An introduction to Chatbots


Posted on: November 27, 2017 by Eric Monchalin

We can safely say that we are in or at least entering the era of the chatbot. The big players, Google, Facebook, Microsoft and others, are all busy developing and improving this innovative user experience technology. Messaging applications are now more widely used than social networks, which is good news for the chatbot as they are such messaging applications!

Chatbots are coming!

So what exactly is a chatbot?

The aim of a chatbot is to conduct conversation which allows people access to information via a lightweight messaging application. There are currently two distinct types of chatbot:

  • Rule-based chatbot: These essentially work as an interactive FAQ. They’re programmed to recognize certain terms and patterns from which they can respond with pre-set answers.
  • AI chatbot: These act as an artificial brain, using sophisticated cognitive and natural language processing capabilities. It not only understands requests but also context, intent, emotion and it continuously gets smarter as it learns from conversations it has with users.

Although the first chatbot, Eliza, was created in 1966 by the Massachusetts Institute of Technology to simulate a psychotherapist, their emergence in our day-to-day lives has only been in the last two years or so. They are swiftly becoming more sophisticated but they are nowhere near having reached their full potential.

Even if the most successful give an illusion of simplicity on the front end, there are a lot of challenges to solve to create a seamless customer experience: analytics, flow optimization, error checking, integration to APIs, routing and escalation to live human conversation.

The rules for a good rule-based chatbot

A good rule-based chatbot requires a thorough input of possible patterns and responses. The more sophisticated versions use a knowledge base to get a list of potential responses and then score them to choose the most appropriate one.

This type of chatbot lacks of adaptability and analytics capabilities of a human being to well understand a person’s expectations and intentions. A rule-based chatbot will never be able to compete with his AI cousin.

AI chatbots

Of AI chatbots, there is still much to learn. We are seeing the first in the market from the really big players: Alexa, Google home and Siri are fantastic examples.

However, the obstacles to overcome in their creation are still very much apparent:

  • They have to learn: For them to be able to learn, they have to be used and interacted with. A key difficulty is making it attractive enough for a consumer to dedicate their energy into interacting with, and therefore teaching, one of these chatbots.
  • They will learn what’s being taught: They don’t have a moral compass or understand what’s decent so they need good teachers to be “good”. An example of poor teaching made headlines recently: “Microsoft deletes 'teen girl' AI after it became a Hitler-loving sex robot within 24 hours”.
  • Some things are hard to teach to a machine, for example, learning to understand emotion. This is going to be the hardest challenge of all but is key to developing a great chatbot.

None of the above is insurmountable. With the heavy weights of the industry focused on this task we will see leaps in chatbot evolution over the next few years.

The Atos IT challenge

This is why chatbots are this year’s Atos IT Challenge and primarily we want to see a fantastic idea brought to life with flawless customer experience.

The key to a great chatbot is to ensure your focus is on the user experience. Focus too much on the technology and you can lose this.

One thing is certain: if you master the chatbot, you master the world!

More about the Atos IT Challenge competition and its 2018 edition >>

Share this blog article


About Eric Monchalin

Vice-President, Head of Machine Intelligence at Atos and member of the Scientific Community
Eric is Vice-President, Head of Machine Intelligence at Atos. He’s responsible for solidifying new technological and business directions for Big Data Global Business Line. Eric’s career has been mainly built on numerous R&D positions in several companies, with experience in leading 100+ people organizations and managing large projects in international environments. Eric’s best memory is the first Supercomputer Bull delivered CEA end 2005 and which was ranked number 5 worldwide. He was in charge of this multi tens millions Euros project on behalf of Bull R&D: technical presale, design, development and on site bring up. Eric is a technology-minded person who values wide range of skills and technological knowledge focused on customer wishes to turn them into reality.