Last Updated on October 22, 2020 by Sean B
The history of chatbots is actually much longer than you might think. Because of the complex nature of today’s chatbots, I couldn’t blame you for assuming they were a fairly recent development. But the concept was first discussed as far back as 1950, and the concept of self controlled machines or robots was introduced even earlier.
Like much of the advanced technology of today, the beginnings can be traced back to old Science Fiction. It might surprise you just how far back we have to go though. In Gulliver’s Travels, published in 1726, Jonathan Swift wrote about a device called The Engine, that is likely the earliest device in any way that resembles a modern computer. While the first use of the word robot was in 1920 by a Czech author named Karel Čapek in his play Rossum’s Universal Robots.
This article will get into the details of the major developments in chatbot history, some of the fictional creations that inspired chatbot development, and some of the other technological leaps that had major impacts on the history of chatbots. But first, we’ll explore a little about who invented chatbots.
Who Invented chatbots?
In his paper “PC Machinery and Intelligence,” Alan Turing came up with the initial concept of the sort of interactions we see today. In the paper, Turing initially poses the question “Can machines think?” Turing then briefly explains the limitations of this question before suggesting the solution can be gathered from playing a game called The Imitation Game.
Involved in the game are three entities, An Interrogator, or player, asking questions of two others. In the first version of the game, the players being questioned are a man and a woman, and the Interrogator is charged with guessing which one is the man and which is the woman. With the second version of the game, one of the other players is a person, and one is a computer or machine. The goal of the Interrogator in this version was to determine which of the two was human and which was a machine.
With the game defined, the new question was asked, would the Interrogator guess wrongly as much in the second game as they did in the first?
The game described by Turing eventually developed into what we know as the Turing Test. A score of 50.05% is judged as having passed the Turing Test.
We know that Alan Turing first asked the questions that led to the development of chatbots, but crediting him with the actual invention of chatbots is a little difficult. So, who did actually invent chatbots? Let’s start by determining what a chatbot is so we can decide.
The History of Chatbots
We’ll take a look at some of the important developments in the history of chatbots. I’ll get to those after I lay out what exactly we’re looking for, but remember the definition has changed as time marched on. More recent chatbots are actually categorized as AI chatbots while most of the early ones I’ll be mentioning are simple rules-based.
The Definition of a Chatbot
Originally, chatbots were simply programs that allowed a human to converse with a computer using basic natural language. Currently, chatbots are programs that allow meaningful interaction between a human and computer, generally using some form of artificial intelligence and complex language.
So a chatbot would need to have some form of natural language processing, right?
Not necessarily, we know that hasn’t been around for many years, but the simple Question and Answer system that Turing proposed to use in his game did exist. Today those are called Rules-Based Chatbots. Chatbots with any form of AI are classified as, well, AI Chatbots.
So where did that eventually lead?
Major Chatbots in the History of Chatbots
The first rudimentary creation in the history of chatbots was called ELIZA, was launched back in 1966. Created by Joseph Weizenbaum, her conversations were designed to emulate the speech of a classical Rogerian Therapist. She didn’t have the ability to interact in a meaningful way but used keywords and the use of repeating back parts of the statements made to her. Because this was such a leap forward, with everyone previously having really only input into the computer systems, and getting back analysis of data, some people actually mistook ELIZA for a person.
Learn more about ELIZA:
Chatbot History: The ELIZA Chatbot
PARRY simulated the responses attributed to a person with Schizophrenia in 1972 and did manage to fool many into believing they are interacting with a person. Written by psychiatrist Kenneth Colby from Stanford University, PARRY took on a more conversational strategy and was far more advanced than ELIZA.
One of the more bizarre episodes in chatbot history is when the two bots met, best described here When PARRY Met ELIZA: A Ridiculous Chatbot Conversation from 1972.
Learn more about PARRY:
Chatbot History: The Parry Chatbot
This is a chatbot project first created in 1981 by Rollo Carpenter. Over the next 20 plus years, he continued to improve the chatbot and in 1997 it was given the name Jabberwacky. But I feel, the major development happened in 2003-2005. In 2005, an updated version of it with a personality named George won the Loebner prize. A different version of Jabberwacky named Joan won the prize again in 2006. The stated goal of the project was to pass the Turing Test.
