Introduction
Do you remember your first interaction with a chatbot? To be honest, I don’t, and that means only one thing – it was not impactful enough (either in a good or a bad way).
Chatbots have been around for quite some time, in earlier days mainly as products of academic research, but since the internet Era they started to become more visible, and nowadays they are everywhere.
While some chatbots catch our attention immediately upon visiting a website, others remain a bit shy and hide on the contact page. However, it is quite common to encounter chatbots in the customer support offerings of most businesses.
In this post, I look at the early attempts at creating an intelligent chatbot, and how we got to the recent developments in the field.
The Turing Test
In a previous post, I explained what conversational AI is, which included the concept of machines simulating human conversation – However, a crucial question arises: how do we determine or assess what truly defines a machine in this context?
The answer to the above rhetorical question is The Turing Test.
The Turing Test, proposed by the British mathematician and computer scientist Alan Turing in 1950, is a test designed to assess a machine’s ability to exhibit intelligent behavior indistinguishable from that of a human. The test serves as a benchmark for evaluating the development of artificial intelligence. [1]
The original idea of the test involved a setup where a human evaluator engages in text-based conversations with two participants: a human and a machine. The evaluator is unaware of which participant is human and which is a machine. If the machine is capable of generating responses that are indistinguishable from the human participant, then it is said to have passed the Turing test. [1]
The above interpretation is not consensual, as there is plenty of debate on the original conclusions from Turing’s paper.
The first time I came across the Turing Test was in the movie “Blade Runner“. I was quite young but the entire sequence where Rick Deckard (played by Harrison Ford) is interrogating Leon Kowalski in an attempt to find out if Leon was a Replicant is quite mesmerizing. And if you are screaming – “hey, it’s Voight-Kampff, not Turing”, you are correct, that was the name of the test in the movie, which is in fact an implementation of the Turing Test.
“Are you a replicant?”
Blade Runner 1982
Curious to know which questions [3] the test had? Here you go.
- Itβs your birthday. Someone gives you a calfskin wallet.
- Youβve got a little boy. He shows you his butterfly collection plus the killing jar.
- Youβre watching television. Suddenly you realize thereβs a wasp crawling on your arm.
- Youβre in a desert walking along in the sand when all of a sudden you look down, and you see a tortoise, itβs crawling toward you. You reach down, you flip the tortoise over on its back. The tortoise lays on its back, its belly baking in the hot sun, beating its legs trying to turn itself over, but it canβt, not without your help. But youβre not helping. Why is that?
- Describe in single words, only the good things that come into your mind about your mother.
- You’re reading a magazine. You come across a full-page nude photo of a girl. You show it to your husband. He likes it so much, he hangs it on your bedroom wall.
- You become pregnant by a man who runs off with your best friend, and you decide to get an abortion.
- You’re watching a stage play – a banquet is in progress. The guests are enjoying an appetizer of raw oysters. The entree consists of a boiled dog.
Now go ahead and give your favourite chatbot a run
So, if we want to be able to measure how human-like a chatbot is, then the Turing Test seems to be a solid approach to help us qualify it. Whether or not passing the Turing Test qualifies for a great conversational AI experience is something else, but nonetheless, it has been a benchmark that stood the test of time, and there was even a competition (Loebner Prize) giving awards to the best computer programs.
ELIZA (1966)
ELIZA chatbot was a groundbreaking computer program developed in the 1960s by Joseph Weizenbaum at the MIT Artificial Intelligence Laboratory. It is considered one of the earliest examples of natural language processing and artificial intelligence applications.
The purpose of ELIZA was to simulate a conversation between a human user and a psychotherapist. It aimed to mimic the dialogue of a Rogerian psychotherapist, who typically reflects the user’s statements back to them in the form of questions or prompts, encouraging further exploration of their thoughts and emotions.
ELIZA’s primary purpose was to demonstrate how a computer program could engage in conversation and simulate human-like interaction.
The chatbot worked by analyzing the input text and identifying keywords and phrases using simple parsing techniques. It then used predefined patterns and rules to generate responses. ELIZA’s approach was to respond with open-ended questions or paraphrase the user’s statements to keep the conversation going.
The chatbot’s choice of a Rogerian psychotherapist persona was a smart one. It could recognize keywords in user input and reflect their questions back, acknowledging and validating their original statements. This created understanding and empathy in the conversation, providing a supportive and meaningful experience within the chatbot’s limitations.
The conversations with ELIZA often revolved around the user’s feelings and concerns, and the chatbot would attempt to keep the user engaged by showing empathy and understanding. However, ELIZA had limited understanding and could only provide preprogrammed responses based on patterns it recognized. It did not possess true comprehension or the ability to generate original thoughts.
