You have probably talked with a robot, more often than you might realize. Your phone is likely filled with bots such as Siri or Google Assistant. In this series of blogs, I explore how chatbots work, why they work and explore what that means for us humans. This second episode will be a short introduction to chatbots and how they came to be.
The First Chats
The first notable chatbot is ELIZA, created by Joseph Weizenbaum in 1966. Weizenbaum used Rogerian psychotherapy as a foundation for the program, using keywords from the user to pose reflective questions. He originally wrote the DOCTOR program in ELIZA as a parody, with the intention of showing the superficiality of human-machine interaction, but he was up for a surprise, as people actually enjoyed talking to ELIZA and became attached to her, often forgetting she was a machine.
Jabberwacky, launched in 1997, works a little differently. It stores all conversations and is able to learn to talk more appropriately using contextual pattern recognition techniques. It is even capable of learning foreign languages! It led programmer Rollo Carpenter to create the more recent Cleverbot.
Inspired by ELIZA, Richard Wallace developed A.L.I.C.E. (Artificial Linguistic Internet Computer Entity) in 1995. He created AIML (Artificial Intelligence Markup Language), an XML dialect, designed especially for the development of natural language software agents such as A.L.I.C.E. Since then, AIML has been a popular language, it has also been used for Mitsuku, a chatterbot which has won the Loebner Prize for ‘most human-like’ chatbot 3 times! Mitsuku contains A.L.I.C.E.’s AIML files, but can also play games and is of course, able to learn.
Bot or Not?
One of the ideas that had the most influence on the development of (chat)bots, is the ‘Turing Test’, developed by mathematician Alan Turing in 1950. Turing is often seen as the ‘father of theoretical computer science and artificial intelligence’ and his ideas and inventions have made a profound impact on the world. The Turing Test is a method of measuring the intellectual capacity of a bot. The bot converses textually with a human, who does not know whether his/her conversational partner is human or not. If the bot passes for a human, it has passed the test. Although criticized because of e.g. its dependency on the knowledge of the human, it laid an inspiring basis for the development of a humanlike (chat)bot, although this ambition has existed for centuries. The Loebner Prize uses the Turing test to award prizes to the programs that seem most human.
Accompanied by Personal Assistants
Fast forward to the now, a lot has been developed since. Even Facebook uses ‘chat extensions’. Siri, released by Apple in 2010, uses machine learning techniques such as neural networks to learn. The same counts for e.g. Amazon Alexa (2014) and Google Assistant (2016). These are not simply chatbots anymore, but function as personal assistants, taking text and speech commands on phones, being able to fulfill tasks and control your devices. Now you have your own personal smart chatterbot assistant in your pocket! It makes me wonder, how far will we let them integrate into our lives? How intimate are we able to get with our bots?
In the next episode, I will introduce PIP, the chatbot that talks about love, created by Ine Poppe for ROBOT LOVE.
Further Reading:
- Ine Poppe interviewed Joseph Weizenbaum, creator of ELIZA, for the Dutch newspaper NRC.
- In 1973, PARRY, a paranoid schizophrenic chatbot was developed by Kenneth Colby, and talked with ELIZA…
- More about AIML (Artificial Intelligence Markup Language)
- BBC – Intelligent Machines
- Talk to ELIZA
- Talk to JABBERWACKY
- Talk to CLEVERBOT
- Talk to A.L.I.C.E.
- Talk to MITSUKU
- Talk to PIP
Episodes:
