Artificial intelligence (AI) and learning machines are penetrating the workplace and everyday life an ever-growing rate. An important element of AI is chatbots – robots and software services that perform communication tasks. Let’s take a look at where chatbots come from, how they are implemented and how they need to develop.
Where do the bots chat?
Very generally, chatbots can be implemented wherever there is digital-based communication. This currently spans e-commerce and online banking, web support and new services. On Ebay, for example, a user can interactively flick through the products in dialog with the ShopBot, which includes an integrated image search using posted images to find similar articles. Mastercard provides a bot called Kai to banks and their customers in order to simplify transactions, while Mildred helps Lufthansa customers select their flights.
What’s behind it all?
The majority of today’s chatbots are based on models known as retrieval-based models, meaning a catalog of question-answer routines optimized for a specific purpose. This in itself is not very smart, but it is based on the smartness, the scale and the diligence of writing the dialog script. Put simply, the script is in charge.
When it comes to creating a chatbot there are already many solutions, botkits and open-source libraries which provide ready-made scripts for various target platforms (e.g. Facebook Messenger, Slack or Google Assistant), or even drag-and-drop interfaces.
Who is building chatbots today?
In principle, anybody who gets to grips with the topic can create these scripts and use the odd tool. There are now many agencies and smaller web service providers exploiting the boom for themselves. The target platform is usually Facebook Messenger.
Big players such as Google, Facebook, IBM, Microsoft or Amazon are not emerging as developers; they are providing tools and platforms with which others can develop their own chatbots, for example the Microsoft Bot Framework or Facebook for Developers.
According to a current Oracle study, chatbots will make up most of the customer service sector by 2020, with 80 percent of large brands introducing chatbots as customer advisors.
What do future chatbots need to achieve?
The current generation of chatbots – often still beta versions – marks only the beginning of human-machine communication. We are entitled to expect a lot more from the chatbots of the future. In the near future they will be able to understand us fully and communicate with us much better than existing voice assistants such as Siri and Alexa, which so far only offer answers to simple questions and demands.
It is important to identify future application fields. To work together with the actual providers and users – the companies – to plan, design and develop the next generation chatbots, which will be based more and more on deep learning and other similar technologies.
The requirements and possibilities of the next generation chatbots are also the subject of a talk by *um Senior Data Scientist Ulf Schöneberg at the Chatbot Summit on June 26 in Berlin.
This may also interest you:
The myths and possibilities – Artificial intelligence and learning machines
Enterprise AI – Artificial intelligence in companies
Study about artificial intelligence and machine learning in companies