor said, the language acquisition device, or Lad, attempted to simulate the linguistic functions of a real human brain.
It uses neural networks, linked in sequences that simulate the flow of communication. The upshot is an adaptive system that learns through association and example, building up its capabilities through its interaction with its environment. This, in turn, is influenced by inbuilt drives - such as hunger or curiosity - determined by the programmers.
"Lad will interact with whoever is with him, developing the phrases that he wishes to say in order to solve the problems he faces," Professor Taylor said.
Professor Taylor and King's College colleague Neill Taylor (no relation) have already brought Lad to the level of an 18-month-old child.
Lobal Technologies, founded to exploit the technology, is looking for a second round of funding.
Brady Anderson, chief operating officer, said Lad could provide a software platform for computer games allowing designers to create highly interactive characters.