We used to think that robots aren’t smart enough to replace us in the workplace, but it turns out we’re teaching them how to do exactly that.
Two months ago, Facebook announced the launching of M, its virtual assistant taking residence in the Messenger app. It’s built after Siri’s model, but there’s a new twist to it: there’s a supervising team of actual humans who monitor M’s responses and jump in when M is stuck.
The team behind M is also authorized to make phone calls – and do a little bit of extra research – just to make sure the user receives a personalized answer. But the human help isn’t going to be around forever; Facebook announced it is planning to use the assistants’ help as a way to teach M to make better choices of its own.
What they’re doing with this artificial intelligence is actually inspired from real-life, where people sometimes need to train a younger, faster and cheaper replacement to do their job. So far, M has become accessible only to a few hundred users in San Francisco, which means the information of the cyborg’s abilities is still limited.
But since Alex Kantrowitz at BuzzFeed was given access to M and had a chance to play around, we got a chance to see just how helpful the M virtual assistant can be. While it appeared that M did the initial work on its own, it started to become clear by the end that humans have taken over.
M’s responses are much more advanced than anything Siri could ever be capable of, such as combining information from multiple requests, picking up on vague hints, and using implicit information. M has also proven to be a master of colloquial language, something that a machine would have a lot of trouble learning on its own.
It’s clear that without human assistance AI can be frustrating and sometimes completely misunderstand simple requests. That’s what happened in previous M tests when human technicians weren’t involved; and even when commands were understood, delayed responses were a tell-tale sign of the machine’s difficulties in processing information.
But there’s a new method around the block called deep learning, and thanks to this Facebook’s AI is definitely improving. In order to be able to process massive amounts of data, the machine must be prepared beforehand with said amounts of data.
In the case of Facebook M, humans are accelerating the process by teaching it all that it needs to know. Facebook hopes that one day, with enough data and technological improvements, these assistants will train themselves out of a job.
Image Source: Tech Juice and Twitter