I thought chatbots were pretty old and outdated until I *met* Mitsuku last semester as part of another IM class. It was really interesting that this chatbot was in a human form. The homepage states that she’s a “four-time winner of the Loebner Prize turing test, so I gave her a try – or rather, a lot of tries because she actually turned out to be interesting!
Mitsuku’s answers may not always be fitting, but she has a lot of different answers on stock. I find it interesting that this chatbot almost gives an impression of a real person, which makes the conversation more exciting and interesting, regardless of whether the whole content makes sense or not. It’s interesting because you wouldn’t expect such complicated, sophisticated answers from the computer that just isn’t human.
I read a paper a while back that talks about how believable these chatbots have become. Some people have actually started treating these chatbots as their real counselors because of certain traits that they possess. For example, chatbots often repeat some phrases of the sentence that you write, which is what counselors and psychologists often do to turn the conversation’s focus back to the patient/speaker. The preset algorithm makes these chatbots particularly successful in giving the impression that they possess human qualities. I find this human-machine interaction very fascinating and see a lot of potential in developing this field.
Screenshots from my conversation with Mitsuku
