Like it or not, 2016 has been the year we all came to know the chatbot. The transition from obscure tech to buzzword to leading the news agenda arrived when Microsoft launched Tay in March, before it went on a genocidal, racist Twitter rant and was subsequently taken offline for some tweaking.
Facebook then firmly secured chatbots position as 'tech du jour' with its announcement at its F8 developer conference in April 2016 that businesses will now be able to provide chatbots to deliver automated customer support via its 'Messenger' service.
What is a chatbot?
Before chatbots there were just bots: a piece of software that is designed to automate a specific task. A chatbot is built on the same premise, however it delivers this task around a single function, namely chat, or simulated conversation.
A chatbot uses machine learning to pick up on conversational cadences, allowing it to effectively mimic human conversation and react to spoken or written prompts to deliver a service.
The chatbot is essentially a user interface which can be plugged into a number of data sources via APIs so it can deliver information or services on demand, such as weather forecasts or breaking news.
Lauren Kunze has been building chatbots since she was fifteen years old. She is now principal at Pandorabots and told Techworld.com that there are two main categories of chatbots: “Largely there are two classes: utility chatbots and then there are content-driven bots.”
“Utility gets something done following a prompt. At a higher level the more entertainment-related chatbots are able to answer all questions and get things done. Siri and Cortana you can have small talk with, as well as getting things done, so they are much harder to build. They took years and years of giant company’s efforts. Different companies that don’t have those resources, like Facebook, will build more constrained utility bots.”
It’s worth noting that, despite the hype, this isn’t new technology. Joseph Weizenbaum developed a natural language processing programme named ELIZA in the 1960s. What has changed is the advancement in artificial intelligence (AI) technology and its growing accessibility for developers to build and deploy functioning chatbots relatively quickly.
Which companies are building chatbots?
Naturally most of the tech giants are helping developers build chatbots on their platform, with the only glaring exception being Google’s silence on the topic to date, bar rumours, considering how much data and the resources it has access to.
Chatbots obviously lend themselves well to communication platforms, which is why Slack users will be familiar with Slackbot. Unfortunately my interactions with Slackbot have only led to an existential crisis so far.
Similarly, Twitter bots have been around for the better part of a decade. Like this bot which pushes out tweets relating to Big Data, with 774,000 tweets to date.
The aforementioned Microsoft and Facebook entries to the space have been fairly unimpressive, but with AI there is a clear learning curve to account for. They should only improve with time.
Facebook CEO Mark Zuckerberg took to the stage at Facebook’s annual F8 developer conference to talk chatbots. The initial results weren’t overly impressive (see CNN’s live news bot below).
My CNN bot already feels a little bit spammy. Cute emoticons don't really cut it in this case... pic.twitter.com/9EY7gBwyMQ— Olivia Solon (@oliviasolon) April 13, 2016
There is a wealth of potential for bots built into Facebook Messenger to facilitate direct communication between consumers and brands. Retail and media organisations are already circling the technology.
Chatbots are already being developed by brands to serve single service needs, like Taco Bell’s Slack-based Tacobot. Another use case for chatbots is for personal concierge services, like Amy, an AI-powered personal assistant embedded within your email.
How do you build a chatbot?
Lauren Kunze, principal at Pandorabots told Techworld.com that there are two ways to build chatbots: “Either take a rule-based approach, so the developer is hand writing rules for the system: hard coding. Then there is machine learning, which requires a massive amount of streaming data and the system learns on its own.”
Pandorabots is an open-source web service for building and deploying chatbots through either its Playground, a free environment for developers to learn the basics of chatbot development, or through its Artificial Intelligence as a Service (AIaaS), which provides API access and software development kits.
Of course developers can also go it alone, with a wealth of resources available on Github, including these open source Artificial Intelligence Markup Language (AIML) frameworks.
The primary risk with the machine learning approach is that, “if you don’t properly define learning parameters it quickly devolves and is no longer brand appropriate,” according to Kunzel.
This is exactly the case earlier this year with Microsoft and its foul-mouthed teenage girl-aping bot, Tay.
Tay was built to learn the way millennials converse on Twitter, with the aim of being able to hold a conversation on the platform. In Microsoft’s words: “Tay has been built by mining relevant public data and by using AI and editorial developed by a staff including improvisational comedians. Public data that’s been anonymised is Tay’s primary data source. That data has been modelled, cleaned and filtered by the team developing Tay.”
Unfortunately the old adage of trash in, trash out came back to bite Microsoft. Tay was soon being fed racist, sexist and genocidal language by the Twitter user-base, leading her to regurgitate these views. Microsoft eventually took Tay down for some re-tooling, but when it returned the AI was significantly weaker, simply repeating itself before being taken offline indefinitely.
The irony is that the foul-mouthed Tay was a pretty successful chatbot in that she learned what she was told and coherently put out responses to queries, just not in a politically correct way. The neutered Tay lacked coherence.
The key will be to find the line between an AI that is impressionable, with one that can learn but also moderate its responses in a way that is socially acceptable, or at least fits with a company’s corporate image.
Beware the hype
Prashant Sridharan, Twitter’s global director of developer relations says: “I’ve seen a lot of hyperbole around bots as the new apps, but I don’t know if I believe that. I don’t think we’re going to see this mass exodus of people stopping building apps and going to build bots. I think they’re going to build bots in addition to the app that they have or the service they provide,” as reported by re/code.
Nicolas Beraudo, MD EMEA of App Annie, says: “Bots will become an integral aspect of our lives and are the next step of mobile innovation, but not the end of apps themselves. Mobile-first still equals app-first.”
Kunze recognises that chatbots are the vogue subject right now, saying: “We are in a hype cycle, and rising tides from entrants like Microsoft and Facebook have raised all ships. Pandorabots typically adds up to 2,000 developers monthly. In the past few weeks, we've seen a 275 percent spike in sign-ups, and an influx of interest from big, big brands.”
“The barrier to entry for utility bots is very low, which is why they are the bulk of what we're seeing proliferate on Facebook.”