It seems no matter where you look these days, someone is talking about chatbots.
And if you listen for more than a moment, you'll hear the great chatbot lie - and you might not even notice it.
What is this "great" lie, you ask?
It's simply this: "There are two types of chatbots - rules-based and AI-based."
And that's the lie - there's only one type of chatbot because THEY ARE ALL RULE BASED - even the AI ones!
The Dream of AI
Back when I started programming (circa the late 1970s), AI was a dream, and the idea of a computer talking with a human (and them not realizing they were talking to a computer) was the Turing Test's goal.
Today we have bots that pass that test; many of us have conversational bots in our homes, on our phones, watches and even our entertainment systems.
These bots are classified as conversational AI bots because they understand voice commands and speak to us in return.
But just because they talk doesn't make them any smarter than a bot that does not.
We assume technology is smarter than it is when it displays human behaviours and actions - even if it's only copying us!
A Bad Bot Example
But rather than argue what is and is not intelligence, let's take a look at a very public AI bot and how it fared.
Tay was an AI bot developed by Microsoft. You'd be forgiven if you assumed Microsoft knew a thing or two about programming because they do.
But in the case of AI, things aren't quite that simple.
Tay was "let loose" on social media to learn, and within 16 hours of launch, Microsoft had to shut it down because of the offensive and derogatory remarks it posted online. Tay went from pleasant to punisher in less than a day!
A Bot's Guiding Principle
At its core, an AI bot uses rules to learn the intent of the conversation. It attempts to parse the phrase and discern the meaning. Then in return, it tries to create a response that aligns with the query.
Human language is incredibly challenging to encode because of the nuances we take for granted (and learned over years of trial and error - often embarrassingly so.)
Suffice to say, a bot that learns needs parameters and guidelines to keep the conversation on track.
What's another name for parameters and guidelines?
One Rule To Rule Them All?
Even AI follows this simple programming construct:
"IF something THEN something ELSE something."
This simple If-Then-Else construct, when layered repeatedly, creates sophisticated branching logic that, to the untrained eye, appears "intelligent."
It's not unlike the professional Mentalist who can “read minds.” They are actually using clever language and a series of other tricks to fool you into believing they have super-human abilities.
The Illusion of Intelligence
Remember, any sufficiently advanced technology appears magical, and it's not hard to create extremely complex branching logic with today's bot programming tools.
With sufficient complexity, a bot can engage in conversations and appear intelligent.
But push the bot beyond the boundary of its known rules, and you're back to default replies - no matter how clever, cute, or apparently informed the bot is.
The Evolution of Bots
Today there are many bot-building platforms to choose. Some provide a platform to create simple bots that mimic basic business functions and services.
Other platforms let you create more complex bots capable of responding to dynamic user input (whether typed or spoken.)
In the end, the success of a bot is directly linked to the skill and experience of the bot developer and their ability to understand the conversation domain and the client requirements.
(1) There is only one type of bot because they're all rules-based!
(2) Anyone trying to con-vince you there are two types (and one is better than the other) is not being straight with you! They’re probably a rookie bot develop too, so be warned.
(3) AI is not a silver bullet. It's more like your crazy relative (who’s fun at parties) but you’d never let run your business or look after your kids!