Why chatbots are the talk of the town

Long ago, if you required assistance, service, or information from a business or organization, it was necessary to call them on something called a “telephone.” If you can imagine, these devices worked by pressing a certain sequence of numeric symbols, then connecting you to a human on the other line. Of course, you will most likely be required to remain on hold for ages listening to some muzak that will inadvertently be too loud. Your only respite during all of this will be the frequent reminders that “Your call is important to us, please stay on the line.” Thankfully, these days we have a much more elegant solution — chatbots.

A brief history lesson

Surprisingly, the very first “chatbot” was developed even before the first personal computer. Developed by MIT Artificial Intelligence Laboratory scientist Joseph Weizenbaum in 1966, this technology was affectionately named “Eliza,” and worked by examining keywords which were input, and giving an output based on a defined set of rules. After that came “PARRY” in 1972, developed by Stanford University psychiatrist Kenneth Colby, which was described as “Eliza with attitude.”

The next breakthrough in chatbot technology came in 1995, with Richard Wallace’s “Artificial Linguistic Internet Computer Entity,” or “A.L.I.C.E.” for short. By this time, there existed several methods to quantify the efficacy of chatbots, or in actuality, their ability to mimic genuine human interactions. Among these were the Loebner test, which A.L.I.C.E. won three times, and the more well-known Turing test, which this tech could not pass. The code for A.L.I.C.E. is available as open source, and can be found here from the ALICE A.I. Foundation on Google Code if you’re interested.

To bring us into the modern age, we witnessed the next evolution of chatbots in the form of virtual assistants, such as Apple’s “Siri,” Android’s “Google Assistant,” and Microsoft’s “Cortana.” Expanding on this, smart speakers such as Amazon’s “Alexa” and the “Google Home” were also developed. But enough history — what’s going on in the world of chatbots today?

All chatbots are not created equally

Broadly speaking, there are two main types of chatbots. The first is a “rule-based” chatbot. These function very much in the same way as “PARRY” all those years ago, by relying on a list of pre-written keywords that they are designed to first understand, and subsequently give feedback to. These don’t require lots of training (making implementation faster), the relatively simple technology usually makes them cheaper to launch, and the pre-defined method of communication lets you better control the chatbot’s responses. Conversely, they may not catch typos or misspellings of words, their responses can seem obviously “robotic”, and they’re unable to learn on their own (thus requiring manual improvements).

The other type is an “AI-based” chatbot, which is driven by “Natural Language Processing” (NLP). For those unaware, NLP is a subfield of AI which refers to the ability of a computer to understand human language — both spoken and written. These can better catch grammatical mistakes and typos, they are able to constantly improve, and they offer a more natural-seeming conversation to the client or customer. On the other hand, their learning process demands a more drawn-out implementation period, the lack of a pre-defined structure can make conversations with them unpredictable, and they take some time to “un-learn” the things they are incorrectly taught through interaction.

What are some ways to implement chatbots?

One example of a sector where chatbots are being implemented in a massive way are contact centers, or customer service centers. These chatbots more or less have three main functions. The first is to have a sufficiently-large database of answers to informational requests, such as “What time does your office close?”. The next is to understand and comply with transactional requests, such as “I want to block my credit card”. The third and perhaps most vital function is to simply understand when a question is too complicated, and then pass the conversation on to a human, hopefully along with all of the relevant information the chatbot has already received from the client or customer.

Another field that is benefiting from chatbots is news, or rather the delivery of relevant information. Perhaps the most common examples of these are Telegram Bots. By setting up this chatbot within the messaging service Telegram, your company or organization is able to provide notifications and news, integrate with other services (Gmail Bot, Wiki bot, etc.), and even offer paid services by acting as a virtual storefront. Facebook Messenger offers a similar service, with CNN’s chatbot being one of the most popular in use today. In essence, this type of chatbot serves as the next iteration of RSS feeds, with remarkably better integration and convenience.

You may be surprised to learn that chatbots are even being implemented in entertainment. Again through Facebook Messenger, there are a number of chatbot-powered games which users can play directly in their Messenger app. These include such titles as “MojiHunt” and “Trivia Blast.” There are even entire video game franchises based around the technology of chatbots, such as the popular series “Scribblenauts.” In this game, you are given the task of “summoning” items by typing them out in order to help you complete certain tasks, such as opening doors, building bridges, and of course fighting dragons. Beyond the three rules (no bad words, no proper nouns, no copyrighted words), you’re left to your own devices on what items are necessary to proceed. The game interprets your words, and causes that item to appear in the game and interact with the characters and environment around it in a logical way (“logic” here being a relative term).

Try typing “Conclusion,” or “I’d like to stop reading now”

In this article, we’ve only just scratched the surface of the tip of the iceberg when it comes to how the technology of chatbots has been, is being, and can in the future be implemented. What it really comes down to is convenience for the client or customer, and cost for the company or organization. In the case of chatbots, they can instantly answer up to 80% of customers’ routine questions, and they’ve been shown to reduce customer service costs by up to 30%. Obviously the technology isn’t perfect, but it’s well on its way to making those “telephone” things we discussed previously a forgotten relic of the past. Just don’t expect a convenient chatbot to be there for you anytime soon next time you try to cancel your airline ticket.