• delaware

How to never feel lonely again

People don’t like to be alone, but there is not always a suitable companion available to have a conversation with. And if no one’s around, would you just start to talk to your computer? The concept of chatbots in the context of artificial intelligence has been around for over 50 years and is being explored more every day.

From Eliza to Alexa

The concept of chatbots already originates from the fifties, when Alan Turing developed his Turing test, which tests a machine’s ability to exhibit intelligent behaviour equivalent to that of a human. Created in 1966, Weizenbaum’s ELIZA is supposed to be the first program which appeared to pass this Turing test. Numerous chatbots were developed over the years. Do you remember having funny, meaningless conversations with Cleverbot in the late 2000s?

Due to the recent progress in AI, chatbots have gained many new abilities and popularity. Chatbots and artificial intelligence are becoming increasingly present in everyday life. Today we have Amazon’s Alexa, Apple’s Siri and Microsoft’s Cortana who are glad to assist you with anything you throw at them, from playing music and cracking jokes to informing you about traffic and creating shopping lists. In this video, we see an example of how Google’s Assistant can actually ring up a salon or restaurant to make an appointment for you. Businesses can use chatbots to start off customer support via messaging apps, among many other applications.

The gap between conversations with machines and actual humans is becoming increasingly thin as time goes on. Eventually, artificial intelligence might help us combat loneliness by becoming a dear companion.

Ask your bot for rooster pics

After a talk about AI and ethics on Tuesday morning (read about this here), we got the chance to implement our own simple chat bot, which was capable of searching for pictures given some key words. For example, when telling the bot to “search for a picture of a rooster”, it will search its database for pictures matching your query and display them all. Depending on the words and phrasing used, the chatbot would distinguish between greetings, searching for pictures, ordering and printing.

We accomplished this by using Azure Bot Services with Azure Search and Language Understanding Service, or LUIS for short. In advance, Cognitive Services were used to obtain a caption and some tags about the images. When training this model, you supply some data of the concepts you want it to understand, which results in a trained model for interaction. These captions and tags were stored in stored in a Cosmos DB. During the actual workshop, our goal was to find a picture of a man enjoying a glass of German beer by querying the basic chatbot we developed step by step in just under 3 hours.

With this workshop, we got an insight about how artificial intelligence can be implemented in everyday life using a simple base case.

By Femke Bruckmann and Jasper Van den Bossche, 1st Master Computer Science