A piece of AI software called Alice is proving jolly good at pretending that she is human. The software is part of an artificial intelligence competition being held at the University of Reading, west of London.

Ten volunteer judges will sit at split-screen monitors and carry on two conversations at a time. One will be with a human, while the other might be Alice, or Cleverbot, or maybe Ultra Hal. After five minutes, the judge will be asked to say who — or what — is behind each screen.

Alice, short for Artificial Linguistic Internet Computer Entity, has a two-gigabyte brain and is therefore brighter than my last girlfriend, has a stored range of canned phrases, synonyms and commands for answering easy questions and dodging difficult ones.

Richard Wallace, Alice's creator, said Alice stood a fighting chance of fooling human judges because "People are more like robots than we would like to think."

Wallace said one of his favorite parts of the competition was seeing his creation spit out a startling answer to a tricky question. He said there are times when the computer gives responses that are completely unexpected, and you see a spark of life in there.

We guess until she goes out all night and he does not know where she is.

More here.