Remember when Apple meant iPods and iMacs, Google was just a search engine, and Facebook was a social network you hadn’t heard of because you were still on MySpace? Those days—which really weren’t long ago at all—are long gone. Instead of specializing, these three tech giants scarcely make a move without stepping on each other’s toes. The latest entry in this game of one-upmanship? Facebook M, a virtual personal assistant that joins the ranks of Siri, Cortana, and Google Now (aka “Okay Google”). On the strong side, M is smart—it’s powered not only by machine intelligence, but by actual humans. On the weak side, M is buried inside an app and entirely text-based. Let’s take a closer look at how M stacks up in the suddenly crowded field of virtual assistants.
Currently being tested by Bay Area Facebook employees, M was announced via a status update from Facebook Vice President of Messaging Products, Dave Marcus (who gave up being CEO of PayPal to take his current gig). The virtual assistant “lives” inside of Facebook Messenger, the standalone mobile app that has taken over the role of Facebook Messages (though those using Facebook through desktop computers or tablets can still send Messages without Messenger). From a user perspective, Marcus wrote in his post that M will “[enable] people on Messenger to get things done across a variety of things, so they can get more time to focus on what’s important in their lives.” From a business perspective, however, Marcus explained to Wired what anyone in the search business knows—people querying Siri or saying “Okay Google” are generally far along in the purchase cycle and looking to buy. “Intent often leads to buying something, or to a transaction, and that’s an opportunity for us to [make money] over time,” Marcus explained. In March, Marcus had already announced Businesses on Messenger, which allows retailers to provide customer service and send notifications (like receipts and shipping info) via the Messenger app.
The Human Equation
Facebook is positioning M as the perfect blend of machine intelligence (fast, accurate) and human understanding (empathy, judgment). Like other virtual assistants, M is mainly powered by artificial intelligence—it searches the web to answer your questions. The big difference is that M isn’t working alone. Facebook has hired numerous supervisors—called M trainers—to keep watch over M and help it “learn” how to better help users. Facebook anticipates hiring thousands of M trainers (employed as contractors, for the record) with backgrounds in customer service. The trainers ensure every request is answered, fine-tune computer-generated responses, and in some cases, actually fulfill the request themselves rather than having a computer do it. Interestingly, from the user’s perspective there isn’t a difference—you never find out if a computer, a human, or both helped you. Even the name “M” is purposely ambiguous (unlike say, Siri—even though you can change the settings to a male voice, Siri is generally referred to as a “she”).
This isn’t exactly a brand new business model. There are several paid virtual assistant services that have actual humans fulfill users’ requests, from quick research to actual, physical jobs like picking up dry cleaning. The difference is that those have not been feasible to accomplish on a large scale. TaskRabbit, for example, which is sort of like an Uber for human helpers, has considerable name recognition but is only actually available in a handful of American cities. Also, all of the people-powered assistant services (up to now, anyway) cost users money. Facebook M is free.
How will they make that work, especially given that as M moves beyond the Bay Area, Facebook will have to hire many more contractors? Facebook is banking on the success of M leading to partnerships with businesses (as Marcus implied). It’s not impossible to imagine the company developing some kind of PPC-style scenario where Facebook takes a commission on sales made via M. Additionally, the machine learning aspect of the service—the trainers are literally training the machines—means that over time, even as the amount of data increases the human element would become less necessary. (If that sounds scary, just keep repeating Asimov’s Laws of Robotics until you feel better.) They’ll have more complex AI at their disposal, with less human oversight required.
And speaking of touchy subjects—for now, there’s a “wall” between M and your Facebook profile, feed, and other social data. Though M can learn about your preferences from the chats you have with it (both your current query, and anything you’ve asked M for in the past), it doesn’t pull from your Facebook information. Experts agree that wall will likely disappear at some point, but given Facebook users’ many privacy concerns in the past, it will be tricky.
How It Works
You’ll know you’ve got M when it shows up in your Facebook Messenger app—you’ll be able to tap an icon (a squiggly shape that sort of looks like an M, but also somewhat resembles the symbol for infinity) and send M a message. M messages you back, asks follow-up questions as needed, and updates you when the task is completed. It can take on simple tasks (like recommending a restaurant), but M can also be used for more complex, multi-part tasks (like planning a party). The latter type of task differentiates M from other smartphone-based virtual assistants such as Siri, and it’s also one where more elements are likely to be tackled by humans. One of the most common tasks Facebook beta testers have asked M to handle? Calling customer service lines and waiting on hold, a task that as of now is pretty much immediately delegated to a human.
Something you may have noticed that is a major difference between M and Siri, Cortana, and Google Assist—M is completely text-based. Think about it for a minute: While you might speak to a voice-recognition assistant like Siri in full sentences, how often do you send texts in them? The screen captures of M conversations that Facebook has released show users typing in full sentences with perfect grammar, spelling, and punctuation. No emoji, either. While this is obviously helpful for the human M trainers, it may feel unnatural for many users. It is the way people talk out loud, but that’s not how people “talk” on screens.
If you’re using M for a simple task, it’s also pretty slow. One of the screen capture examples Facebook has provided includes the query, “I’m going to Chicago next week. I’m looking for a great burger. Where should I go?” If you strip away all the conversational overtones, what you’ve really got is a Google search for “best burger chicago.” And instead of taking the two seconds to do that yourself, you’re now going to the trouble of opening an app and having a conversation about it. Whether it’s a computer or a human on the other end of your M conversation, pretty much all they’re going to be doing is that exact Google search. You’re also going to get only one result, unless you decide to get picky and continue the conversation. Facebook’s stated goal is to help users get more done and have extra free time, but it seems like just using M will eat up a decent chunk of time all by itself.
The Future of M
Facebook knows that having to go to an app is a disadvantage—Siri and Cortana are built right in to their respective operating systems and are just one tap away as soon as you activate your device. What they’re hoping is that M will be so significantly better—more powerful and more personalized—that it will become your go-to. For more complicated tasks, M can act like a concierge, helping you make purchases, buy tickets, book reservations, or navigate the voice menus and long hold times to make changes to your cable service. For anyone who’s not already deeply attached to Facebook Messenger, however—and anyone who isn’t often faced with the kind of First World Problems the beta testers have used M to handle—this is likely to be a hard sell. The more likely angle? Just like the transcribed voicemail messages you get through Google Voice help the search giant make leaps and bounds in voice recognition (natural language, regional accents, slang—they couldn’t ask for better data), if enough users adopt M, Facebook will get massive amounts of data about consumer preferences, buying patterns, and more. Just because it’s a free service doesn’t mean that the social giant isn’t getting something in return.
It’s also fairly likely that the text-only aspect of M won’t last long. In January, Facebook acquired Wit.ai, am API-based natural-language platform that allows developers to incorporate voice recognition into any app, hardware, or software. Previously, anyone who wanted to have speech recognition and natural language had to develop it independently, a la Google Voice. Wit.ai CEO Alex Lebrun has a background that’s exactly that—he previously founded VirtuOz, which built voice-controlled software for major clients. Wit.ai takes a crowdsourcing approach, allowing developers access to their service so long as they share their own data with the community of Wit.ai users. Though Wit.ai remains separate from Facebook—and its business model hasn’t changed—it’s probable that its speech recognition database will be integrated into the social giant’s offerings at some point.
With social platforms coming to function more like search engines, allowing users to find services and discover new businesses, it may be time to reevaluate your social strategy. How up-to-date are your profiles? Do you have a presence on the right social networks for your audience? Higher Power SEO can help you answer all of these questions and more with our social media services. Call us at 760-881-4736 today to learn more.