Chatbots were not born overnight
Personal voice assistants weren’t always as smart as you thought they would be.
“Alexa, play my top song from 2019.”
In a few seconds, Alexa, Amazon’s smart speaker, will be playing that tune you were obsessed with four years ago.
This is not the only task Alexa can do for you. Aside from asking it to play a song, you can also request for the latest information on traffic, weather and local news, as well as a rundown of messages and emails sent to your phone.
Smart speakers have become a necessity for numerous homes around the world. With the rising popularity of the Internet of Things (IoT), artificial intelligence (AI) went one step further to make technology seamlessly blend in with society.
Are these smart speakers just a stroke of genius?
Big Tech might like us to believe that these technologies were created out of thin air—but that’s not the case. Alexa, Siri and other smart speakers are just chatbots that come in the form of a voice assistant.
Sadly, these smart speakers have fallen behind since the release of new AI technology in the block: ChatGPT.
But what makes these smart speakers different?
To better answer that question, we must look at the brief history of chatbots.
Like many AI technologies, chatbots were born out of people’s determination to build machines that resemble human intelligence. This interest can be traced back to the 1950s with mathematician and computer scientist Alan Turing, whose legacy lives on in the form of the “Turing Test.”
Almost two decades after its debut, chatbots went through a series of milestones:
1966: ELIZA, the first chatbot
Its primary goal was to simulate a psychotherapist’s operation by asking the user a specific set of questions. Back then, ELIZA’s communication skills were limited as it was unable to withstand long conversations nor retrieve a wide range of knowledge.
1972: PARRY, the schizophrenia-like chatbot
One of the biggest challenges chatbot developers need to address is mimicking human emotions.
PARRY was introduced as the chatbot that resembled a schizophrenic patient. It was supposed to display “personality” through the way it stutters. Five psychiatrists were given the challenge to see whether they can differentiate PARRY from a real schizophrenic patient. Despite the inconsistency of their answers, some of them seemed to believe that PARRY was an actual schizophrenic patient.
Aside from its low response speed, PARRY fails to learn anything from the conversation.
1988: Introducing AI in chatbots
Slowly but surely, the chatbots in this period were able to retrieve information from previous discussions.
Jabberwacky, for example, was written in CleverScript. This language lets the chatbot use contextual pattern-matching to retrieve information from a previous discussion.
What is the pattern-matching approach? This technique requires the chatbot to match the user input with a predefined answer using a rule pattern. Their knowledge had been limited as developers—and not the software themselves—oversaw the creation of new answers.
1995: ALICE; the first online chatbot inspired by ELIZA
While ALICE still retrieved information using pattern-matching, it elevated people’s expectations of chatbots.
Unlike the previous chatbots, ALICE was developed using a new language called Artificial Intelligence Markup Language (AIML). It allowed ALICE to access a wider range of patterns (41,000 templates and related patterns for ALICE vs. 200 keywords and rules for ELIZA).
Over time, more chatbots were developed to simplify people’s daily tasks (i.e., smart speakers) or solve our problems online. Instead of using the pattern-matching approach, the chatbots we encounter now operate with AI via Natural Language Understanding (NLU), a subcategory of Natural Language Processing (NLP).
However, it’s important to remember that not all chatbots function the same way, nor can they fully replace human beings in the workforce.
As mentioned in the New York Times, the likes of Alexa, Siri and Google Assistant, are what we now call command-and-control systems. While they can amass more knowledge by using the internet, their functionalities do not go beyond the simple tasks one would like a virtual personal assistant would do for their home.
As you can see, smart chatbots were not born overnight. It took us more than 70 years for AI to reach the likes of ChatGPT’s level.
Reference:
1 Adamopoulou, E., & Moussiades, L. (2020). Chatbots: History, technology, and applications. Machine Learning with Applications, 2. | Link to full article
2 Weise, K. (2023, March 15). How Siri, Alexa and Google Assistant Lost the A.I. Race. The New York Times. | Link to full article
