Facebook founder Mark Zuckerberg explains their goal is to help computers “understand language more like a human would with context,” rather than just rote memorization. To achieve this, Facebook has the AI Research team (FAIR) focused on developing systems with human-level intelligence. Their research includes AI theory, algorithms, applications, and both software and hardware infrastructure for data knowledge extraction.
Data is given from the news February 2017, user numbers continue to rise for Facebook as there are now 1.86 billion monthly active users, a 17 percent increase from this time last year. 1.23 billion people use the site daily, up 18 percent annually. Mobile adoption also continues its upward momentum with 1.74 billion monthly active users, a 21 percent rise.
Image by: [venturebeat.com]
All of the people are uploading 136,000 photos and updating their status 293,000 times per minute until recently Facebook could only hope to draw value from a tiny fraction of its unstructured data – information which isn’t easily quantified and put into rows and tables for computer analysis. [1]
And no one has more data about us than Facebook.
Facebook AI Multiple to various working areas. According to Kevin Lee, a technical program manager at Facebook and author of the blog.
“If you’ve logged into Facebook, it’s very likely you’ve used some type of AI system we’ve been developing”
Let’s begin the fundamental part of AI:
Deep Learning
The significant part, Deep learning can help improve things like speech recognition and object recognition, and it can play an important role in advancing AI-driven deep research in fields as diverse as physics, engineering, biology and medicine.
Deep learning enables machines to classify data independently. For example, an image analysis tool can learn to recognize cats by analyzing numerous images, without being explicitly taught what a cat looks like. It learns from the context—such as what other objects are present in cat images or what text or metadata might indicate the presence of a cat.
Deep learning allows computers to automatically recognize basic word relationships, such as associating “taxi” with “ride.” However, human input is still required to provide context. For example, a human must help the computer distinguish between “I need a taxi” and “I just got out of a taxi.” The Deep Text system combines unsupervised learning for basic word connections and supervised learning for detecting complex patterns in text.
According to Lecun’s explanations;
But language is more complicated for computers to parse than simply identifying objects in pictures.
Explore more on image recognition and image classification.
1- Deep Text
User’s comments and posts feed Facebook text data and the machines should be able to discern relationships between words on their own by breaking down the texts into individual letters and even exclamation marks.
If we take an example, Deep Text ventures to understand how people use slang and determine the specific meaning of a word that may have multiple meanings by understanding the context. Additionally, this technology will help match users with advertisers, weed out prohibited content, rank search results, and identify trending topics.
2- Translation
Facebook has users worldwide, with over half not speaking English. To break communication barriers, the Applied Machine Learning team developed an AI-based translation system that helps 800 million people monthly see translated posts. Facebook’s platform language is unique, with new expressions, informal spellings, regional variations, and emojis, reflecting human-to-human communication in real time.
In 2011, Facebook announced the launch of a new translation tool powered by Microsoft Bing. Translate that lets users select to view Page posts in their native language. Page admins can select to show only machine translated posts, or they can select to allow Facebook users to submit their own translations.
3- Photo image search
When recalling favorite memories, it’s tough to remember when an event occurred or who took the photo. With Facebook’s automatic image classifiers, imagine searching through friends’ photos based on image content instead of tags or text. This would allow you to find specific photos more easily—pretty cool!
Facebook is working on artificial intelligence (AI) feature that can understand what’s going on in your photos, which in turn would help your Newsfeed perform better. A more helpful area; this feature will allow it to answer questions about a photo, a feature aimed at helping blind people “see” images uploaded to the social network.
4- Talking pictures
Facebook is developing systems that understand images at the pixel level through image segmentation. This enables recognition of individual objects and their relationships within images. Using this technology, Facebook plans to create more immersive experiences for the visually impaired, like “talking images” that can be read with fingertips, as well as improve image search capabilities.
You can find experimental videos here.
Image by: [techcrunch]
Also, Facebook has used the Applied Machine Learning team’s platform to apply computer vision models to satellite images to create population density maps and ultimately determine where it needs to deliver broadband in the developing world. And its video-captioning efforts have proven to increase engagement, as measured in shares or likes, by 15% and boost viewing time by 40%.
