I don’t know if it has been proposed earlier, but the other day I had an interesting idea about creating a variant of Turing Test and mashing it up in a Chataroulette kind of environment. As you may be aware, Turing test comprises of a game where a human chats with an unknown entity (which may be a computer or another human). The job of a human is to figure out whether the entity is human or a computer. It is proposed that if a computer passes Turing test (that is, if it manages to fool the human into thinking that it is a human), then we may call that computer “intelligent”. Unrelated to Turing test, Chataroulette is an online site where random strangers are paired up for chatting.
What if we mix Turing test and Chataroulette and then turn the whole concept upside down? Well, I think the consequences will be pretty interesting. Read on.
THE SETTING: An environment with multiple human and computer (chatbots) participants. A human participant is randomly paired with either a human or a computer.
THE RULES: Human has to convince during the chat that s/he is indeed a computer (chatbot)! After the chat is over (or during it), human can decide whether the other entity is human (who also makes a similar decision) or is a computer (which doesn’t make any decision). For a human, if the other entity decides that s/he is a chatbot, he wins 10 points and other entity loses 10 points. On the other hand, if a human makes a decision the other entity is a chatbot (and it is indeed true), s/he wins 10 points and chatbot (creator) loses 10 points. And if a human decides that chatbot is indeed a human, s/he loses 10 points while chatbot (creator) gains 10 points. In the end, the human or chatbot with most points wins the game (and a prize money?).
THE ESSENCE: So, every human in the game has to act like a chatbot while at the same time discern whether his chatting partner is an actual human (disguising as chatbot) or a chatbot (disguising as human). And every chatbot has to act like a human.
What makes this game interesting is the dynamics of chat and element of suspense. Imagine if you were in this game, you will suddenly start chatting in a goofy fashion (because you are emulating a chatbot). You won’t try to be too-smart (so that you are detected as a human) and you won’t try to be too-dumb (because you will be caught as an obvious deception). As a chat bot creator, you are not just relying on intelligence of your program to win the game. You’d also like to exploit the fact that a human will be over-cautious while declaring the program as chatbot (because humans also disguise as chatbot). So, the actual chatbot need not be a perfect chatting machine!
THE QUESTION: As a human, how do you win the game? What sort of conversation would you hold in this game to disguise yourself as chatbot? And what clues during the chat will tell you that the other entity is a chatbot?