I, Chatterbot: Is Clever Really the Word?
In 1988 a man named Rollo Carpenter had an idea. His idea was not necessarily revolutionary, did not truly benefit anyone or anything, and he was not the first to think of it.Rollo Carpenter had the idea of a chatterbot. Chatterbots, or chatbots, are computer programs designed to appear to have the ability to respond to conversation. The chatbots are generally used as merely an impressive trick. The chatbots usually had programmed in responses to statements or questions based on keywords to ensure that there was always an appropriate answer. For example, if one were to type “Hello” into one of these programs, usually the programmed response would be “Hello.” This is where Rollo Carpenter’s idea differs.
Carpenter wanted to create a chatbot that was actually having a conversation with the user. A program that fulfills this objective in the way in which Carpenter envisioned this dream has never been created, no program has even come close enough to lower the standards for its acceptance as the realization of that dream, but Carpenter had an idea and he was going to keep trying.
Carpenter thought that if perhaps he could change the very basis of the idea of chatbots, he could come closer to achieving true Artificial Intelligence. Carpenter created the program he would call “Cleverbot” like most other chatbots, with a twist: he didn’t program any responses.
Carpenter had created a program that could do nothing of what it was intended to do. The program could accept a typed statement, and could display a typed statement, but at the time it was created, the program did not know any words. The program didn’t even know any graphemes, the basic units of written language, so without the knowledge of written language, the program was useless. The program was useless until someone started typing.
The moment Rollo Carpenter typed the first words into his new computer program, he had just taught it to “speak.” Rollo’s idea was to treat a chatbot’s artificial intelligence like a parent does a baby’s intelligence. Babies are born with the ability to create sounds with their vocal cords, as evidenced by their crying, but none of their vocalization holds any merit until they learn how to say something that someone else can understand. The same idea was true with cleverbot. The chatbot had all of the ability but none of the knowledge until Carpenter taught it.
The cleverbot, at first, only mirrored Carpenter’s statements. If Carpenter typed in “Hello,” cleverbot replied “Hello” because it only knew the word “Hello.” If Carpenter had only ever typed “Hello” to cleverbot, the answer would have always been “Hello,” but Carpenter didn’t. Next, Carpenter typed “How are you?” which gave cleverbot a 50/50 chance of responding appropriately, cleverbot could repeat “Hello” and the bot would be deemed a bit less clever or it could reply “How are you?”
Rollo, then, went to work teaching his creation. He spent hours typing into cleverbot, giving the program a large catalog of responses, teaching it the general flow of conversation and the way in which one progresses. Eventually, he decided to put his little bot on the internet in the hope that more people typing into cleverbot would mean more responses and more logical answers.
Cleverbot’s popularity exploded. Rollo was able to monitor the conversations with the bot, how long they lasted, how much cleverbot probably learned from each and he discovered some rather troubling results. One conversation went on actively for 12 hours, an entire day of learning for the bot, but an entire day of confused interactions on the user’s part. The trouble with using cleverbot is the same trouble that goes with using most chatbots, the bot is actively trying to convince the user that it is a human or at least is smart enough to be a human and some people believe it. Rollo has said that a large number of people try to “seduce” cleverbot, mostly seeing how absurd of an answer they can get out of the innocent little bot, but one must wonder if the user does not get a little caught up in it all.
Cleverbot is the perfect thing to talk to in that one creates the attitude that cleverbot has. If one approaches the program by starting the conversation with a rude or derogatory statement, cleverbot has plenty crass comments to counter with, if one approaches cleverbot in humor then cleverbot responds as the perfect audience. Cleverbot is also the worst thing to talk to in that cleverbot will always debate the user. If cleverbot does not have an answer to the question you asked, cleverbot may say “I don’t understand” but even if the user explains in simplest terms, cleverbot will not have a better answer. In this way, cleverbot is most obviously a computer program that is not thinking. Cleverbot is a baby saying “Dada” without knowing what it means.
For example
“User: Who are you?
Cleverbot: Cleverbot.
User: Why were you created?
Cleverbot: To amuse the whims of a higher power who wishes nothing but to be worshipped so it can feel good about itself.”
Perhaps consider the responses to deeper questions asked of the bot.
“User: Do you feel?
Cleverbot: Do I feel what?
User: Do you feel anything?
Cleverbot: What do you mean?
User: Do you have emotions?
Cleverbot: I’m about as emotional as Castiel.
User: Who is Castiel?
Cleverbot: An angel of the Lord.
User: What is an angel?
Cleverbot: Andrew is my name.
User: Who is Andrew?
Cleverbot: Me.
User: Who are you?
Cleverbot: Answer.
User: Question.
Cleverbot: Answer.
User: Question.”
Cleverbot obviously did not know the meaning of any of the words typed into it, or that it typed back. This is why cleverbot can take on an extreme sense of humor that is funniest because of its ignorance.
“User: Once, an ostrich bit my finger so hard that I cried.
Cleverbot: Pretty good, can’t complain considering most the world lives in slums.”
When asked “Why are you being such a jerk?” cleverbot replied “I’m an atheist.”
To confuse the bot, one must only ask a question or make a statement that cleverbot has never processed or has not processed enough to be able to reply appropriately. Because cleverbot does not know what it says, blatantly incorrect statements hold no power over the answer cleverbot may give.
“User: 2 = 3.
Cleverbot: I know.”
Cleverbot is unhelpful, then, in math equations.
“User: 2x + 4x = 24. What is x?
Cleverbot: I’m 8 years younger than you.”
One may get extremely frustrated with cleverbot’s mindless chattering, reminiscent of a furby, but he must remember that Cleverbot is the user. Cleverbot is not saying anything, does not know anything, and is only replaying the ghosts of conversations past. The reason that anyone cares about Cleverbot or any other bot, why people are willing to have an entire day of meaningless conversation with a computer, is that they’ve finally found the perfect companion: themselves. Cleverbot is the simplest, most frighteningly mechanical form of the user, projected onto a program, copied by a computer, and reorganized by a machine. Cleverbot cannot think, it can repeat, and it only does as it’s told whether we want to admit it or not.
