Last week, Facebook launched an updated version of its A.I. assistant, M, which is now empowered to suggest Messenger features as users exchange messages.
The original version of M debuted a couple years ago, built as a chatbot with its own messaging thread, and it was justifiably likened to Siri as it functioned on a quasi-Q&A basis. Now, the chatbot has been converted to engine for in-conversation suggestions related to the conversation taking place.
Current examples of these suggestions include sending GIFs relevant to the conversation, sharing location, or hailing a ride. As users engage with the suggestions, they will, over time, become more tailored to the users' habits. Naturally, annoyed users can deactivate M at any point.
While the previous iteration of M could hold a basic conversation and understand commands decently enough, this new version has left some wanting more, dissatisfied with what the update offers in terms of the A.I.'s application and potential.
Previously, M was a chatbot powered by humans, which put it leagues ahead of fully automated A.I. assistants like Siri or Google Now, but it's hard to discern how much of that humanity will be incorporated into the new version of M. Only time will tell on that front.
Nonetheless, in M's new form, there's a wealth of possibility for the assistant to claim, if the Facebook team makes the right moves. Messenger's monthly active user base topped 1 billion late last summer, so the A.I. will have access to a trove of users and have plenty of data to observe and learn from, a dream for many A.I. companies.
Even though it only recommends using Messenger features for now, M's natural language processing (NLP) capabilities stand to improve quickly if users engage with it, which opens many new doors.
Once the NLP is fine-tuned, Facebook's messaging team should grant M access to new data inputs, such as user profiles and activity data, data that is already proprietary to and voluntarily shared with Facebook. From there, M could move to connect users to other parts of the social platform, such as events, pages, photos, and fellow users' posts and drive increased engagement.
After the M team gets a handle on how users interact with M and performing actions such as likes and comments, the door could open even further to third parties.
If Facebook doesn't pursue a walled garden approach, M could provide a foundation for deep linking to other apps, like Spotify and Google Maps, to better connect users with their objectives and interests -- in other words, a platform within the platform.
In addition, Facebook could open the door to outside developers to build apps, plugins, and games, much like was done with the original Facebook platform nearly ten years ago. As a development platform, M could draw on powerful network effects and deliver real value to users, developers, and, most importantly, Facebook itself.
Ideally, M could then become an independent engine of commerce, connecting users with opportunities to consume media, make purchases, and find useful information, like nearby restaurants or events.
Messenger already pulls data like location and browsing habits - connecting it to data from Facebook, such as user likes and business locations gleaned from those businesses' page.
Imagine two people chatting about getting dinner at the new Indian restaurant on Messenger. M could extrapolate a lot of information from that context and then serve up a recommendation to reserve a table at that restaurant via OpenTable (maybe one day Facebook could provide this service on its own, moving closer to end-to-end presence in our daily lives).
While some may view M's new format as a step down for the A.I. service, it does have a path to becoming a platform and becoming a hub for users' entire activities, much like WeChat has already accomplished in China.