Judging by the comments section of this blog, some of you readers are actually robots. Well, today's your day because we're going to be talking all about you. Chatterbots are computer programs designed to have conversations with humans. The best chatterbots use some form of artificial intelligence to learn how to generate better dialogue over time. Rollo Carpenter has developed a chatterbot which learns from the most chaotic and difficult conversationalists: internet users. Cleverbot is Carpenter's latest chatting AI and uses a growing database of 20+ million online conversations to talk with anyone who goes to its website. At this stage in its development, talking to Cleverbot is like having a text conversation with a monkey tied to a typewriter as it is being flung down a flight of stairs. Eventually though, automated conversationalists will become a staple of entertainment and business. Already robotic voices answer the phone at many large corporations. One day Cleverbots may provide companionship that we won't be able to discern from the real thing.
Rollo Carpenter has developed other chatterbots in the past. Jaberwacky is a program much like Cleverbot, though it learns differently. Avatars generated from Jaberwacky, Joan or George, have won the Loebner Prize for chatterbots (2005, 2006), and Cleverbot took second place in 2009. Carpenter's companies Icogno and Existor are developing virtual presences for businesses. What makes Cleverbot so deliciously different, however, is that it learns only from the people it talks to. There is no other input or restriction on its growth. As iCub is doing with robots, Cleverbot is taking chatterbots to their most basic learning strategy: just repeat what other's have said or done. While it is in its infancy, this makes Cleverbot very primitive, but as it continues to aggregate more and more conversations, it may soon have a database of dialogue so complex that it can talk about any subject with anyone.
In a way, talking to Cleverbot is a little like talking with the collective community of the internet. And it's just as horrific, funny, and boring as that may sound. Carpenter has even placed a disclaimer right on the website warning users that they are getting the regurgitated musings of the world, and the program should be used "at YOUR OWN RISK".
But let's assume for now that you can handle talking to the echoes of the online masses - how does one go about interacting with Cleverbot? You have three options. Typing into the Google-like box and selecting "Think About It!" will prompt Cleverbot to read your words and generate a response. You can continue to type chat with Cleverbot for as long as you like. "Think for Me" will have Cleverbot respond to itself for the last statement it made. This is handy when you find yourself flabbergasted by Cleverbot's lack of coherency. Clicking "Thoughts So Far" will recall past conversations you've had with Cleverbot. For some really great examples of Cleverbot witticisms, follow the "cleverness" link. There are humorous records of conversations that the bot has shared with others.
As you may expect, people are less than kind, and often less sensible, when they know they are talking to an artificial entity. Since Cleverbot is just the sum of its past conversations, this means that it is often less than kind and very rarely sticks to any topic. It once spontaneously made fun of me for not knowing Tagalog. Still, even with all the crazy sentences, non sequitors, and idiotic ravings, I found myself wanting to keep chatting with Cleverbot. I wasted most of an hour testing if it knew various bits of physics lingo. It did. After a while, I really enjoyed interacting with cleverbot.
I'm not alone. Chatterbots have been known to keep people interested, and keep them coming back to talk again and again. According to Existor, the longest real world conversation with a chatterbot lasted more than 11 hours...with almost continuous interaction between the computer and human. This bodes well for the programmers looking to improve the way that humans respond to artificial programs. Certainly most of us dread the robotic voice on the phone when we call a bank or insurance company. But what about witnesses we interview for a fictional crime? Carpenter's chatbots are used for a facebook application (221 B) that lets you play the roles of Sherlock Holmes and Dr. Watson investigating a mystery. When Cleverbot becomes as clever as Holmes, we may all be clamoring to hear what it has to say. For now, I'm interested to know what the spybots, data miners, and spamming programs have to share in the comments section. It's like my own little private AI experiment in there.
[screen capture credits: Cleverbot, Icogno]