5- Facebook’s bot API for the Messenger Platform
Zuckerberg had joined a “chatbot arms race” with Microsoft CEO Satya Nadella. Bots can provide anything from automated subscription content like weather and traffic updates, to customized communications like receipts, shipping notifications, and live automated messages all by interacting directly with the people who want to get them.
Three Main Capabilities (*):
Send/Receive API: This new capability includes the ability to send and receive text, images, and rich bubbles containing multiple calls-to-action. Developers can also set a welcome screen for their threads to set context as well as different controls
Generic Message Templates: We think people prefer to tap buttons and see beautiful images, rather than learn a new programming language to interact with your bot. That’s why Facebook have built structured messages with call to actions, horizontal scroll, urls, and post backs.
Welcome screen + Null state CTAs: Messenger app is giving you the real estate and the tools to customize your experience. This starts with the welcome screen. People discover our featured bots and enter the conversation. Then, they see your brand, your Messenger greeting, and a call to action to “Get Started”.
6- What is Caffe2go?
Facebook mobile app is transform video using ML in real time and bringing artsy networks to your phone with its Caffe2go feature.
Facebook’s new feature lets users record smartphone videos in the style of famous artists like Van Gogh and Picasso using style transfer. The system transforms live video to resemble iconic artwork, similar to Prisma’s photo filters. Unlike Prisma, which requires an internet connection for some filters, Facebook’s system works offline and renders live, providing a seamless experience for users.
This technology is a framework that lets Facebook engineers run deep neural nets on a smartphone, and it can have applications beyond just video interpretation. While the system’s resulting images are snazzy, they also show off the potential to bring machine learning closer to users’ daily lives.
The recent new from Facebook: We use AI to Thwart Suicide
The company is making reporting a potentially suicidal user easier, but it has also developed technology where a potential case could be dedected even before it is reported. Company has developed technology that scans users’ posts and comments left by friends, looking for trouble signs.
Image by: [bbc news]
Facebook product manager Vanessa Callison-Burch told the BBC.
“We know that speed is critical when things are urgent”
It has now developed pattern-recognition algorithms to recognize if someone is struggling, by training them with examples of the posts that have previously been flagged.
Talk of sadness and pain, for example, would be one signal. Responses from friends with phrases such as “Are you OK?” or “I’m worried about you,” would be another.
7- Gaming
Facebook AI open sourced their professional world champion beating AI software in 2018.
AI Investments
Based on Steve Toth’s October 2016 data
Mark Zuckerberg has acquired 68 companies till date. Facebook’s largest acquisition so far has been WhatsApp Messenger, which they purchased at $19 billion in February 2015. Other notable acquisitions include Instagram ($1 billion in April 2012) and Oculus Virtual Reality ($2 billion in March 2014).
The total cost of Facebook’s acquisitions to date is $ 23,124,700,000. But today Facebook’s community in the figure and daily active user amounts is rising day by day.
Some of Facebook’s smaller purchases include the domain fb.com, which it bought for $8.5 million in November 2010 and the hosting and sharing platform Drop.io, which it bought for $10 million in October 2010. Just a few million dollars each, no big deal for Facebook, whose IPO started out at $38 per share (in May 2012), which sits at roughly $93 per share in mid/late-August 2015.
Facebook is a perfect example of a company that has more buying power than entire nations. And it is leveraging that power to buy AI startups like these:
Image by: [cbinsights]
Now that you know what Facebook is up to, get ready to explore how Google is using AI!
For a bit more info on how AI is used in Facebook, you can listen to Facebook head of AI Research Yann Lecun.
You can also our list of AI tools and services:
References:
https://research.fb.com/category/facebook-ai-research-fair/
https://code.facebook.com/posts/804694409665581/powering-facebook-experiences-with-ai/
https://fortune.com/facebook-machine-learning/
[2] https://fortune.com/2016/06/01/facebook-artificial-intelligence/
(*)https://developers.facebook.com/blog/post/2016/04/12/bots-for-messenger/
Today Investments part: https://www.techwyse.com/blog/infographics/65-facebook-acquisitions-the-complete-list-infographic/
Comments
Your email address will not be published. All fields are required.