“Chatbots” is a combination of the word “chat” and the abbreviation for robots, “bots.”
As an emerging technology, there is a lack of common consensus which explains why there are so many different definitions floating around the internet.
A straightforward interpretation for chatbots is, “a computer program that uses Artificial Intelligence (AI) to complete tasks through human-like conversations with users either by text or audio as a medium.”
Furthermore, in the technology industry, there are numerous technical buzzwords like chatbots, virtual assistants and AI. One may easily get overwhelmed, so read this to find out the differences betwee chatbots, virtual assistants and virtual human agents.
An Extension of AI
Chatbots were initially built by developers just for fun.
Early chatbots used a keyword-matching technique to generate an output for a specific input. For example, simple chatbots like werewolfbot operated on the same principles. Keyword functions such as /startgame or /lynch <user> enables werewolfbot to act as a game master for a group-based game within a telegram group chat.
Although chatbots can function without AI, usability is a major problem that undermines the overall user experience. Take this photo for instance:
Most modern-day chatbots have AI incorporated in their programming to eliminate this fundamental usability issue. Although most chatbots today can adequately sieve out the essence of a message, understanding nuances in speech and text is still a mammoth task they have yet to fully overcome.
Understanding Chatbots: Natural Language Understanding
A sub-field of AI, Natural-Language Understanding (NLU) is the engine which powers chatbots’ comprehension. It is also classified as one of the 4 “AI-Complete” problems, the hardest category of AI-related problems according to researchers.
The emergence of NLU in the timeline of AI isn’t new, but voice-operated assistants like Siri and Cortana fuelled the explosion of interest in NLU.
History, Timeline and Milestones of Chatbots
1966 – ELIZA
The first ever chatbot, ELIZA was created by Massachusetts Institute of Technology. She responds in the same way as werewolfbot, replying with a pre-written script when keywords are identified in the input text. One of the first milestones in chatbot history when ELIZA became capable of attempting the Turing Test – developed by the “Grandfather of AI,” Alan Turing. This test determined a machine’s ability to exhibit “intelligent behaviour” equivalent to, or indistinguishable from, that of a human. Subsequently, in an experiment with human test subjects, many were convinced she was human.
A few years later, modifications were made to ELIZA’s original programming script to simulate a psychotherapist. Thus, the new electronic therapist ELIZA was conceived. You can talk to her here.
1972 – PARRY
Another early chatbot synonymous with ELIZA is PARRY, written by psychiatrist Kenneth Colby. In contrast to ELIZA, PARRY was a chatbot simulating a person with paranoid schizophrenia. Likewise, PARRY equally exhibited “intelligent behaviour” from passing the Turing test. But what happens when you put a psychotherapist and a mentally ill patient together?
1995 – ALICE
Developing “Human-like” elements in chatbots was another milestone. This can be seen through ALICE’s (Artificial Linguistic Internet Computer Entity) conversations. She was a revolutionary breakthrough in her time that won the Loebner Prize 3 times – a competition for “the most human-like computer programmes.”
Despite that, she was unable to pass the Turing test unlike her predecessors. However, ALICE will be forever immortalised as the inspiration behind award-winning movie “Her” a story of a man falling in love with a chatbot.
1998 – MegaHAL
MegaHAL was the earliest chatbot with a primitive AI capability, albeit the controversial claims. As the conversation progressed, it was able to learn new words and substituted them into different sentence structures. MegaHAL’s conversations often led to humorous responses as they were generated based on sequential and mathematical relationships rather than conversation rules and norms.
2000s – Siri, Alexa, Google Assistant
Today, the three ubiquitous chatbots are a far cry from MegaHAL. With each use, they improve by learning and adapting themselves to provide a more personalised and customised experience.
Will the next milestone be consciousness in chatbots? It would revolutionize the world yet prove to be a frightening sight to behold when things like this happen.
Where can chatbots shine?
According to the 2018 State of Chatbots Report, online respondents feel that websites are hard to navigate and they couldn’t find answers to their simple questions or queries. In a sleepless online world, 24-hour services aren’t an unreasonable expectation.
Chatbots can be operational 365 days a year and 24 hours a day to conduct smart searches to deliver the necessary information, quickly.
Behind the scenes of every e-commerce is an intricate network of system and process like payment, promotion, logistics, and sales. Online chatbots can help boost revenues by making recommendations, upsell and cross-sell relevant products.
When a customer searches for a new laptop, the friendly chatbot can prompt: “Would you like to opt for an extended warranty?” Followed by another recommendation to cross-sell: “Protect your brand-new laptop with this padded carry case.”
