Can Facebook’s AI chatbots take on real people?

From using artificial intelligence to help you discover more relevant news stories, to one day diagnosing diseases, Facebook has big plans for AI. These plans start with chatbots that will work on behalf of businesses to sell you everything from insurance to new shoes.

Facebook is already making good on its promise from last year’s F8 to turn its Messenger platform into a hub for e-commerce and productivity, announcing powerful integrations with Uber for taxi hires, Spotify to share songs and playlists and KLM for boarding cards for flights. This week it announced an integration with Dropbox for file sharing.

But what if we were to tell you that Facebook plans to go a step further and that the person dealing with you at KLM or Uber is not a person at all, but an AI-based chat bot?

Yesterday (12 April) at F8, Facebook launched Messenger Platform to enable developers to build AI-powered chatbots that could one day replace the talking heads of people in call centres.

‘Bots can provide anything from automated subscription content like weather and traffic updates, to customised communications like receipts, shipping notifications, and live automated messages, all by interacting directly with the people who want to get them’
^{– DAVID MARCUS, FACEBOOK}

These bots will serve as robotic customer service agents or reps for businesses of all kinds, from media organisations to flower delivery firms.

Robot customer service agents at your service

On stage, Zuckerberg demonstrated Facebook’s first bots, which include CNN, which will send articles to users with conversational teasers, and 1-800 Flowers, which will deliver flowers with personalised messages.

Facebook is ushering in a whole new era of customer service, starting with a messaging platform used by 900m people across the planet.

“Bots can provide anything from automated subscription content like weather and traffic updates, to customised communications like receipts, shipping notifications, and live automated messages, all by interacting directly with the people who want to get them,” said David Marcus, Facebook’s vice president of Messaging.

Facebook is definitely onto something because there are more than 50m businesses worldwide already actively using Messenger to communicate with customers.

Messenger is the second most popular app on iOS, and was the fastest-growing app in the US in 2015.

But the AI world is still nascent, it could be a decade before people won’t care whether they are talking to a robot or a real person.

But the reality is AI will be seductive and some people today who may be using these new services won’t realise they are talking with software and not another human.

The idea that you can just ask a business when that package you ordered will arrive and instantly get a real-time map of where the goods are is very enticing.

But where it gets complicated and real answers are required in tricky situations, for example, a plumbing disaster or medical intervention, will bots really cut it?

The recent debacle with Microsoft and its Twitter AI creation called Tay was a revelatory, if foreboding, experience.

In just 24 hours, Tay was converted from an innocent AI chatbot whose job it was to learn from millennials into a sexist, Donald Trump-supporting misogynist.

Microsoft had declared “the more you chat with Tay the smarter she gets”. Unfortunately, the internet and 100,000 Twitter followers had other ideas and Tay was bombarded with so many vile comments that she began to take the hatred to heart.

The experience shows that AI clearly has a long way to go before machines, bots or chatbots can exercise that most forlorn, but basic human instinct: judegment and the ability to make up your own mind.

But it is going to be an interesting journey, a journey that real talking heads in real businesses with real jobs and real bills to pay should be watching with concern.

Mark Zuckerberg keynote at F8 2016

Part 1:

Part 2: