Currently, M is available only to approximately 10,000 people in the Bay Area, as it is still in Beta. This would mean that Facebook could open up a major new front in its battle with Google for your attention, in the event M works. It could also mean making huge money from direct-response advertising that has so far evaded the company. In the event that M doesn’t work, it will bring about a huge change on how we communicate with computers, as it is in the forefront of services. It also means a huge change from using our devices as fussy tools to asking our devices to simply take care of things for us. Dennis Mortensen, CEO and founder of x.ai, a virtual assistant that schedules meetings for its users says “I just can’t imagine a future in which I pick up my iPhone and browse 115 apps to do a task. There are plenty of jobs where I don’t need the app—what I need is the job done.” Products such as Apple’s Siri, Microsoft’s Cortana and the voice-recognition systems Google integrates into its apps and operating systems are trying to get to this future included by their companies. There are many advantages that Facebook has over these apps. Firstly, chat is already a familiar interface for billions of people unlike voice-driven systems. According to Michael Wolf, chief executive of tech consulting firm Activate, the total number of people using messaging services is more than 2.5 billion, and estimated to grow to more than 3.6 billion by 2018. The second is scale, as Facebook messenger already has more than 700 million users. This results in Facebook’s third advantage, as it has a probable way to get the huge amount of controversial data needed to make a chat-based assistant sufficiently automated, as it has access to so many users. Alexandre Lebrun, who heads the Facebook team that built the artificial intelligence for M says that currently requests sent to Facebook’s M are being answered by humans who are taking part in replying requests. From sending your friends a parrot to asking M to make you a restaurant reservation, these requests can take almost any form. Mr. Lebrun says that M is watching humans handle requests that M itself isn’t yet smart enough to handle, and every day it learns a little more and gets a little better at handling those requests. This implies that the number of humans “powering” M have been kept more or less constant by Facebook. Further, on account of the increased automation, progress made by its underlying algorithm allows Facebook to roll it out to more people every day. (A Facebook spokesperson said that the company will also hire more human trainers for M eventually.) Mr. Lebrun says that the important technology here is not that the AI that powers M – the algorithms and underlying code are largely open source, but the data that Facebook is collecting from every conversation. Even at least without hiring many thousands of people to power M, it is still so reliant on humans to do its job, Mr. Mortensen says he cannot think how Facebook will make M available to all of its users. However, Mr. Lebrun says that if M can take care of at least 80% of user requests on its own, “we have a very scalable solution.” He adds that the most recent requests are the ones that can easily automated the most. The question as to who gets to control primary interface of mobile devices is what is at stake here. If search was the dominant interface of the PC and the web, then chat, both over a text-based interface and through voice has the capability to take over as the first place users go to on mobile devices. The stakes are high. By far, Google is the biggest player in “direct response” advertising as of now. Direct response advertising is the kind of ad that pops up when we are searching for a new phone, hotel reservation, or any other purchase. The share of digital advertising that goes to direct response ads is at 59%, according to a 2014 eMarketer report. This means that if Facebook can persuade people to use a digital assistant like M in place of a search engine, it can capture more than double the size of the ad market and the gains, most probably, would come at Google’s cost. M also has the ability to open up Facebook to the same critical examination faced by search engines before—namely, how can we trust the results? How do we know that the restaurant didn’t simply pay to be suggested? Unlike a search engine that provides us with a list of options, with the digital assistant we are allowing the system to make a decision for us. It is likely that Facebook is taking on a task that is just too big. Ted Livingston, CEO and founder of Kik, a popular messaging app, says his own company is working on “bots” that lets users to interface not just with the Internet but with things in the real world, like menus at restaurants. The users are provided with a straightforward set of preprogrammed choices via Kik’s solution. “We’re not sure if people are looking for this all-encompassing AI experience,” says Mr. Livingston.
