January 10, 2018
It appears Facebook bit off more than they could chew with their AI experiment, the powered virtual assistant, better known mysteriously as ‘M’.
Released in August 2015, M is a text-based virtual assistant that involves using human employees to train an artificial intelligence system. After being offered to around 2,000 people in the San Francisco area, Facebook are bringing the human-assisted part of the service to an end.
Facebook’s goal was to develop artificial intelligence technology that could automate the tasks M was operating, without human interaction, for instance, book Ubers, make dinner reservations, change flights & send gifts. Pretty impressive! Unfortunately, Facebook learnt the hard way when M only surpassed 30% of automation. It could be argued that it was a great achievement that Facebook were able to keep the ambitious M in operation for two years, seeing as it required a human workforce, and the high cost factor would definitely have contributed to Facebook falling short.
Facebook are viewing M coming to an end as the end of an experiment, and they’re using the insights gained to power & guide strategy for other AI projects going forward. The project fulfilled what it was supposedly set out to do, learn what Facebook users needed & expected from a virtual assistant. All is not lost, as some features that M inspired, for instance, M suggestions, will continue to run.
This shouldn’t be regarded as a failed attempt at AI for Facebook. It is always going to be challenging to integrate AI seamlessly on social media platforms, particularly when up againstcompetition like Amazon’s Alexa & Google Assistant. With any luck, Facebook will take it one step at a time, and adopt a ‘slow and steady wins the race’ mantra when it comes to constructing a new virtual assistant challenge. We definitely haven’t seen the last of AI & chatbots on Facebook…