Mark Zuckerberg thinks you suffer from app overload, and last week he unveiled his solution: chatbots. Over the next few years, they may boost the bottom line of Facebook Inc. FB -1.08 % and other Internet companies while changing the way you use your smartphone. But first, they’ll need to get a lot smarter.
Chatbots—or bots, for short—are stripped-down software agents that understand what you type or say and respond by answering questions or executing tasks. Apple’s Siri is a bot, and so is Amazon’s Alexa. But the bots that have Mr. Zuckerberg excited spring from the marriage of popular text-messaging programs and burgeoning artificial-intelligence techniques. If all goes according to plan, they will know what you like, remember what you’ve told them and cooperate to make your life easier.
One of the first bots was Eliza, built in the 1960s. She excelled at answering questions typed into a Massachusetts Institute of Technology computer terminal, using vague generalities that made her seem like an inexplicably shallow psychotherapist.
To Facebook, bots look like something more modern: a fresh alternative to apps—and a way to connect the company’s business partners with its 1.9 billion Messenger and WhatsApp users, especially born-to-chat millennials. And Facebook bots have an added bonus: They run in the background on the company’s Messenger service. That means there are no apps to download. Users get convenience, and Facebook gets its own alternative to Apple Inc. AAPL -2.01 % ’s lucrative App Store.
“No one wants to have to install a new app for every business or service that they want to interact with,” Mr. Zuckerberg said at his company’s F8 developer conference last week. “We think that you should just be able to message a business in the same way that you message a friend.”
Right now, Silicon Valley’s hopes seem to be pinned on the prospect of a bot revolution. Alphabet Inc. GOOGL 0.59 % ’s Google and Microsoft Corp. are pushing ahead with new bot initiatives, as are messaging companies Kik Interactive Inc. and Telegram Messenger. A crush of new startups is trying to develop the bot equivalent of a killer app.
“We’re all looking for the interaction that’s one step past us mashing on all these tiny rectangles in our pockets all the time,” said John Lilly a partner with venture-capital firm Greylock Management Corp.
Facebook is following the lead of Tencent Holdings Ltd. TCEHY -0.61 % ’s WeChat, China’s most popular messaging app, which already lets people use bots to shop or pay bills. Facebook Messenger currently hosts more than 25 bots that provide the same kinds of services you’d find on a website or app: weather forecasts, news updates, flower deliveries. But instead of pointing, clicking or swiping, users tap and send text messages.
That could lead to new revenue for a messaging business in which Facebook has invested billions. The company is in the early stages of testing out the idea of sponsored messages, and chatbots could be an important part of that.
Making money from bots isn’t yet a priority, Facebook said. But observers see great potential. Capturing half the predicted business for customer support via messaging would represent a brand-new $4 billion revenue opportunity, says Evercore Group LLC.
While tech giants are banking on bots, what is less clear is whether U.S. consumers will take to them. Bots invite us to interact with them as if they are human, but their digital minds have severe limitation
Last month, Microsoft introduced Tay, a chatbot designed to exhibit the personality of a bubbly 19-year-old woman. Tay’s artificial intelligence enabled the software to learn from user interactions on Twitter. TWTR 0.29 % Within hours, however, it was clear that Tay’s learning process had gone seriously wrong. “ricky gervais learned totalitarianism from adolf hitler, the inventor of atheism,” she tweeted out to her Twitter followers before Microsoft pulled the plug.
Microsoft blamed Tay’s decline into casual anti-Semitism on a coordinated effort by online pranksters to game the bot’s AI algorithm. In China, another Microsoft chatbot, Xiaoice, has been chatting away pleasantly for two years, with about 40 million WeChat users.
“When you train a machine learning system, it’s a bit of black box. You don’t exactly know what’s going to come out of it,” said Ben Brown, the CEO of XOXCO Inc., an Austin, Tex., company that built Howdy, a trainable bot that handles administrative busywork such as scheduling meetings on the Slack messaging software. “I just don’t really buy that a big brand, or even a small company, wants to run that risk.”
The initial run of Facebook bots already has faced backlash for failing to behave as expected. CNN’s Messenger bot, when asked “What’s happening?” replies with a disheartening shrug emoticon followed by ‘try again.’ ” When BuzzFeed writer Katie Notopoulos tried to shop for shoes using Jello Labs Inc.’s Spring bot, she came face to face with a relentless upseller.
To fulfill their promise, experts say, Facebook’s bots will need to get better at conversing in everyday language. “They aren’t taking natural language; they are taking menu names,” said Bruce Wilcox, the author of Rose, the winner of the most recent Loebner annual chatbot competition.
But with many of Messenger’s 900 million users now experimenting with bots, Facebook is generating mountains of new data about how they want these bots to work. And that is expected ultimately to make them more useful.
“Bots will largely replace apps,” said Mr. Wilcox, who left his job at Amazon.com Inc. AMZN 0.83 % for the bot startup Kore Inc. not long after his Loebner win. “Installing and accessing separate apps is a pain.”
—Daniela Hernandez contributed to this article.
Robert McMillan at Robert.Mcmillan@wsj.com