QA1WATT

[From David Hamill] The purpose of QA1WATT is to produce a subset of the full Turing test, i.e. a particular sort of restricted Turing test. Turing's original idea was an interrogation, with the judge asking questions and two entities (one human, one machine) answering them. The judge is meant to decide which entity is the human.

Two restrictions are proposed: 1) each entity must provide a single-word answer; 2) the single word is to be chosen from a specified wordlist. Currently we're only arguing over the size and scope of the wordlist; some think a small list would be good, others a large list.

My thought is that in a Turing test, the judge would have a copy of the list, as would the human confederate (with a suitable interface to ensure compliance), and the bot would use an electronic version of it. The judge would be able to ask questions in free form on any subject, but the entities would have to use the agreed wordlist for replies.