Facebook's AI Tools Now Recognize Different Languages - and Can Negotiate Deals
While Facebook's Messenger Bots haven't gained the momentum some thought they would at this stage, The Social Network is continuing to develop its artificial intelligence capacity, adding new options and processes which could provide significant benefits - both now and in future.
This week, Facebook has announced a couple of new developments on this front - firstly, Facebook has added new Spanish language support to their 'M' artificial assistant tool.
Facebook first introduced 'M' back in 2015 - similar to Apple's 'Siri', M is designed to be able to provide assistance on a wide variety of tasks, with the initial vision being for M to be an all-encompassing personal assistant that could not only help you find information, but could also order products on your behalf, pay your bills, organize the best deals on flights, etc.
After an initial trial with users in the Bay area, M was eventually scaled back and rolled out to all U.S. Messenger users as 'M Suggestions' back in April.
The system still provides automated recommendations, but it now works by recommending ideas based on message context.
The new Spanish language capacity expands upon this, opening up the system to a whole new range of users.
As Facebook continues its global expansion, providing multi-language support for all functions will become increasingly important. In fact, it's already become a key element - as noted by TechCrunch, over 800 million Facebook users per month now utilize the platform's multilingual sharing button, which enables Pages to post in one language and have it appear to people in their native tongues. And with more than 437 million people speaking Spanish as their native language, the option provides a much wider use case for M Suggestions.
The new feature is rolling out in Mexico from today.
The second major announcement on the machine learning front comes via the Facebook Engineering team, who've announced that they've successfully taught bots to negotiate in order to achieve specific outcomes.
Sounds a bit concerning, right? Now bots can actually understand, and negotiate a better deal?
Yes, they can - as explained by Facebook researchers Mike Lewis and Dhruv Batra, they've been able to successfully establish a process where machines can be taught to negotiate their way to an optimal outcome based on numerical values placed on the available options.
You can see the process more clearly in this diagram - the system is taught to aim for the highest possible score in each negotiation, and given a set range of questions within which it needs to establish an outcome.
If the system can't make a deal, it gets zero points, so it's incentivized to come up with the best solution possible, within logical reason.
And the process has provided some interesting outcomes - as noted by Recode:
"Facebook says that the bots even learned to bluff, pretending to care about an outcome they didn't actually want in order to have the upper hand down the line. "This behavior was not programmed by the researchers but was discovered by the bot as a method for trying to achieve its goals," says Facebook."
While right now such systems don't provide much immediate functionality, in future, an automated system capable of negotiation could be used to get you better deals when making purchases online, to help settle disputes and achieve better outcomes without taking up your time, or even organize business deals on your behalf.
Sounds extreme? This is where such tools are headed - and worth noting too, Facebook says that:
"Interestingly, in the experiments, most people did not realize they were talking to a bot rather than another person - showing that the bots had learned to hold fluent conversations in English in this domain."
Processes like this could provide significant benefits - both from the business and consumer perspective.
It's a little scary, sure, a little concerning that you might soon be dealing with bots and not know it. But it's interesting to consider the potential, and where such tools might provide benefit.
Follow Andrew Hutchinson on Twitter