The purpose of this blog is to document my experiments in building robot brains. My hobby and my dream is to build robots like the ones in science fiction; the robots of Asimov’s robot stories or Robby the Robot in Forbidden Planet. Such a robot would need a brain with encyclopedic knowledge, common sense reasoning and natural language understanding. This kind of technology has been “20 years away” since the 1960s. Now, in the twenty-first century we don’t seem much closer to building intelligent autonomous robots. I don’t really expect to see robots like this in my lifetime much less be able to build one myself as a lone hobbyist. My realistic goal is to learn a bit about linguistics and cognitive science while working on some interesting software. If I happen to build something useful along the way, that would be a bonus.
Natural language conversation is my current focus. Perhaps this is the hardest part of building a robot brain. Language seems so natural that human beings take it for granted but language is a terribly complex behavior for a robot. I’m starting out with a study of linguistics rather than current techniques in Natural Language Processing or Computational Linguistics. Most current work focuses on statistical methods or machine learning. Statistical methods are effective for applications when the underlying problems are not well understood but statistical methods are a brute-force approach that give little insight into the processes being modelled. To achieve the linguistic performance that is my goal, I believe a deeper understanding of language is required.
I have no illusions of creating a robot like Robby in Forbidden Planet, who spoke fluent English and was “at your disposal with 187 other languages along with their various dialects and sub-tongues.” But as Commander J. J. Adams noted, “Colloquial English will do fine, thank you.” If I can even endow a robot with that linguistic skill, it would be a major accomplishment.