Companionship & Entertainment
Chatbots can do fantastic things like scheduling an appointment or finding solutions to your problems. But they can also hold a conversation and hear about your day. Chatbots are the ideal social companions who’ll never judge your unconventional opinions.
Limitations of chatbots
Chatbots are complex and they don’t come cheap. They can cost up to $340,000 if advanced functions are involved. This amount excludes other operational costs such as maintenance, tech support, infrastructure and downtime.
Can they be too smart for their own good? Microsoft’s AI chatbot experiment, Tay was a flop. If AI learns from interacting with users, how can we stop it from learning from the worst of humanity?
Apart from always passing the buck to chatbots, we can assist them by making slight adjustments in our behaviour. A systematic approach can be used to deal with chatbots by providing appropriate information like message content, location and subject matter.
Existing Chatbots in Singapore
Endorsed by the Singapore Tourism Board, Zumata is a chatbot that conducts smart searches for travel related information. Zumata also increases efficiency by automating and cutting down the typical search time from 45-60mins to mere seconds. It can even generate air-ticket quotations, handle purchases of air tickets and conduct a comparison between travel insurance plans.
The resident bot-in-training at Singapore Management University supports the student’s journey by providing all the queries a student has about school, financial services and more.
Integrated Health Information Systems (IHiS)’s Chatbot
Patients who are unwell could, in the future, be chatting with software that can assess their conditions and advise them to visit the right hospital or clinic. When patients type in basic information such as their symptoms and medical history, the chatbot can even prioritise seriously ill patients and cut their waiting times at hospitals.
The Rising Demand For Chatbots
“By 2020, non-human interactions between business and customers will amount to 85%”
– A bold prediction by Gartner, an objective resource for industry insights.
Organisations are beginning to capitalise on this prediction and swiftly leveraging on the potentials of chatbots. It’s no surprise that in another survey consisting of 800 C-level executives, revealed that 80% have indicated that they already have or plan to use chatbots in their companies by 2020.
Moreover, in 2016, Facebook Messenger unveiled a platform to develop chatbots and in just 2 years, there are now 300,000 active bots on Facebook. That’s more chatbots created in 2 years than in the entire history of chatbots.
The future is now.
How To Build One
Before diving into building a chatbot, one must first consider the fundamental question: Is it necessary? Just because every other company in the industry is doing so, doesn’t equate to a strong case to follow suit.
There is a simple guideline to answer this question:
1) Will it add value
2) Does it have growth potential?
3) Can it help you achieve your goals?
If the answer is “Yes” to all, you can now move on to building one.
Basics of chatbot building
This phase is another round of questions and finding the “right-fit”. Should the chatbot simply entertain or will it provide a suite of support services? Which platform should it operate on – Facebook Messenger, your dedicated website or a stand-alone mobile app? The next step is deciding if the bot is to be built from scratch or to use a list of ready-made platforms.
Some popular bot creating platforms include Chatfuel, Beep Boop, ChattyPeople,Botsify and Smooch.io.
- Minimum Viable Product
Before launching the actual chatbot to the world, the chatbot should contain the essential features discussed upon in the pre-development stage. The chatbot should be exposed to a small, controlled group of end-users for testing and assessment. Feedback should be gathered at this stage to improve the system as it is relatively cheaper as compared to when the bot is released.
Don’t forget to double-check everything and get opinions from the pilot study group. The final chatbot is now ready to be unveiled to the world. Chatbots must be trained continually to enable them to maximise their potential. They must be able to ascertain the right learning points, lest they turn up like Microsoft’s Tay.
Global Trends For Chatbots
Google Trends has shown that a growing interest level in the search term “chatbot,” which peaked during august 2017. Not to mention, Singapore held first place for the highest interest by region followed by other Asia Pacific nations like China, Hong Kong and India.
As for a potential investor’s perspective on the chatbot industry, the market size is expected to exceed USD 1.34 billion by 2024. Although North America held 57% of the global chatbot market shares, Asia Pacific is promising region that is growing fast and catching up. The E-commerce market which amounts to 35%, is the largest in the end-use category. Whereas for application wise, customer service tops the sector with a 43% global market share.
Chatbots are unstoppable. There are many compelling reasons and advantages to integrate chatbots into organisations. Even though there are limitations to chatbots, there’s no denying that chatbots are the future. Chatbots are also a visual manifestation of AI which people can easily interact with.
Now that you’ve gained an invaluable insight and knowledge of chatbots after reading this article, what will you do today with this advantage?