In fairness
Facebook, the social network created in 2007 by Mark Zuckerberg et al, has changed the way people connect and how information is disseminated. Today, with an estimated 2.9 billion who are active on Facebook, user data input in the app/site had been greater.
With billions of information, one can imagine the tedious work in managing data. This is were artificial intelligence has become very beneficial and advantageous in handling colossal information. And with these information, Facebook AI has become a vital part in developing in-app technologies for a helpful and cool user experience.
All About Facebook AI Research
Zuckerberg had been sharing his vision that would help the company last for a longer time. He started getting his hands on artificial intelligence, and being impressed with the technology, planned to make an AI R&D laboratory.
He had talks with Mike Schroepfer, and Yann LeCun, the one who made convolutional neural networks (CNNs), a very helpful tool in image identification. With their similar take on the future of technology, LeCunn became FAIR’s CEO and pooled in exemplary tech scientists and researchers across the globe.
And in 2013, FAIR was made. Today, Facebook AI Research has more or less than 100 employees, with research laboratories in New York, California, Paris, Montreal, Seattle, Pittsburgh, London, and Tel Aviv.
9 Technologies Facebook AI Research Has Developed
The Facebook AI arm has quite a handful of artificial intelligence tools and applications. We look into their 9 cool techs FAIR has developed and/or invented which are being used in the Facebook app/site.
DeepText
Did you know that Facebook’s suggestions after your post is powered by AI? When a Facebook user posts an item for sale, the app might suggest things that can boost the user’s chance of making a sale. The app understands the context behind of your posts, and this has improved one’s experience in using the app or site.
The company boasts of the tool’s near-human accuracy, which can process thousands of posts per second all through 20 languages. This cool feature uses deep learning, a more advanced method than natural language processing for understanding text.
DeepFace
Ever wonder how Facebook can detect who is in a posted picture? The face-tagging feature in the app runs on DeepFace, a facial recognition tool with artificial intelligence in it that was developed by FAIR.
DeepFace boasts of about 97% accuracy, which is better than actual humans in identifying people through face. It was trained with 4 million actual real-world data of faces from Facebook users. The AI technique behind this tech is a neural network of 9-layers with more than 120M connection weights.
Automatic Alternative Text
For a Facebook user who is blind, this integrated tool is very useful in describing images in appropriate detail. The Automatic Alternative Text is notches up to the classical alt text when labelling images that one uploads. The description is automatic; you do not have to manually type a description of an image.
It employs the use of advanced object recognition, a neural network technology with parameters in billions and trained with millions of raw and organic images with description.
M2M-100
When a Facebook user posts a text in his/her native language, you will see at the bottom the “see translation” link. It turns out, artificial intelligence is also behind this feature.
Also known as (MMT) Multilingual Machine Translation, M2M-100 can translate 100 languages with each other directly, without having to process its English translation first. This is a big leap in language translation because it allows faster data processing and other languages can become a training data for computers besides English.
M
M is an automated AI assistant within Facebook Messenger. If you have noticed some pop ups after your reply in the messaging app, it is M’s working. M is assisting on some cool features in-app like instant reply, reply suggestions, gif images sharing, location sharing, cab hailing, and event planning.
M is a very handy AI assistant and its main objective is to help, though it has downsides such as a user might having feelings of being violated his/her personal space and information. If you have such feeling, you can simply turn off this non-voice AI assistant.
Hate Speech Detector
With the nature of sharing sentiments in social media, Facebook AI has developed a tool which can detect hate speech, bullying, and targeted attacks. Perhaps it implements a method similar to DeepText, but is more specified in distinguishing metaphors and exaggerated expressions from real life-threatening harassments.
Suicide Prevention Tools
Suicide is one of the leading causes of deaths in the U.S., of which clues before the devastating deed can be read or seen on a Facebook user’s profile. In order to hinder such regretful action, FAIR has developed a way to help those who are on the verge of suicide. Facebook detects such genuine calls for help, and, similar to M, suggests steps or provides comforting words for the would-be victim.
fastText
fastText is an open-source and lightweight implementation which improves vector representation and text classification. It is helpful in detecting spam posts, and words with similar meanings needed in categorizing posts.
Facebook Marketplace Indexing System
Facebook has developed an improved indexing system and autosuggestion in their marketplace. After making a search query, buyers can also view other items for sale similar to what they are searching. For sellers, when posting a listing, the app or site will suggest the price range and category the product is in. Photo enhancement is also possible in listings with images.
Wrap Up
Facebook has come a very long way in being just a social media site. They have cleverly used their resources to aid institutions, businesses, people, communities, and countries where Facebook is openly operating. With delving into artificial intelligence, Facebook might be able to sustain its legacy as a leading IT company many years from now.
There are news of data exploitation by the social network, but to be fair, information posted on the app/site have also been vital in developing innovative technologies, improving their services, and targeting illegal, dangerous, and life-threatening activities. They can only use what we post on Facebook, so there is a call for responsible, careful and mindful posting of text, image, videos, and events.