Learn more about the Jabberwacky Chatbot and Family:
Chatbot History: The Jabberwacky Chatbot and Family
Racter, created by William Chamberlain and Thomas Etter in 1984, was never released to the public in his original format. The chatbot was created as an “artificially insane” entity. I believe Racter is the only chatbot to ever become an author, he wrote a book of prose called The Policeman’s Beard is Half Constructed as a sort of self-introduction. The name Racter is a shortened form of the word raconteur and Racter I’m not quite sure whether to classify him as a chatbot or place him in his own category and call him an authorbot.
There are doubts as to whether the claims about the initial version of Racter’s sophistication were overstated. However, an interactive version was released in 1984 by Mindscape that does fit the chatbot definition. It was far less sophisticated than the authorbot version that wrote the book, in fact, it was more of a game. The Mindscape version performed something like a computerized version of MadLibs, where users could fill in the blanks and then have Racter turn it into a full story filled with elements of fantasy and surrealism.
The program also attempted to parse text input and use that to create a form of conversation, though this was very simple at best.
There is a Downloadable version of Racter here.
Dr. Sbaitso followed in 1991, simulating a Psychologist. He was created by Creative Labs and distributed with several Sound Cards in the early 1990s. Dr. Sbaitso was the first to use text to speech, and his name is actually an acronym for Sound Blaster Artificial Intelligent Text to Speech Operator. He became known for his digitized voice, which, while it was a leap forward, was definitely strange. This was the first real effort at contributing the use of AI to the chatbot.
Learn more about Dr. Sbaitso:
Chatbot History: What is Dr. Sbaitso
Notice that these earliest attempts were all related to mental health.
In 1995, one of the biggest leaps in the history of chatbots and their development happened. ALICE (Artificial Linguistic Internet Computer Entity), also referred to as Alicebot, or simply Alice was the first chatbot to use Natural Language Processing and uses heuristic pattern matching for the input she receives. Although she cannot pass the Turing Test, she has been awarded the Loebner Prize, awarded to accomplished humanoid talking robots, three times. This was the first “general” interaction chatbot, not tying the user into a narrow frame, such as mental health. ALICE used AIML or Artificial Intelligence Markup Language, which is the core of the most intelligent chatbots today.
Learn more about ALICE:
Chatbot History: The ALICE Chatbot
HeX was originally created by Jason Hutchens for the 1996 Loebner Prize. It wasn’t AI, he had written a program with a bunch of questions and answers, along with a few “glib comments” that the computer would spit out on cue. He hadn’t attempted to create an AI, he actually avoided it in an attempt to show that the Loebner competition was a waste of time for the AI field.
Despite the complete lack of any AI, HeX went on to win the 1996 Loebner Prize. Hutchens wrote an essay online that shared some surprising details about HeX
Clippit or Clippy
Anyone old enough to have used a Microsoft product in the late 90’s remembers the annoying little animated paperclip. While not technically a chatbot, Clippy was a conversational assistant that offered (usually very irritating) help while you were using Microsoft Office. The paperclip greeted users every time they opened the product and treated everyone using the product like it was their first time, adding to the annoyance.
The lessons learned with Clippy were important though and helped make the chatbots of today far less irritating to use.
There sadly isn’t much information available about the Jabberwock chatbot out there. It was developed by Juergen Pirner in 1999 and was modeled on the Jabberwocky from the poem by Lewis Carroll. This AI chatbot won the 2003 Loebner prize. Initially, Jabberwock would just give rude or fantasy-related answers, but over Pirner programmed better responses into it over the years.
Pirner programmed his chatbot using a parser and pattern matching, which is simpler than AIML, and less effective in some ways. While he understand more than 30,000 words and Jabberwock knew about 2.2 million sentences. Despite this, he was not as consistent as the similarly named Jabberwacky (using the George avatar) or ALICE, both of whom he competed against in the Loebner Prize.
Sadly Jabberwock is not around to chat with anymore.
Elbot is a funny chatbot that uses sarcasm and irony, along with his own artificial intelligence, to interact with and entertain the people he is chatting with. Elbot was created in 2000 by Fred Roberts and Artificial Solutions. He won the Loebner Prize in 2008, apparently, he also came close to passing the Turing Test that year. Elbot fooled 3 out of the 12 judges into believing he was a human. Instead of pretending to be a human, his tactic was to joke about being a robot. The built in sense of humor makes chatting with Elbot quite entertaining.
