Hear from CTO Mike Schroepfer and other Facebook leaders on Facebook's latest technology and innovation
Posted by Facebook for Developers on Wednesday, May 2, 2018
Day one of the F8 developer’s conference came with many fun and interesting new features. Much of day one had to do with Facebook’s privacy plans. This was expected given the recent troubles that the company faced. If you missed out on day one of the F8 developer’s conference, be sure to click here to check out our post on the highlights of it. If you weren’t able to tune in to day two, we’ll talk about the highlights and all you need to know!
F8 day two got a bit more technical. Throughout the talk of technical aspects, the presenters made sure to emphasize the ethics being practiced behind it all. Day one breached on talk of AI and VR, but on day two they dug a little bit deeper. The focus of day two was long-term technology investments relative to connectivity, AI (artificial intelligence), and AR/VR (augmented reality/virtual reality).
The past couple of years, there has been a lot of talk about AI, or, artificial intelligence. And as the technology is continuing to grow and being tested on various platforms, we are finally starting to touch the surface on experiencing the results for ourselves. Facebook’s AI researchers had a slight break through with a project that conducted a successful training of an image recognition system using a database of 3.5 billion photos (ones that are publically available). This system worked via the hashtags on the photos rather than human captions.
AI and Ethics
According to Facebook, the implementation of this technology will be extremely beneficial to Facebook’s researchers, as it will allow them to work more quickly. Facebook has stated that the program has already been used in order to help assist the company in identifying content that violates their policies. The talk of the AI system also played a part in Zuckerberg’s emphasis on ethics, in that it contributes to various moderation tactics such as the attempt to stop hate speech. Although, there was an admittance that certain hateful phrases, for example, “I’m going to beat you!”, and, “look at that pig!” will still need human review as they could be misconstrued by the AI system.
Another small issue is that of the potential for biases with AI advancements. To address this, Facebook has created the Fairness Flow. The Fairness Flow exists to help measure potential biases built into Facebook’s algorithm. This was likely a follow-up rebuttal to the April hearings where Facebook was scrutinized and accused of improperly moderated content.
Despite the efforts and large ethical emphasis, we think it will still take a bit of time for users to regain trust in the network when it comes to the safety and protection of their data. Regardless of any lack of trust, we have no doubt that people will continue to use the social network platform as frequently as they have before.
One of the biggest announcements at day two of the F8 conference was of the newly available version of the open source AI framework, PyTorch 1.0. This AI research and production tool aided in the image recognition project as mentioned above, as did the production tools: Caffe2 and ONNX. Facebook also expanded the open source offerings for AI developers with the implementation of the new online hub, Facebook.ai. As for the new version of PyTorch, it is said to be quicker and more efficient than the original version that was released about a year ago. The claim is that the new version will aid in accelerating AI from the research stage to the production stage.
A new tool we call DensePose generates full 3D surfaces that can be applied, in real time, to footage of human bodies in motion.
Posted by Facebook Engineering on Monday, April 30, 2018
When using PyTorch 1.0, AI developers will be able to easily transition from the highly responsive execution mode for their research to the recognizably optimizable graph execution mode for their production of said research. The program is already in use performing almost 6 billion text translations each day in light of the most commonly used languages on Facebook, 48 to be exact. The beta version will be available in the next few months.
As for ONNX (Open Neural Network Exchange), the format upgrades will allow AI engineers easier movement between models and frameworks. There will also be support features for additional available AI tools. The additions will allow ONNX to now work with the large majority of model types and will have the ability to be distributed to millions of mobile devices.
Facebook prides itself on advancing its technological capabilities in order to help keep people connected. One of the ways in which they are doing so is by creating better sharing experiences. This includes things such as the possibility of capturing more realistic-looking surroundings to the production of more highly evolved avatars.
Our teams have been working on state-of-the-art research to help computers generate lifelike avatars. This example shows two researchers testing the tool in real-time.
Posted by Facebook Engineering on Wednesday, May 2, 2018
The company’s research scientists have been able to create a prototype of a system that can generate 3D reconstructions of a physical space. This was one of the coolest features that was displayed at day two of the F8 conference. By the use of machine learning technology and “point-cloud reconstruction,” the ability to rebuild two-dimensional photos and videos into 3D environments was born.
Not only is Facebook concerned with the software involved in connecting people, but they have taken into their own hands the actual internet connectivity that must exist in order to utilize those connection tools. The company is working with countries with low internet connection in implementing fiber internet connectivity to large-scale areas, in places such as Uganda. Efforts continue as they also are working with networks in other countries to expand advanced Wi-Fi capabilities.
We worked with operators to design and build an open-access, shared backhaul network that will improve connectivity services in a region covering 3 million people in Northwest Uganda.
Posted by Facebook Engineering on Tuesday, May 1, 2018
Facebook users are finally getting their wish
On a final note, Facebook is finally granting the people their wish and adding the ability to vote up and down on posts. Facebook originally came up with an alternative solution to this request by adding the various options of voting “love, laughing, crying, and angry” on a post, but it looks like they are giving in. We think they have been avoiding this for a while, trying to avoid imitating the Reddit-like feature, but the people have been asking, and they are finally getting what they want.
What does this all mean for you and your business?
After learning about all the plans that Facebook has for their social network platforms and various AI, AR/VR software’s, it is quite plain to see that the only thing in site for their future is expansion and advancement. Knowing this, we think it safe to say that Facebook won’t be disappearing into the distance, but only continuing to evolve and grow digitally, just as we do.
This goes to show how important it is for businesses to make sure that they are up-to-date and involved in the digital world. This is the reason why so many companies are turning to a trusted and experienced digital marketing agency. Get on board with an agency that will use their knowledge and capabilities to develop a unique and successful digital marketing plan for you and your company. Here at Tech-Critic, we’ve got you covered. Contact us today and let us show you the growth that your business is capable of.