Battle of the chatbots

Silicon Valley giants including Apple, Google, Microsoft, Facebook and IBM share a vision in which AI is at home both driving smartphone assistants like Siri or chatbots like those revealed at Facebook’s F8 developer conference.

It’s not yet perfect — only weeks ago Microsoft suffered a little embarrassment when it introduced its Tay chatbot. This was supposed to chat with younger people across multiple social networks, but within hours of its release pranksters had undermined the AI so it began spouting racist rhetoric and Tay was switched off for “adjustment”.

As the technology behind chatbots evolves such errors should become more constrained and in future Microsoft CEO Satya Nadella wants developers to create virtual assistants and intelligent chatbots for a range of tasks, from calendar management to booking hotel reservations and beyond.

But the technology isn’t there yet. In most cases bots haven’t passed the Turing Test. Facebook’s limited availability Facebook M chatbot could deliver only limited responses, so it’s thought they used both robotic and human operators. However, technologies are evolving — natural language processing (NLP) is rapidly becoming accurate.

F8 saw Facebook introduce technologies with which to develop robotic customer service reps for news, weather and retail. Accessed using Facebook’s Messenger product, these bots link into Facebook’s deep trove of personal knowledge with HP, Staples, CNN and Expedia among the first to offer these solutions, but these more limited machines are just a taste of what’s to come.

“We think you should be able to text message a business like you would a friend, and get a quick response,” Facebook co-founder and chief Mark Zuckerberg said.

Chatbots are already popular in China, where a Microsoft system provides services for messaging app WeChat. The popularity of messaging services like WeChat, Viber and Telegram is part of the environment that enables these solutions.

"Bots are conversational so they are a natural extension of how we like to communicate and what we like to do," Forrester analyst, Julie Ask, told Computerworld. Comparing bots to an assistant, she notes: “You can chat with the bot, ask the bot to do things for you, like order take-out or get a new lipstick."

Another trend is internationalisation. When a French person calls a customer support line in the US they usually expect to have to speak in US English, and if they can’t they have limited opportunity to get the help they need. Intelligent chatbots will respond in the caller’s own language, reaching out to skilled internal staff as and when specific requests are made.

It’s a competitive space. As well as Facebook, Microsoft, Apple and IBM, there are many smaller players with skin in the game: Pandorasbots, eGain, SecondEgo and other players are all working hard to develop useful bots.

Stakeholders are focused on figuring out how to deploy such automation across multiple industries, not just in customer service: enterprise communications, medical, public information and beyond. For example, Workato works with Slack to enable access to key enterprise applications (such as your Salesforce CRM records or Zendesk customer support data) from inside Slack.

Proliferation poses its own challenge: When every company has its own bot system vying for attention, how will the buyer decide which bot is best to use? This will generate a fresh opportunity for yet another layer of intelligence to moderate all those relevant bots in order to ensure consumers get exactly what they need, but there is the danger consumers will end up needing to deal with multiple bots in order to handle simple tasks.

We have come a long way since 1966, when MIT’s Joseph Weizenbaum created the Eliza chatbot. One of the first examples of natural language processing, Eliza delighted thousands with its imitation of a psychotherapist in text-based conversation. The difference between Eliza and the chatbots of today is that the solutions now in development have foundations in neural networks and deep learning systems as evidenced by IBM’s Watson or DeepMind’s Alpha Go AI; with the intelligence boosted by access to big data mining technologies. Future generations of intelligent bot will be able to draw insights based on real-time analysis of huge and disparate data sets incomprehensible to human minds.

This shift to context is key. Google CEO Sundar Pichai believes chatbots herald a huge evolution in the nature of computing: “Over time, the computer itself - whatever its form factor - will be an intelligent assistant helping you through your day," he told Computerworld. "We will move from a mobile-first to an AI-first world."

It remains to be seen how much impact these intelligent machines will have on traditional commercial and consumer cultures. With automation of customer service, robotics and a wave of other AI impacts only now beginning to disrupt existing systems and shrink employment needs, it’s difficult to predict the effect. While it’s certain soft skills (machine and human customer service reps, for example) will be required in order to humanize events, the future evolution of employment models in an automated world remains open to debate.

Are thinking machines a threat to humanity or a business opportunity? Orange Business will attend the forthcoming InnoRobo event in Paris, France, where the world’s leading experts in robotics and associated technologies will share their ideas at this exciting time.

Stewart Baines
Stewart Baines

I've been writing about technology for nearly 20 years, including editing industry magazines Connect and Communications International. In 2002 I co-founded Futurity Media with Anthony Plewes. My focus in Futurity Media is in emerging technologies, social media and future gazing. As a graduate of philosophy & science, I have studied futurology & foresight to the post-grad level.