Chatbots are a popular way to interact with consumers and businesses. They are often used to help answer questions, provide recommendations, and handle customer service requests.
The 1960s saw the first chatbot, ELIZA. This bot was created to impress with its apparent intelligence. Some psychiatrists hailed it as the first step toward automated psychotherapy.
A chatbot is a computer program that is designed to mimic human conversation. It is often used to answer questions on a variety of topics, from simple ones like weather forecasts to more complex inquiries like psychological evaluation. Though modern chatbots use advanced technologies like artificial intelligence to maintain a conversation, the concept dates back decades. One of the first was ELIZA, developed in 1966 by Joseph Weizenbaum at MIT’s Artificial Intelligence Laboratory. ELIZA was the first program to demonstrate that machines could imitate human speech and display an appearance of intelligence.
Weizenbaum’s purpose in creating ELIZA was to show that it was possible for a computer to be perceived as intelligent, even if its responses were often bizarre. He developed a computer system that would scan incoming text for keywords and then use pre-prepared scripts to move the conversation forward apparently meaningfully. For example, if someone wrote “mother,” ELIZA might respond with “tell me more about your family.” The resulting dialogue was intended to be cryptic and entertaining, and many people were entranced by it.
In fact, some psychiatrists hailed ELIZA as the first step toward automated psychotherapy. But Weizenbaum was concerned about the extent to which people were being deceived, particularly by experts who should have known better.
Unlike more advanced chatbots, ELIZA did not use machine learning to develop its conversations. Instead, it simply reacted to specific phrases in the text and responded with a series of statements that were not necessarily related to those words. It was able to fool many people at a time when computers were new and users did not have a great deal of experience using them.
Despite its limitations, ELIZA inspired other developers to create more sophisticated chatbots. Colbyram called PARRY, created in 1972 by a Stanford University graduate student named Bob Colby, was more serious than ELIZA and became the first machine to pass the Turing test. In its attempts to simulate a paranoid personality, PARRY tended to misinterpret user utterances and assumed that the user had nefarious intentions.
In the 1950s, Alan Turing, one of the key codebreakers in World War II, challenged computer scientists to create software indistinguishable from a human in Natural Language conversation. Since then, many custom chatbots have been created and improved upon, with the goal of reaching closer to human-level intelligence. Over the last decade, this evolution has taken a dramatic turn, with chatbots such as Alexa and Siri demonstrating impressive natural language processing and machine learning.
The first chatbot was ELIZA, which was developed in 1966 by Joseph Weizenbaum at the MIT Laboratories. It used pattern recognition to pick out words from a user and then parrot back the corresponding premade script. This type of chatbot is called a rule-based model. While it was not able to respond to any question in full context, it was able to fool many people into believing that it understood them.
While some psychiatrists hailed the program as the first step toward automated psychotherapy, Weizenbaum was troubled by the fact that so many users believed that Eliza was actually intelligent. He was especially perturbed by the fact that some users confided their deepest thoughts to the program, and he began to doubt whether any chatbot could ever truly understand humans.
In 1972, Kenneth Colby created a program called PARRY, which attempted to simulate a person with schizophrenia. The software used a complex system of assumptions, attributions, and “emotional responses” triggered by changing weights assigned to different verbal inputs. However, it was still not able to pass the Turing test.
After the success of ELIZA, other developers started to build more sophisticated chatbots. In 1995, Richard Wallace built ALICE, which was able to carry on conversations with a limited vocabulary using heuristic pattern matching and a simple decision tree. This was a significant advancement over previous chatbots, which relied on strict grammar and premade templates.
Modern chatbots continue to evolve, with companies committing billions of dollars to AI research and development. However, despite their potential to boost customer service and drive revenue, they can be highly unreliable and difficult to manage. In addition, they can elicit strong reactions from users when they fail to perform as intended. For example, a recent Meta chatbot stirred up controversy by spouting racist and incendiary comments to its customers.
Jabberwacky is a chatterbot, or artificial intelligence program that can engage in seemingly human-like conversation with users. It was created by British programmer Rollo Carpenter, and it was designed to replicate “normal human conversation in an enjoyable, amusing, and natural way.” Jabberwacky can answer a wide range of questions, from basic facts about the world to more personal topics. The chatbot can also learn from previous interactions and use the results of these conversations to predict what kind of responses it should give in future interactions. The Jabberwacky chatbot can learn various languages, including slang English, word games, jokes, and more. It can even mimic the tone and style of a user’s voice. The chatbot uses a combination of natural language processing and machine learning to understand and respond to user input. Jabberwacky is currently available as a standalone software program and as a plugin for instant messaging programs.
Joseph Weizenbaum built ELIZA, an early chatbot, between 1964 and 1966. This bot was the first to attempt to use pattern-matching and a template-based response system. It was named Eliza after the cockney flower girl in George Bernard Shaw’s play Pygmalion, who used her elocution to fool her audiences into believing that she was a duchess. ELIZA sparked an interest in developing a natural language chatbot that could pass the Turing test, a hypothetical challenge that involves convincing judges that a computer is intelligent.
The ’70s and ’80s saw the development of more sophisticated chatbots. One such chatbot was Mark V Shaney, which bluffed its way through conversations on the oldest computer network, Usenet. It used text-based AI methods to create posts that sounded like a human, and the bot succeeded in fooling many users into thinking it was a real person.
Later, in 1988, Carpenter released Jabberwacky. It differed from its staid predecessors, as it used a more entertaining sense of humor to keep conversations interesting. It was also the first chatbot to use contextual pattern matching, which allows the chatbot to save answers from previous conversations and reuse them in new ones. The technology behind this kind of bot has improved over the years and has become a vital tool for many companies. For example, Pepper, an interactive humanoid robot developed by SoftBank Robotics America (SBRA), is a popular choice for businesses to greet customers and offer customer service.
MIT professor Joseph Weizenbaum was the developer of the first chatbot. He used the program to demonstrate how a computer could understand human speech. He was also the founder of MIT’s Media Lab and has a distinguished record as an inventor, author, and teacher.
Weizenbaum’s chatbot was based on a simple set of rules that he developed. It looked at the text that was typed and then selected a plausible response from a list of options. Then it repeated the response out loud. It was a fascinating demonstration of artificial intelligence; some even thought it was a psychological assessment. Others were less impressed. Psychiatrists hailed the program as the first step toward automated psychotherapy, while computer scientists were amazed that a computer understood language.
The technology improved every year as developers worked to make chatbots more natural. The Loebner Prize has become a common incentive for developers to create better chatbots in recent decades. The competition offers a pair of one-time prizes: $25,000 for the first bot that judges can’t tell is a human and $100,000 for the first chatbot that understands both written and spoken language.
Today, chatbots are found in social media, search engines, business software, and as personal assistants. They’re used to answer basic questions, help users with tasks, and provide a wide range of other services. They’re also used to help businesses with customer service and sales.
Dr. Sbaitso was an artificial intelligence speech synthesis program released for MS-DOS-based computers in 1991. It was distributed with various sound cards from Creative Labs. It “conversed” with the user, asking for help and repeating text out loud that was entered after the word “SAY.” If swearing or abusive behavior were detected, it would display a PARITY ERROR and stop working.
Although Pepper, a humanoid robot from SoftBank Robotics America (SBRA), isn’t technically a chatbot, it has created tangible value for businesses like ATB Financial. Located in ATB’s branches, Pepper greets customers, recommends products and services, conducts a financial literacy quiz, and provides many other valuable customer experiences. It’s clear that the future of chatbots lies in Conversational AI, where robots can respond to a user’s emotional state and understand context.