What’s the history of chatbots? Chatbots are something novel using AI, or are they? Without adding a spoiler to the article’s details, the first chatbots might even be older than the average person reading this article in front of you now. But what is a chatbot? Simply put, a chatbot is a program that can respond to a user’s query in the form of conservation. Think of it as talking to another person on the other end who responds to your questions or chat but they’re powered by a machine.
These programs are currently used for a variety of purposes. It can be used for an automated help desk, which is quite simple because users only need to click pre-determined options to create a ticket for an issue. For a personal use-case, AI assistants like Siri or Google Assistant can help you book a meeting, create a reminder, set an alarm, and many other things. There are more use cases for them now that they can provide better answers and responses using natural language processing, machine learning, and AI.
The first chatbot
While chatbots have risen to fame only recently, it was in 1950 that Alan Turing developed a method for testing a machine’s ability to perform and converse like a human. He invented the Turing Test, which has become the standard for testing AI by determining if it can talk to people without them realizing it is artificial. After that, the technology was finally introduced first in the 1960s.
The first chatbot, ELIZA, was developed by Joseph Weizenbaum in the Massachusetts Institute of Technology’s (MIT) Artificial Intelligence lab in 1966. Compared to current chatbots, ELIZA uses pre-determined responses based on a keyword. As such, it cannot truly hold a conversation like a human being. There have been other developments in creating a chatbot since ELIZA, such as PARRY in 1972. The huge leap came in the 1990s when we got both A.L.I.C.E and Jabberwacky.
The leap forward
Richard Wallace’s A.L.I.C.E. (Artificial Linguistic Internet Computer Entity) is a natural language processing (NLP) chatbot capable of holding a conversation with a human. It was probably the most advanced of its time when it was released in 1995. Despite failing the Turing test, it was capable of telling a user about its hobbies and other information. Meanwhile, Rollo Carpenter created Jabberwacky in 1981. The main design is to solely pass the Turing test, which involves conversing with a human being. In 1997, an internet version of the same name was released.
They released Cleverbot, Jabberwacky’s successor, a humanoid robot with deeper knowledge in 2008. SmarterChild, another chatbot that debuted in the early 2000s, is likely Siri’s ancestor. It became famous as people who used to have Instant Messaging apps such as AOL, IM, and MSN Messenger were using it for entertainment within the service.
Very early days … working on a new #Cleverbot … A Very Clev? We shall see … a #large #language #model #deep #learning #neural #network #NLP #AI … a #chatbot built to #chat ! pic.twitter.com/qR4DhvxvkU
— Cleverbot (@cleverbot) August 16, 2023
Apple eventually created its own version, Siri, in 2010. People who own an Apple device use it as a personal assistant. As mentioned earlier, it can help you set a reminder, set an alarm, and do other things. Of course, Google launched its own in 2012 under the name Google Now, which was later renamed Google Assistant for Android in 2017.
Finally, we arrived at Mitsuku, a female chatbot created by Steve Worswick. It currently has the most Loebner Prize victories, which is a Turing test competition for artificial intelligence. It won in 2013, and again from 2016 to 2019. It makes use of the framework from Alice’s AIML files, as well as additional user-generated content. Mitsuki is great because the developers actively worked to provide better responses. Currently, it has been rebranded as Kuki.
Now and soon
Currently, ChatGPT and Bard are available to us, both of which passed the Turing Test, indicating significant growth in chatbots. Both can carry on a conversation with a human being, perform research on their behalf, and even debug programming codes. However, because these language models are still in their infancy, people should not use them to obtain factual information.
We've just launched fine-tuning for GPT-3.5 Turbo! Fine-tuning lets you train the model on your company's data and run it at scale. Early tests have shown that fine-tuned GPT-3.5 Turbo can match or exceed GPT-4 on narrow tasks: https://t.co/VaageW9Kaw pic.twitter.com/nndOyxS2xs
— OpenAI (@OpenAI) August 22, 2023
These chatbots cannot distinguish between misinformation and genuine information, as they respond to user inquiries based on what they have learned from their training. Consider chatbots as a high-intelligence search engine that can assist you with research, but you must still do your due diligence to fact-check your sources. After all, fully relying on AI can lead to people exploiting the program.
YouTube: What are the first chatbots in history?
Photo credit: The feature image and pictures, except the screenshot, are symbolic and have been done by Christopher Isak with Midjourney for TechAcute. They are merely decorative and do not depict technology that exists or existed at any point in time.