Despite its limitations, ELIZA had a significant impact on the field of artificial intelligence and human-computer interaction. It showcased the potential for computers to engage in conversation and raised questions about the nature of human-computer interaction, language understanding, and the boundaries of machine intelligence.
ELIZA inspired the development of subsequent chatbots and It laid the foundation for the evolution of chatbot technology and its applications in various fields.
You can give ELIZA a test drive here: http://psych.fullerton.edu/mbirnbaum/psych101/Eliza.htm
Trying something from 1966 in 2023 for the first time definitely yields a different emotional response, but one can imagine the impact such a program had in these early AI days.
It used a rule-based approach, where certain works would trigger certain responses, in the good old fashion of conditional expression. ELIZA was written with MAD.
PARRY (1972)
PARRY was one of the early chatbot programs developed in the 1970s by psychiatrist Kenneth Colby. It was designed to simulate a person with paranoid schizophrenia and engage in conversation with users. PARRY aimed to demonstrate how a computer program could exhibit human-like responses and behavior, particularly in the context of mental health.
It seems there was a trend in early chatbots to address mental health issues, but their primary purpose was to advance the understanding of human-computer interactions rather than serve as practical therapeutic tools.
PARRY employed a rule-based approach, using a set of pre-defined patterns and responses to generate conversation (same technique as ELIZA). PARRY would respond to user inputs based on these rules, attempting to emulate the speech and behavior of an individual with paranoid delusions.
PARRY was noteworthy for its ability to engage users in coherent and interactive dialogue, even though its understanding and responses were limited to the predefined rules it operated on.
While PARRY was a significant contribution to the field of artificial intelligence and natural language processing at the time, it should be noted that its purpose was more centered on simulating a specific mental condition rather than serving practical or therapeutic purposes.
Parrys’s source code can be found here: http://www.cs.cmu.edu/afs/cs/project/ai-repository/ai/areas/classics/parry/
PARRY meets ELIZA (1972)
PARRY and ELIZA interacted several times. The most famous of these exchanges occurred at the ICCC 1972, where PARRY and ELIZA were hooked up over ARPANET and responded to each other. [6]
Dr. Sbaitso (1992)
Dr. Sbaitso is an early chatbot program developed by Creative Labs in the 1990s for MS-DOS-based personal computers. The name was an acronym for Sound Blaster Acting Intelligent Text-to-Speech Operator
It was bundled with the Sound Blaster sound cards, which were popular audio devices for personal computers at that time. Dr. Sbaitso was primarily designed to showcase the capabilities of the sound card and demonstrate text-to-speech technology.
The fact that is come bundled with a sounds card meant it was the first one to starting to hit the masses, outside the dome of research institutes.
The chatbot played the role of a “psychiatrist” – and engaged users in text-based conversations. Users could type messages to Dr. Sbaitso, and the chatbot would respond with synthesized speech, using the computer’s audio capabilities to audibly communicate its messages.
While Dr. Sbaitso may not have had a significant impact on the development of modern chatbot technology, it holds nostalgic value (the 90s have this effect on many people) for many computer users who encountered it during the early days of personal computing.
The sound is painful to hear given our current expectations but I reckon that this was phenomenal at the time.
From a technology standpoint, Dr. Sbaitso used the same rule-based approach. In the below GitHub page, you can actually see a JSON file that holds the keywords and respective possible answers.
https://github.com/danreeves/sbaitso/blob/master/strings.json
It’s an implementation aimed at bringing Dr. Sbaitso to Twitter.
ALICE (1995)
ALICE is an acronym for Artificial Linguistic Internet Computer Entity and is a chatbot that was developed in the late 1990s by Richard Wallace. Its main goal was to simulate natural language conversation with users.
ALICE follows the line of all its antecessors and also uses a pattern-matching rule-based technique. It analyzes the words and phrases that users input and searches for specific keywords or patterns in their messages. Based on these patterns, ALICE generates predefined responses that are programmed into its system.
The chatbot was designed to engage in text-based conversations and provide answers or responses to user queries or statements. ALICE’s responses were primarily based on the patterns it recognized in user inputs. It aimed to create the illusion of understanding and meaningful conversation.
While ALICE gained popularity and was one of the most well-known chatbots of its time, it had limitations. It lacked the ability to truly understand the context or meaning behind user messages, and its responses were limited to the pre-programmed patterns it recognized.
