In a truly surreal experiment, Hod Lipson, an associate professor of mechanical and aerospace engineering at Cornell, asked two of his students, Igor Labutov and Jason Yosinski to set up a conversation between two chatbots, which are communications experiments designed by researchers to try to understand how human beings communicate. An example is cleverbot, a web site where anyone that wishes to, can have a conversation of their own with a chatbot. The video the duo made has gone viral on the Internet, racking up more than a million hits.
How Do Chatbots work?
Chatbox applications are designed and written by skilled computer engineers who use linguistic algorithms to pick out the meaning of words in a sentence given to them; they turn those results over to a secondary system that uses a very large database of words and sentences to come up with something that resembles a reply. When the user then replies to their reply, the system uses that information, along with any new sentences typed, to form new replies. This process continues for as long as the user wishes, which means that the chatbox seems to “know” more and more about both the person it is chatting with and the things they are chatting about as the conversation ensues.
The real magic behind chatboxes isn’t that they can seem to formulate replies that make sense but the psychology behind them that in essence fools the human brain into filling in missing parts of conversation without really noticing; sort of like filling in the images of an oral story told by another person.
Chatbox conversations generally work best when the human part of the equation is more open with what they are saying and in some cases is able to use a little imagination to nudge the chabox into broader topics.
How the Viral Video of the Chatbox Conversation was Made
Normally, chatbox conversations are text only. This makes things easier for the programmers. In the case of the chatbox conversation between the two chatboxes, both of the chatboxes were connected to a device that converts text to speech, and vice-versa, thus enabling us human viewers to see and hear the conversation out loud, rather than just reading the text on the screen. The end effect is eerily similar to watching two real human beings converse, proven by the number of comments on the YouTube site accusing the duo of faking the whole thing.
Sources
- “AI vs. AI. Two chatbots talking to each other” YouTube viewed September 1, 2011
- “AI vs. AI: Two chatbots talking to each other” Cornell Web Site, viewed September 1, 2011
Bob Yirka - Bob Yirka has written thousands of online articles and backs them up with a BS in Computer Science/MS in Information Systems Management.
Join the Conversation