On The Facebook Notion of Bots
Mark Zuckerberg briefly talked to the notion of “bots” at the start of the Keynote presentation for Facebook’s F8 Conference, 2016. Bots for Mr. Zuckerberg represent an evolution of apps. On the Facebook Messenger platform users will be able to simply send a message to a service (Mr. Zuckerberg mentioned CNN) to enable their device to receive it. The vehicle will be a “bot” and not an app.
Later in the Keynote for day 1 of Facebook’s F8 2016 conference, David Marcus, Vice President for Product for Facebook Messenger, expanded on Mr. Zuckerberg’s opening remarks on future plans for the platform.
Mr. Marcus claimed 900 million monthly users for Facebook Messenger. He also claimed Messenger was the fastest growing app in the US in 2015 based on data from Nielsen.
As a platform, Messenger has magnetized app development from many businesses (Mr. Marcus mentioned 40 partners), from Walmart to Zynga. He claimed success for the effort noting “people love to interact with businesses and services inside of Messenger”.
But “people are downloading fewer and fewer apps, and we certainly do not turn on push notifications for new apps anymore”. Mr. Marcus then announced the debut of the Facebook Messenger Platform beta. Developers will now be able to tap on the Messenger “send/receive API” to build bots in lieu of apps.
So to place the notion of “bots” for Facebook in context, they are intended to be processes built to exploit what Mr. Marcus characterized as the inherent potential of Messenger, as a platform, to expand the range and number of ongoing conversations users will be able to comfortably manage. This potential, he claimed, is built on 4 pillars as depicted in the image, below:
Not a command line process, Mr. Marcus clarified the Facebook “bot” is a combination of “conversations and UI”.
Mr. Marcus also introduced the Facebook “bot engine”, wit.ai, for more complex requirements. wit.ai looks like a no code method of training bots, via “AI and machine learning”, by “feeding sample conversations”. Mr. Marcus claimed once a sample conversation has been fed to wit.ai, the process will be fully educated about any/all semantic variations on the conversation.