During its recent F8 conference, Facebook announced that it’s opening Messenger in beta to allow business chatbots into the app. What it means, is that people are now able to interact with AI-powered company representatives within Messenger, although the number of available bots may be limited, depending on your location. I feel like they’re an important and interesting issue to discuss, and constitute another step in the human-machine interaction process. The following post presents brief history of chatbots, as well as discusses their present state and future. Take a look at what to expect from the latest developments in NLP systems for business.
Facebook backs the move with some solid numbers: over 50 million companies operate within its social network, and more than 1 billion business messages are sent every month.
The new guests to the Messenger party will send and receive text, images, buttons, bubbles and calls to action, combining conversation with actual user interface. Chatbots will engage customers (more than half of the 1.59 billion active monthly users are going mobile) via channel that these people are comfortable in.
That’s about it, in a nutshell, when it comes to new things introduced for Messenger. Chatbots, or some forms thereof, have been around for some time though, and will grow way beyond Facebook’s instant messaging platform. With that being said, let’s have a look at the broader picture of what we’re in for.
What are chatbots and where do they come from?
In short, chatbots are computer programs designed to hold conversation using textual or auditory means. They’re built to resemble human behavior in a convincing manner. Most often, you’ll find chatbots used for customer service or information acquisition purposes. Natural language processing systems are employed for some of the most advanced chatbots, but many of the simpler ones scan for keywords and pull a reply from the database that best matches the input.
This is precisely how the first prototype of a chatterbot – ELIZA (1966) – operated. It was designed to create an illusion of understanding and steer the conversation forward in an apparently meaningful way. The emphasis was on vagueness and unclarity, providing replies that were considered intelligent in a wide variety of contexts.
ELIZA proved that it’s fairly easy to achieve this. Human judges asked to assess whether they were talking to another individual or a machine were quick to give the benefit of the doubt when the replies could be considered as somewhat intelligent.
Human tendency to interpret the digital output in much the same way as natural conversation opened up the gates for multiple applications of these rather simple systems. From information elicitation, toys, mobile apps, customer support, to any other online help systems. Chatbots have now grown into more complex NLP systems that can perform a wider range of tasks and act much like your personal assistants.
Chatbots have arrived
When the Internet first became popular, everyone wanted a website. Then it was apps, and now it’s a new era – the era of business bots. What they’ve done so far to surpass more traditional applications is they’ve offered more interactivity and human-like contact.
I’m pretty sure that if you’re reading this, there’s a great chance you’ve interacted with a chatbot before and weren’t too impressed with what it can do. I agree, me neither. But when you compare some of the original sites and apps with what we have today, they don’t stack up well against each other.
The problem is that today’s chatbots still aren’t very good with understanding natural language, be it written or spoken, which may translate into poor user experience. In fact, Facebook M, the personal assistant competing against Cortana and Siri, has actual people handling some of the more complicated requests.
What this probably means, is that full automation of certain aspects of interaction won’t be possible anytime soon. Still, I reckon we shouldn’t be too quick to give the final verdict on chatbots.
These programs hold a promise of letting you interact with businesses you care about in much the same way you do with friends and family. The intention is to help you do more and quicker than when using multiple apps.
I think we shouldn’t expect too much of chatbots just yet. When you think of it, the basic AI that’s behind them already is a pretty amazing technology. With advancements in deep learning (analyzing huge amounts of digital data), we will probably see the emergence of bots that will make contacting businesses from a platform like Messenger seamless.
Where are chatbots headed to now?
What chatbots are sure to continue having an impact on, moving into the future, is consumer-business communication, as well as the overall human-machine interaction.
The ultimate goal for many businesslike chatbots is to become personal assistants that serve as entry points to the Internet. Some of the more optimistic forecasts mention a time frame of about five years for us to see a significant shift in how we use bots.
What works in favor of these systems’ growth is that some of them are able to learn new things based on the interactions they’re having with humans, instead of just dully pulling replies out of a static database.
There’s a number of issues that speak for advancement of chatbots:
– Current customer support often requires the client to operate across a number of platforms simultaneously – a combination of online search, talking to a human, navigating menus, verifications, navigating websites, etc., whereas chatbot experience generally entails typing a simple command to get a rich media answer.
– Chatbots offer a standardized way of reaching a business. Client verification takes place through Facebook, no phone transfers are necessary. AI-based help systems are a modern way to present rich information – media, links, etc., in contrast to messages transmitted using only voice.
– A chatbot creates an edge for a businesses. It becomes expected by the customer, and also, it’s cost-cutting – trimming down on call centers related spendings.
– And maybe most importantly, chatbots will provide huge data sets to improve natural language processing systems. It’ll work in a way similar to Google searches, where users don’t even think about it anymore, how search results have really improved over the years.
At this very moment, the best thing we can hope for are bots specialized in certain types of conversations or tasks, for instance, booking plane tickets.
As their users, we shouldn’t expect very sophisticated conversations, but rather tools that facilitate certain tasks and make them as painless as possible. Essentially, chatbots are supposed to create a simpler way to interact with businesses.
It’s no wonder that chatbots are becoming increasingly popular in business environments, serving as points of first contact instead of call centers.
The idea of bots is great, but for now, manned live chat will still beat them in terms of pure user experience – a thing to remember especially for smaller and medium businesses. Plus, group live chat also has other applications, like facilitating online community growth.
Some of the more complex client support interactions are best handled by real humans, and excellent customer service is always sought after. Chatbots may prove to be frustrating and create negative experience. With that being said, customers are becoming increasingly comfortable with such interfaces – the nature of human-machine interaction evolves.
I’ve tried toying with one of the new Messenger bots and found Poncho – a weather forecast bot that should be available to you regardless of your location. It proved to be pretty limited in terms of conversation, but still neat, and it gave me the forecast. Much as I expected.
If you want to try it for yourself, just type Poncho in Messenger’s search bar and scroll down to where chatbots are listed. There you go! Feel free to let me know about the experience you had with this, or any other bots in the comments, or tweet me @Chatwee