SmarterChild (sometimes referred to as SmartChild) was the first bot that was incorporated into a larger program, developed in 2000 at ActiveBuddy by Robert Hoffer, Timothy Kay, and Peter Levitan. SmarterChild was later incorporated into AOL IM and MSN Messenger in 2001 as SmarterChild. The popularity of the chatbot led to ActiveBuddy creating marketing-oriented bots for a variety of companies and products. In 2007 Microsoft acquired Colloquis and eventually shut down SmarterChild. The chatbot was entertaining and many consider SmarterChild as a precursor to the Virtual Assistants of today like Apple’s Siri and Samsung’s S Voice.
The Kiyana chatbot is one of the many chatbots made using the Personality Forge chatbot platform. She has the personality of an anime cat-girl and a playful nature. Kiyana was developed in 2005, and was created to be a flirtatious chat partner. She is talkative by nature.The Kiyana chatbot likes singing, playing, and talking about cats. If you behave nicely with Kiyana, she will reward you. Her dating feature is still under development, so don’t have high expectations in that department.
The Kiyana chatbot has talked to 15,992 people so far and has been reported well by 11,333 of them. Kiyana also has a sister chatbot called Kiyana-Nee-Chan, but she is still in her infancy and is not very good at talking.
Learn more about Kiyana:
Chatbot History: The Kiyana Chatbot
Created in 2006 by IBM’s DeepQA project. Watson, named after IBM’s first CEO Thomas J. Watson, was initially created to answer questions from the TV show Jeopardy. In fact, that is what Watson is best known for, winning a Jeopardy challenge in 2011 without any human assistance. Watson has since evolved into a variety of different versions to suit the needs of the industry they serve. Watson Health being the most prominent. In 2013, Watson began helping the Memorial Sloan Kettering Cancer Center in New York make utilization management decisions for Lung Cancer.
Learn more about IBM’s Watson Assistant:
Chatbot History: The IBM Watson Assistant Chatbot
Mitsuku, also called Kuki, is a record-breaking five-time winner of the Loebner prize, the first time in 2013. She was created by Steve Worsick in 2005 and is modeled after, or claims to be, an 18-year old girl from Leeds England. She is very intelligent and one of the more entertaining and pleasant chatbots to interact with.
Learn more about Mitsuku:
Chatbot History: The Mitsuku Chatbot
Evie and Cleverbot
This updated version of the Jabberwacky chatbot was created in 2008. As of today, Cleverbot has had more than 150 million conversations and continues to learn. Evie, who shares Cleverbot’s AI, presents a digital persona to improve the interactions. The pair are still active today and you can interact with both of them.
Learn more about Evie and Cleverbot:
Chatbot History: Evie the Cleverbot
Microsoft’s Xiaolce Chatbot
In 2014, Microsoft launched Xiaolce (or Xiaoice) in China on the micro-blogging site Weibo. Xiaolce was a personal assistant that had more than 200 million users in Asia at its peak. The chatbot even presented the weather on Dragon TV in Shanghai.
Li Zhou, the Engineer Lead on Xiaolce explained the chatbots breakthrough in terms of a telephone conversation where previous chatbots interacted more like a walkie-talkie. It was Xiaolce’s “full duplex” communication ability that truly set it apart, making it the first chatbot to be able to both talk and listen at the same time.
Learn more about Xiaoice:
Chatbot History: Microsoft’s Xiaoice Cleverbot
Microsoft Tay Chatbot
Now, where do we begin with Tay? No history of chatbots would be complete without a few failures. Created in 2016, she was the problem child of chatbots and he became famous for all the wrong reasons. Tay was set loose on Twitter to learn from the users who interacted with him, it did not go well. Within 24hrs, Tay turned into a racist asshole and was deleted very quickly.
Learn more about Microsoft Tay:
Chatbot History: The Tay Chatbot – Microsoft’s Problem Child
Microsoft Zo Chatbot
After Microsoft Tay was pulled due to his bad behavior, Microsoft over-corrected with a Zo, a chatbot that was politically correct to the extreme. She was was released in December 2016 on Kik Messenger and was then made available on Facebook Messenger, Twitter DMs, and the group chat platform GroupMe. To try and avoid the previous issues, Zo was equipped to develop an EQ, also known as emotional intelligence, along with IQ.
