M, AI and FB: How Facebook is Going to Become an Even Bigger Part of Your Everyday Life

Back in August, Facebook introduced 'M', its new personal assistant tool which will live inside Messenger. Like Apple's 'Siri' or Microsoft's 'Cortana', users will be able to ask questions of M and have the system respond, via message, providing detailed information and assistance as required. Where M goes beyond those previous personal assistant services, however, is that M can do more complex tasks – like booking plane tickets, chasing up a refund on your behalf, or even finding the best deals and making purchases for you, at your request.

In constructing M, Facebook's taken a slightly different path. While M is largely AI based, and responds to most queries based on algorithms, M also has a team of people working to help the system deal with more complex requests, stepping in when the system gets confused and providing additional assistance and advice on top of M's binary workings. And in the process, M is learning.

For example, if you were to say to a computer 'I want to go out on a date', an algorithm may not necessarily be able to determine if you mean 'date', as in a measure of time, or 'date', as in a romantic meeting (or the fruit, for that matter). To assist, one of M's trainers would be alerted to this confusion and would step in to clarify your intent, helping streamline the process and provide the right, contextual, response. When this happens, M's system takes note and correlates that interaction to improve the likelihood of a correct automated assessment next time it receives a similar query. This is a rudimentary example, for sure – no doubt the system has enough contextual, conversational data to understand the differing meanings of 'date' – but you get the idea, for every interaction and every trainer intervention in the process, M's system learns and alters it's predictive algorithm to better weight the most probable and logical response.

Over time, by refining M's 'brain' (for lack of a better word), M's trainers are effectively training themselves out of a job. The more they help the system learn, the less it'll need them to intervene – but given Facebook's eventual desire to roll out M to millions of users, that system learning could take some time to refine. Facebook Messenger lead David Marcus has already said the company may have to hire thousands of assistants for M along the way, so Facebook is prepared for this, but manual intervention at that scale would come at a high cost. 

So why is Facebook so keen to spend big in order to make M a success? Part of that answer lies in the latest testing of M, which highlights the capacity of the new system, while also flagging what it could become.

Over on BuzzFeed, journalist Alex Kantrowitz has been using M for the past few weeks, putting the new system through its paces and seeing what it's capable of. Kantrowitz has presented some pretty amazing examples of what M can do:

"Over the past few days, I asked M to find and book me a cheap flight to New York, to plan an itinerary for myself and a friend to watch the Mets in the World Series in person (too expensive), to monitor game and flight tickets and alert me if they drop in price, to get me a refund on a streaming package after it stopped working (success!), to find a bar close to a concert venue, and to shop the internet for products and buy me the ones with the lowest prices. (For Amazon purchases, M claims it will ship them to you for free via Amazon Prime, regardless of whether you are a Prime member or not.) It doesn't stop there. Every morning at 7 a.m., M sent me a story about the New York Jets; it also sent updates on Mets' slugger Yoenis Cespedes' health. It sent me four messages so I wouldn't forget my backpack at the end of the concert. Last Wednesday, I told M I had a bad day and asked it to send me something nice. It responded with a photo of a bulldog overlaid with text: "Cheer up buttercup, I'm here to make your day better."

Some of Kantrowitz's interactions with M clearly required human interaction – for example, he asked M to draw a picture of San Francisco Giants pitcher Barry Zito:

(image via BuzzFeed)

Obviously, there's a clear level of human intervention there, but at the same time, Kantrowitz notes that he's often not aware of when he's interacting with a human or a bot.

But the most interesting point in Kantrowitz's observations is contained within his final note:

"Without question, these assistants will make your life easier, but they will also give tech companies a much deeper understanding of your identity. Facebook already knows your birthday, friend list, and much more; with M it could soon know when you wake up, where you like to sit on airplanes, what movies you see, and whether you prefer road or mountain bikes. That's a lot of information to entrust to Facebook, which has a long and storied history of privacy missteps, so M is best approached with caution and common sense. That said, excuse me while I ask M to plan out my evening."

In this one summary, Kantrowitz captures the most likely end-game scenario for M – that the utility and convenience of a using a service like M will outweigh the associated privacy concerns and risks that may come with giving over all that data to a company – a company which, at core, will use that information for profit. We've already seen this situation play out with Facebook's main platform – while people are concerned about giving over their personal information and having their digital activities tracked, most are quite willing to do so, because the alternative option would be to avoid it, which would mean denying yourself the ability to utilize such a powerful, entertaining and omnipresent tool.

But using M would give Facebook a vast range of new data. As noted by Kantrowitz, M is learning everything about his day-to-day life, from the time he wakes up, the products he's interested in, the places he's traveling to. The more you utilize M's services, the more it will be able to learn and store such insights, matched to each users' Facebook identity. That's an amazingly powerful tool – right now, Facebook has all the info on who your friends are, what you're interested in and what you talk about. But through M, it would also have access to what you search for, what specific details you're interested in when considering certain products, your takeaway food preferences. Every query you enter is another detail, which, when combined with the potentially billions of insights M might be able to glean from its interactions with other users, will start to form even more definitive profiles of users, enabling the most accurate ad targeting and outreach opportunities you could possibly imagine. In future, Facebook's algorithms may literally know what you want before you've even considered it.

This is why Facebook is putting so much effort (and money) into the M project – to gather more insight. As marketing and advertising evolves towards more data-driven process, the company that has the most data wins. And Facebook is winning on that front, with a wide range of insights into their more than 1.44 billion users. But, of course, data isn't everything – a platform's only able to obtain that level of information if they provide services that users feel compelled to use. Facebook knows this – in a recent profile of Facebook's bigger plans for Messenger in Wired, Facebook's Head of Management for Messaging Products Stan Chunovsky noted that "user experience is all" in their development process.

"The best user interfaces and least friction translates to happy consumers who spend more time on us." 

Interestingly, in the same piece, Facebook's Vice President of Messaging Products David Marcus noted that they're not looking to make money from payments made on Messenger, they're more interested in getting as much traffic as possible to the app, then monetizing that attention.

"eBay takes a cut of every transaction and listing; Alibaba does all that for free, and makes money from advertising. Alibaba is bigger than eBay and Amazon combined, and is growing much faster. We take the same approach. We want the maximum number of transactions on the platform, while enabling the best possible mobile experience for commerce. The margins on payments aren't that high, and we want the broadest reach. Businesses will want to pay to be featured or promoted – which is a bigger opportunity for us."

This is where Facebook is looking to dominate, not from the incremental monetization of their offerings, but through advanced use of the data and insights they obtain in order to improve their ad targeting and audience outreach. Those efforts are advanced by gathering more data, something Facebook is set to boost significantly if it's able to find a workable and scalable way to make M available to everyone.

Whether that happens or not remains to be seen, but the framework is there, the platforms are being constructed. While much of the initial response to M will likely be based on novelty – people sending the through crazy queries just to see what comes out – it's the long-term use of the system that will prove of most value. Streamlining those everyday queries, managing your common hassles – these are the areas where a service like M could provide a solution. And in providing that utility, they also work to build Facebook's empire and improve both time spent on platform and Facebook's ever-expanded understanding of each and every detail of its audience.

Right now, M still feels like a distant reality. But it's not. And it could change things more than you know.  

View the original article here

Original source: M, AI and FB: How Facebook is Going to Become an Even Bigger Part of Your Everyday Life.
http://www.jretechnology.com