What Chatbots talk when Humans aren´t around-Video-NSMag

The following information is used for educational purposes only.






What Chatbots talk when Humans aren´t around


Computers can converse well enough to fool people into falling in love with an imagined human. But in listening to the random conversation of one chatbot talking to itself, a robot-on-robot love connection seems rather unlikely.

Igor Labutov, Jason Yosinski and Hod Lipson at the Creative Machines Lab at Cornell University connected a chatbot called Cleverbot to itself - just for an "afternoon hack", Yosinki says.

Cleverbot is a web server that has been running since 1997. A human enters text and the bot searches its memory of over 20 million past conversations for the most appropriate response.

In the video, two Cleverbots exchange text and a program converts the text to audible speech. Another program syncs an avatar's movements with the speech.

During their conversation, the two bots quickly discover that the other is a robot, although one declares itself a unicorn. After some passive aggressive comments, the conversation turns to God and philosophy, with one bot regressing to name calling after deciding the other is not helpful. Ninety seconds later, one abruptly blurts "Au revoir".

v The team ended the conversation there to keep it from lasting indefinitely, as Cleverbot does not understand that "goodbye" politely ends a conversation.

What's most interesting is that viewers' reactions to the video, as reflected in the comments on YouTube, attribute human characteristics to the avatars, Lipson says.

The group is thinking about releasing an extended conversation or planning a public debate between many different chatbots.

Artificially intelligent chat programs have yet to completely con humans. An annual competition, called the Loebner Prize, offers $25,000 for the first program to trick judges into believing that they're conversing with a human. The prize has gone unclaimed for 20 years.