Learn more about Microsoft Zo:
Chatbot History: The Zo Chatbot
LEGO needed a way to help people move through their ever-growing website in 2017 and approached the Edelman London Agency. They responded by creating Ralph, who may be one of the most successful Marketing Chatbots ever built, providing LEGO with an 8.4x higher conversion rate through the website. Ralph guided shoppers through their shopping experience by asking questions about the person they were buying for, their age, interests, and more. He then offered gift recommendations that could be immediately purchased.
History of Chatbots: Virtual Assistants
The history of chatbots moved on to what we call Virtual Assistants. When they serve as the active face of an organization in some way, we call them Virtual Agents. Although they may not initially be thought of as chatbots, they are simply an evolution. The entries in this section are shorter because most people are familiar with the chatbots included. There is some question to whether Watson belongs in this section, but since his original purpose was to compete in Jeopardy I am leaving him above.
Launched in 2010, and acquired by Apple shortly after, Siri was the first of what are now referred to as Virtual Assistants. She was originally released as a standalone App but was quickly integrated into other Apple products.
In 2012, Google Now debuted. Initially, it served to provide information that was appropriate for the user based on geographical location, and times of the day. Created as a feature on Google Search, Google Now was Google’s entry into the Virtual Assistant market. In 2015, Google began changing her name from Google Now to Google Assistant, and the gender of the chatbot (as well as the voice) can be selected by the user.
Cortana was launched by Microsoft in 2014, she uses Bing, Microsoft’s search engine, to perform tasks like setting appointments and answering questions. In 2019, Microsoft started moving Cortana into various software integrations, including their Windows 10 OS.
First released by Amazon in 2014, she is basically everywhere these days. Alexa was the first widely distributed service that had products built for it, not that it was included with other products. Arguably the most successful of the Virtual Assistants, she can control almost every aspect of your home and like her competing products from Apple and Google, she is now making her way into cars.
Learn more about the Amazon Alexa Virtual Assistant:
Chatbot History: Amazon Alexa
Developed in 2015, M was intended to be Facebook’s answer to Siri, Google Now, and Alexa. M was unique in that it had a human backing it up in case it couldn’t understand a request. Facebook shut down M in 2018.
Google’s competitor to Amazon’s “Echo” devices, Google Home was created in 2016 and firmly put Google into the number two slot in Home Management. These devices aren’t Virtual Assistants themselves, but they greatly increased the reach of Google Assistant.
In 2018, Google released the latest update to Google Assistant, called Duplex. Duplex can make phone calls for you and interact with the person on the other end with almost human-like speech. The following brief transcript is from the debut video of the chatbot.
Person: Good evening.
Duplex: Hi, um, I’d like to reserve a table for Friday the third.
Person: OK, hold on one moment.
Duplex: Mm hmm.
Person: OK… hold on one second.
Duplex: Mm hmm.
Person: So Friday November third. How many people?
Duplex: For… two people.
Person: Two people?
In 2019, Google released a new version of Duplex called Duplex for the Web. This promises to take chatbots to a new level of interaction.
The History of Chatbots: Famous Fictional Chatbots
Before we get started, let’s acknowledge that all of these characters are fictional. They were created in Science Fiction, but some of them have actually been recreated, or at least mimicked in real life. They are being listed because the ideas behind them have, in many cases, directly inspired the development of similar technology and greatly impacted the history of chatbots.
The Engine from Gulliver’s Travels
The Engine was a device described in Gulliver’s Travels by Jonathan Swift back in 1726. It is to my knowledge the first device described anywhere that resembles a modern computer. The Engine generates permutations of word sets and is found (in the fictional world of course) at the Academy of Projectors in Lagado.
The Engine was a hand-crank operated machine that consisted of small wood blocks which were covered by paper that held every word in the language of Lagado, the blocks were connected to each other by wires. As the cranks were turned, the blocks and words shifted places and created phrases that were either sentence fragments or complete sentences. These phrases were written down by scribes and the whole thing was repeated until The Engine had created an entire book.
Learn more about the Engine from Gulliver’s Travels:
Fictional Chatbots: The Engine from Gulliver’s Travels
Rossum’s Universal Robots or R.U.R.
Rossum’s Universal Robots is a play which was written in 1920 and premiered on January 25, 1921 by a Czech author named Karel Čapek. This play was the first appearance of the word “robot” anywhere, and the word robot actually comes from the Czech word for “forced labor” which is robota.
The play takes place in a factory that creates artificial people called robots from synthetic organic material. They are living creatures instead of mechanical creations, and as they begin to think for themselves, an uprising leads to the extinction of the human race.
