The Future of Chatbots: Will they break the ‘illusion approach’?

By: Eskender Tamerat

At first, it was all about creating illusions. Asking questions endlessly was the golden trick back when a computer parody by the name of Eliza kick-started the era of computers conversing with human beings in the 1960s. With a restricted set of scripted rules, the bot had no clues to grasp the user input, let alone being a good friend of a human.

The next few decades saw the rise of a meaning-based human text interaction.

It reduced the trouble of indulging in a heartfelt communication for mere machines responding to a set of choices waiting for human instructions. This facet evolved to what we could see in the modern video games, in which a user gets visual feedbacks by controlling an avatar using a mouse or joystick.

Returning to the topic on hand, more than half a century later, illusion still fits the bill to describe the current state of chatbots. The domain of knowledge base showed rapid progress – with all the data out there via internet, sentence parsers using natural language processing, and changes in hardware causing machines to be faster and massive in storage – but we are still lagging behind in achieving the grand goal of simulating intelligent conversation between a machine and a human.

Instantiated by the Turing Test, the Loebner competition takes place annually with four finalists battling for the big prize. Unlike the strict requirements of fooling the judges set by Alan Turing all those years ago, the judges will look for the “most human” from the participants, who make it into the final round after dealing with a set of human knowledge questions in the qualifiers.

Bruce Wilcox leads the team who used Chatscript to enter the Loebner prize contest on a regular basis. They developed different chatbots using their language platform, which they decided to open source for fellow chatbot enthusiasts. In addition to winning the Loebner competition four times (2010, 2011, 2014, 2015), they also tasted success in ChatbotBattles in the best 15-minute conversation category.

But the biggest achievement came in 2010 when they were able to fool  the judge in the Loebner competition with the bot named ‘Suzette’. It happens to be the first time they entered the race and they were forced to rewrite the whole script at the eleventh hour following a licensing hullabaloo restricted them from using their previous contractual development works in the competition.

In a paper titled “Suzette, the Most Human Computer”, Bruce revealed his designing methods before concluding with some advice to the judges on the list of things to do and don’t. As a guy who knows the inside out of the chatbot world, Bruce stresses that the task of a chatbot is to create an illusion. Recently Suzette’s feat of causing the judge to believe that it’s a human is repeated in other competitions.

But the main issue is fairly unexplored after all these experiments. It’s one thing to prepare a bot for a limited conversation to be adjudged by few people for whom you could make the chatbot to address the potential queries beforehand. It’d be a different proposition to develop a chatbot that could converse (or, at least, give a sensible response) in any raised topics for something resembling a dialogue between two humans.

As such, Bruce’s team is trying to step up their effort in making the engine automatically write scripts by understanding enormous texts, like from Wikipedia contents. Since authoring the chatbot by hand– albeit with the help of different functionalities of the Chatscript platform to ease and facilitate the process – will not be the ideal move, the making of the machine learning from its experience and labeling a set of information by itself will be invaluable.

Apart from the general purpose conversation bots, there are also efforts on building responsive chatbots in specific areas of expertise. Part of its commercial projects, health-based and customer service chatbots are under development here at iCog Labs. With all the limitations of the most advanced products in the area, there are a number of reasons to contemplate on what could be achieved in the near future, but there is little doubt as to chatbot being one of the frontiers in the field of artificial intelligence.


From the editors of

This Article was originally written for the bi annual magazine (iCog Makers) by Eskender Tamerat and was published on hard copy in collaboration with the Federal Democratic Republic of Ethiopia’s Ministry of Science and Technology on 5, August, 2017. It is republished here with the permission of the iCog Makers magazine editor and the verbal consent of the Ministry. 

Social tagging:

2 Responses to The Future of Chatbots: Will they break the ‘illusion approach’?

  1. Bruce Wilcox says:

    Bruce Wilcox has won the loebners 4x, not 2x

  2. Eskender Tamrat says:

    Thanks for sending the correction. It will be edited accordingly. I am glad you read the piece. Thanks.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: