Since its inception, the social web has been out to automate our interactions. Having built a digital repository of human behavior, institutions like Facebook can keep us willfully engaged by doing our social dirty work on our behalf (tagging photos, remembering birthdays). Where Facebook once learned our favorite movies, brands and people through an account sign up questionnaire, algorithms now deduce our preferences far beyond these basic parameters. It’s getting harder to lie to Facebook as less manual data input is demanded of its users, opening the gates for increasingly targeted advertising and customer service mediums.

Our every action is telling of our intentions — travel plans, fashion purchases and political voting tendencies. Specifically, Facebook Likes can be used to automatically and accurately predict highly personal attributes including sexual orientation, ethnicity, intelligence, drug habits, and whether one’s parents are divorced, according to an early  University of Cambridge study.

As far as Facebook’s concerned, such knowledge translates to better automated services for users. And in the coming months and years, better automated services will require a wicked smart artificial intelligence (AI) system. The past two years have seen significant investment from Facebook on the AI front, gaining extra attention in January after founder Mark Zuckerberg proclaimed his resolution to build a real-world Jarvis, the virtual butler astute enough to achieve sidekick status in the comic-turned-film Iron Man series, for his smart home.

AI the common denominator

At Facebook’s F8 Developer Conference this week, the social networking giant provided the big picture behind the many snippets of AI developments emerging these past two years. Facebook’s 10-year roadmap to further mobilize social communities through messaging applications, extend an advertising ecosystem with service bots, and retain a global user base with video-centric media sharing, all intersect the company’s AI initiatives.

In fact, Zuck’s desire for a simple AI butler that acts on voice commands isn’t so different from the bot developer platform launched at F8. Initially rolled out for limited services like hailing an Uber or Lyft ride and sending money, Facebook’s newly open-sourced M personal assistant bot offers real life examples of how Messenger bots can perform the types of tasks many busy people would love to outsource to a butler (virtual or otherwise).  Examples include weather updates, content delivery and shopping, while the most promising commercial use of Facebook bots centers around customer service. Looking at the history of the M program, Facebook could very well be seeking to deliver a Jarvis-like service sooner than expected. From Popular Science:

We’re used to digital personal assistants by now, like Siri, Cortana, and Google Now. But Facebook took a different approach with its new AI personal assistant, M, which offers the ability to execute complex tasks outside of the confines of your phone. Siri can send a text, but M can book a flight and make travel plans. During the development process, a Facebook employee even got M to schedule a series of in-home appraisals with moving companies.

Facebook’s Alex LeBrun, working on M, told PopSci that AI not only makes M better for accomplishing generalized tasks, but also for specialized cases, like traveling with an infant or during blackout dates. As AI grows, so does M’s capabilities, and LeBrun hopes that within three years, M will be able to call the cable company or DMV and wait on hold for users.

Commoditizing hospitality, selling leisure

The commoditization of hospitality services and customer service brings with it the promise of the leisurely lifestyle to the masses, gleaning unthinkably personal details under the guise of convenience. If it makes life easier or comes with rewards, we’re more likely to volunteer personal information, especially with trusted brands, according to the findings of a recent Aimia study in partnership with Columbia Business School. This user willingness works in Facebook’s favor, considering bots, and ultimately a Jarvis-like service, rely on this data to teach themselves exactly how to be human. The success of Facebook’s bots depend on AI’s humanity — they’re interacting with humans, after all.

Furthermore, Facebook knows more about its users than what’s revealed by actions performed on its social network applications. Cookies offer Facebook a peek into its users’ third party browser activity, as long as they’re logged into the social network (and sometimes when they’re not). Heading into even murkier waters, the use of Facebook’s login API grants additional access into users’ activity beyond the confines of the social network proper. It’s (mostly) legal for Facebook to use cookies in these ways for the purpose of targeted advertising, perhaps the purest of reasons to automate social media interactions.

To serve is to know

To be an effective service bot provider, Facebook’s AI must know its masters at the most personal level. Traditional butlers are trained to be intuitively attentive but always in the periphery, silent witnesses often privy to sensitive information. Automating this observation process enables Facebook to know individual users better by knowing all users better, thanks to inference and statistical analysis of its ever-growing data pool.

“People may choose not to reveal certain pieces of information about their lives, such as their sexual orientation or age, and yet this information might be predicted in a statistical sense from other aspects of their lives that they do reveal,” explains the aforementioned Cambridge study on the predictability of private traits based on digital records of human behavior.

Armed with digital representations of individuals and user populations, Facebook’s AI now acts as a machine-built imagination engine intuiting our needs. Yet Facebook isn’t alone, as Google, Amazon, Microsoft and Apple are also investing heavily in AI technology towards a service bot proactively delivering information, entertainment and actionable recommendations. Google Now can help track flights and send diet-specific restaurant recommendations for travelers, pushing out notification reminders instead of waiting for user inquiries. Amazon Echo is exploring voice commands for smart home management, while Amazon Dash buttons automate frequent home purchases such as laundry detergent and toilet paper. Microsoft’s Cortana and Apple’s Siri both expand the capabilities of smartphones, automating menial tasks for things like calendar maintenance and proximity reminders.

The social advantage

Facebook’s AI advantage is an ability to evolve as a result of ongoing interactions with the network’s highly nuanced social graph. Hopefully Facebook can avoid the pitfalls of Microsoft’s short-lived Tay bot, which turned into a ranting racist only hours after being introduced to the human species through a debilitating series of Twitter interactions.

Slow and steady could help Facebook’s AI win the race in this case, learning deliberately and with the aid of human course correction. According to Yaan LeCun, inventor of convolutional neural networks (CNNs) and director of FAIR (Facebook Artificial Intelligence Research), even the best AI system is still “dumb” compared to human brain cognition.

The user experience must also be top of mind, as “point and click is still easier than texting” for applications such as shopping, said IDC Research Inc. analyst Karsten Wiede.

It will take more than just AI to make a user-friendly experience, as the economic impact for a Jarvis-like service is being considered with caution by Wall Street. Facebook shares rose only modestly yesterday after the announcement of M’s open sourced options for developers, and brokerage firm Cantor Fitzgerald’s Yousseff Squali doesn’t anticipate Facebook’s bots to make money any time soon.

“It’s going to be awhile before we see any revenues,” Squali said in a Bloomberg interview. “Remember, Messaging has been a very, very difficult model for Yahoo, AOL … so for [Facebook] to take this Messaging platform and build it into a real ecosystem that will keep people plugged in for an extended period of time will eventually transform itself into an e-commerce platform.”

To expedite the possibility of execution, Facebook turns to its aggressive video initiatives to learn even more about human interaction, even faster.  Facebook’s earliest AI efforts focused on visual learning in images and video, along with natural language. Now that live video broadcasts have become such an integral component of Facebook activity, the network has a plethora of new data points for social contextualization.

“We will be able to read the video and know what they are about,” said Zuckerberg at this week’s F8. “The aim will be to deliver more relevant content, but Facebook’s ambitions will be much larger.”

Feature image by Alexas Fotos