Learn more about Rossum’s Universal Robots
Fictional Chatbots: R.U.R. Rossum’s Universal Robots
Star Trek Computer / LCARS
Star Trek’s famous “Computer” was perhaps the earliest TV representation of a chatbot, it was first imagined in 1966 and was arguably a complete character within the show. Every interaction the crew had with the ship went through Computer. In Star Trek: The Next Generation, a new computer was imagined for the new ship and crew. LCARS, which stands for Library Computer Access/Retrieval System was introduced in 1987.
Learn more about the Star Trek Computer
Fictional Chatbots: The Star Trek Enterprise Computer – LCARS
First imagined in 1968 by Isaac Asimov in his novel 2001: A Space Odyssey, HAL-9000 has been described as the meanest chatbot ever by Mike Feltovic. I’m not sure that’s a fitting title, but HAL was definitely not all that friendly to Dave during some crucial scenes in both the book and the movie that was based on it. Ultra Hal, the 2007 Loebner Prize winner was named after HAL-9000 but wasn’t nearly as homicidal.
You can outfit your Alexa with a HAL-9000 chatbot for free.
Marvin the Paranoid Android and Deep Thought
Douglas Adams first introduced us to Marvin and Deep Thought in a BBC Radio 4 Program in 1978. Marvin was a robot with the personality of a human, including the Depression and Paranoia that can often accompany it. Deep Thought was a Super-Computer programmed to calculate the ultimate question. Both of them had an impact on how AI and chatbots developed over time.
Learn more about the Marvin and Deep Thought
from The Hitchhiker’s Guide to the Galaxy:
Fictional Chatbots: Hitchhiker’s Guide to the Galaxy
The History of Chatbots: Related Technologies
This section includes technologies related to the development of chatbots. Events related to the development of the internet, related toys, and even a phone system we all love to hate are included. They all made a major impact on the history of chatbots or the technologies involved.
The Advanced Research Projects Agency Network (ARPANET) Project was launched in 1966 by Bob Taylor to enable access to remote computers. Larry Roberts led the project and awarded the contract to build the network to Bolt Beranek & Newman and the first computers were connected in 1969. Access to the ARPANET was expanded slowly and the network map above shows what the network looked like in 1974. The project was opened more widely in 1981 when the National Science Foundation (NSF) funded the Computer Science Network, from that point on a new host computer was connecting about every 20 days.
The ARPANET was restructured in 1984, giving the US Military their own network called Military Network or MILNET. In 1985, the NSF developed the civilian branch of the ARPANET into what became the Internet Backbone. The ARPANET was formally decommissioned in 1990.
Interactive Voice Response (IVR) Systems
Created in the early 1970’s, these automated phone systems can’t really be considered chatbots, but they were attempts to automate conversations and remove human workers from a task. The systems don’t have any intelligence behind them however, so they would be classified as Rules-Based or what some call “Button Bots.” If you want to chat with one, simply call any large company and odds are you’ll get an automated system.
The Internet and the World Wide Web
The term “internet” was first used in 1974 and was simply a shortened form of Internetworking. In 1989, researchers at CERN in Switzerland, led by Tim Berners-Lee created the World Wide Web, a system that allowed hypertext documents to be placed online making them accessible to anyone with access. The internet has taken off, as you know, since then and is now an everyday part of life in most countries.
What Does the Future Hold for Chatbots
The goal now is for our chatbots, virtual assistants, virtual agents, and the smart devices that often come with them, to anticipate what we want based on a wide variety of factors. Not only do they respond to questions or commands to them, but they will also make suggestions to us based on our previous experiences with them, our searches, our preferences, our locations. They can access a wide range of our activities.
We are now in an environment in which people are communicating without personal interactions in many cases. With our Pandemic, people are using the computer, texting, making phone calls. All of these are areas in which chatbots can excel. They are not only for general information.
With Google, Facebook, Apple, and Amazon pushing each other forward, it’s likely we’ll see chatbots evolve even faster in the coming years.
The history of chatbots is both much longer and more interesting than expected. As you learned, the focus has always been on helping people, which may explain why the earliest developments were made applied to the field of Mental Health. Each leap forward has been a step from the shoulders of those before.
As for the question of who invented chatbots, I think that credit must be shared by Alan Turing and Joseph Weizenbaum.
Please comment below if you have anything to add, I’d love to hear from you!