The above can be said about all of the chatbots that preeeded ALICE
ALICE played a significant role in inspiring further developments in the field of chatbot technology. It demonstrated the potential of using pattern-matching techniques for basic conversation and laid the foundation for more advanced chatbots that incorporate machine learning and natural language processing to achieve a higher level of understanding and interaction with users.
SmarterChild (2001)
SmarterChild was developed by ActiveBuddy in the early 2000s, SmarterChild was a chatbot primarily used on instant messaging platforms like AOL Instant Messenger and MSN Messenger. It provided various functionalities, including weather updates, news, trivia, and general conversation. SmarterChild gained popularity and was widely used by millions of users.
Users could engage in text-based conversations with SmarterChild, asking questions or requesting information on various topics. SmarterChild provided responses based on its programmed knowledge base and algorithms.
One of the key aspects of SmarterChild was its versatility (finally a chatbot that steps over the doctor-patient interaction π). It offered a range of capabilities, including weather updates, news headlines, sports scores, stock market information, and even trivia games. Users could ask SmarterChild about specific topics or engage in casual conversation.
SmarterChild became popular among users of instant messaging platforms, particularly teenagers and young adults. It provided a convenient and interactive way to access information and engage in conversation without needing to visit multiple websites or applications.
While SmarterChild had its limitations, such as relying on predefined responses and lacking the ability to truly understand context, it was regarded as an innovative and engaging chatbot during its time. It showcased the potential for chatbots to provide various services and be integrated into existing communication platforms.
SIRI (2010)
The smartphone Era brought the latest big innovation in terms of conversational AI, with Siri leading the way as the pioneer among many others to follow.
When Siri was launched in 2010, it was initially introduced as a standalone app for the iPhone. The Siri app, developed by Siri Inc., was designed to provide voice-controlled features and perform tasks such as making restaurant reservations, setting reminders, sending messages, and more. It utilized a combination of voice recognition, natural language processing, and integration with various online services to accomplish these tasks.
The original Siri app was available on the App Store (this was the time of iPhones 3/4 and iPod touch). However, in 2010, Siri Inc. was acquired by Apple, and Siri’s technology became integrated into the iOS system.
When Siri was first integrated into iOS it gained access to a wider range of system features and became deeply integrated with Apple’s services. This integration allowed Siri to provide information on weather, sports, stocks, and other topics, as well as perform device-specific tasks like setting reminders, making calls, sending texts, and playing music.
While the early version of Siri in 2010 was groundbreaking in terms of its voice-controlled functionality, it was (obviously) not comparable to the more modern version as we all know it. Over the years, Apple has continuously improved Siri’s capabilities by enhancing its natural language understanding, expanding its integration with third-party apps and services, and incorporating more advanced machine learning techniques.
After Siri followed Alexa, Google Now, and Cortana (among others).
ChatGPT (2022)
I was not sure whether to include ChatGPT in this section of history, but it does deserve recognition, beyond doubt. It has had a profound impact on public perception by showcasing the potential of conversational AI at a scale never before seen.
ChatGPT, developed by OpenAI, is an AI chatbot that builds upon OpenAI’s foundational large language models (LLMs). The initial release of OpenAI’s language model was GPT-1 in 2018. However, the model that underlies GPT-3, called GPT, is the core framework. ChatGPT, in particular, is a specialized version of the GPT-3 model, meticulously fine-tuned and optimized specifically for interactive conversations.
Conclusion
The initial three decades of chatbot development predominantly relied on rule-based systems.
Nevertheless, the diversity, ambiguity, and creative nature of human language posed challenges for rule-based and template-based approaches, as they were incapable of covering all the possibilities and exceptions that language presents. Language is too ambiguous to try and code every possible outcome.
However, with the advent of artificial intelligence (AI) and natural language processing (NLP), coupled with the rapid progress of the internet and mobile eras, the stage was set for a remarkable surge in conversational AI capabilities.
References
- [1] https://en.wikipedia.org/wiki/Turing_test
- [2] https://en.wikipedia.org/wiki/ELIZA
- [3] https://scifi.stackexchange.com/questions/252993/what-are-all-the-known-questions-that-have-been-asked-as-part-of-a-voigt-kampff
- [4] https://botwiki.org/bot/smarterchild/
- https://13chats.com/blog/a-brief-history-of-chatbots
- [6] https://en.wikipedia.org/wiki/PARRY
- [7] https://en.wikipedia.org/wiki/Dr._Sbaitso
- [8] https://en.wikipedia.org/wiki/Siri
- [9] https://www.forbes.com/sites/bernardmarr/2023/05/19/a-short-history-of-chatgpt-how-we-got-to-where-we-are-today/?sh=12b56878674f