One such chatterbot is Eliza. Eliza attempts to sound like a Rogerian Psychotherapist, and sounds most intelligent in the kind of conversation that would be had in such a therapy session. (Rogerian Psychotherapy is a technique that revolves around trying to understand the client by asking questions and reflecting their statements back to clarify that they"ve been understood).
Click here to use the Eliza chatterbot
You will probably agree that while Eliza sounds somewhat intelligent at times, she gives very vague and general replies, forgets earlier parts of the conversation, and says some things that sound very strange! As an example, here is a conversation with Eliza:
Eliza: Hello, I am Eliza. Human: Hello I am a human Eliza: Do you believe it is normal to be a human? Human: Yes Eliza: Are you Sure? Human: Yes Eliza: Why did you repeat yourself?
Eliza often gives very general replies by looking for suitable rules based only on the line you"ve just typed, and that often leads to some very weird conversations, such as the one above! In this case Eliza used pattern matching to determine that it might be appropriate to ask me "Do you believe it is normal to be a human?", as I had said I was a human; if I had said "I am a banana" then the same rule would have responded with "Do you believe it is normal to be a banana?". When I answered "Yes", Eliza didn"t actually understand the meaning of me saying "Yes", so simply asked me if I was sure, as it has a general response that says a good way of responding to somebody saying "Yes" is to ask them if they were sure. Eliza also has rules that are triggered if somebody repeats themselves that say she should ask them why. Because Eliza doesn’t actually understand meanings but instead just follows rules that say what might be appropriate responses, she had no way of knowing that it actually was meaningful for me to respond with "Yes" twice! This is a big giveaway that Eliza really isn’t that intelligent, and doesn’t actually understand meanings in conversation, or even whether or not a response to her own question is actually reasonable.