Lab informatiekunde

Publicatieruimte voor studenten informatiekunde (faculteit Letteren, KU Leuven)

Chatbots - Chat to No One and Get Things Done

I once wanted to buy something online, but I wasn't really sure so I fired up the site's online support. I wanted to talk to a person and get some insight in what I was buying. The person helping me in the support chat was really friendly, asking my name, addressing me with my name, making jokes and providing me with the right information. I ended up making the purchase. It wasn't until much later I learned that I never talked to a real person, I talked with an artificial intelligence.

The field of AI is big but one thing I particularly am interested in are chat bots, and how they'll replace common interactions normally performed by people on social media

The Rise of Bots

Automatization is a common practice: we replace people doing work that can be broken down into steps by machines doing the same work, for a lower price and sometimes even faster. The same thing is starting to happen online too. Bots interact with people online to do common tasks. This might sound weird but for a lot of industries this opens up a lot of possibilities.

Research into AI and the construction of languages has improved a lot over the years allowing applications to chat with someone in a natural way, learning what the user wants, parsing the relevant information and at the end of the road even learning how the application can improve itself. These chatbots grow more larger lately, not only because technology has improved but also because people use messenger applications a lot. I've written about LINE before, but even outside of Japan the popularity of applications such as Messenger and WhatsApp is booming. People like to chat, 1-on-1 with someone, and bots can be really helpful with that.

Case Study: Cortana

Cortana is Microsoft's virtual assistant with quite a bit of personality. Aside from cracking jokes and finding you nearby restaurants, it can also track parcels, suggest alternate routes during traffic and pickup important info from your e-mail to add to your calendar. It's a natural-to-use and clever virtual assistant similar to Apple's Siri.

You can order pizza for instance via an online chat system. In a friendly and human way, a bot very much looking like a normal employee will talk you through which pizza you want to order, ask your toppings, give you the total price and more.  As a customer you won’t notice you’re not talking to a real person, unless maybe you start to tell your life story and you'd notice the "employee" you're talking to will always steer the conversation back to whether you want a small, medium or large pizza. All other times it’ll look like you just had a normal quick conversation to order your pizza with a "real" employee, making things feel much more personal.

Lately there’s a big push for artificial intelligence and using natural language with modern technologies. Examples of this include Siri and Cortana interacting with people in a natural way. On the behind-the-scenes side of things technologies become better at searching big data and learning how customers behave and improving itself with machine learning, making the AI become smarter over time.

The field of artificial intelligence is a big one with an exciting future ahead. Self-driving cars are making great progress, robots are improving and becoming more human (even in Japanese hotels!), and facial recognition is becoming a realistic alternative to a password.

No AI is perfect

Case Study: Tay

Tay was Microsoft's attempt at an AI bot with machine learning by learning from what other people told her on Twitter... this didn't end well...

Doesn't all of this have a negative side too? Nobody saw all those old science-fiction movies where AIs go rampant. Nobody saw The Matrix trilogy?

It’s argued that artificial intelligence isn't smart enough to act with a human consciousness during moments where an exception to their rules occur. When a self-driving car can't decelerate fast enough and is bound to hit a person, will it pick the young girl walking on the left, or the older lady on the right? Which one is the moral right choice? No AI is perfect, Tesla learned this the hard way, but at the same time one could argue the AI just does what it was told to do, and the person who programmed it is still at fault.

Unlike what sci-fi movies might suggest, chat bots aren't going to run rampant and destroy the earth. I'm amazed at how they start to chat in a natural way, and just how far their possibilities are. It'll be exciting to see where they head next as the technology behind it improves. Next time when I'll order a pizza online, I'll remember I might've talked to a bot and not a store employee taking my order. Who knows, maybe in 10 years from now we'll have bots or drones deliver the pizza to us.