Aitarget shares key lessons from Day 2 of the F8 2018 conference
Aitarget experts joined developers from around the world in San Jose on 1 and 2 May for F8, Facebook’s annual two-day conference. At the conference, developers and businesses explored the future of technology during a full slate of 50 sessions plus demonstrations.
We’ve addressed founder and CEO Mark Zuckerberg’s overall comments about the future focus of Facebook, his key points, and highlights of Day 1 of F8 2018 in Part 1.
In Part 2 we’re looking at the highlights of the conference’s second day.
On day 2 the focus shifted to product innovation and major long-term investments being made in artificial intelligence (AI), augmented and virtual reality (AR & VR), and connectivity.
Learning to learn
Facebook’s Chief Technology Officer Mike Schroepfer spoke about viewing AI as a “foundational technology”, and their deep investments made in scientist-directed research.
Examples were given of AI tools being used to automatically deal with harmful content: 300,000 bots and assistants proactively remove millions of fake accounts every day, and 99% of Al Qaeda and Isis content was zealously removed by AI tools in the first quarter.
In order to streamline the complicated and intensive process of taking AI development from research to production, Facebook has introduced open source AI framework, PyTorch 1.0. The goal is to “provide a fast, seamless path from research prototyping to production deployment for a broad range of AI projects”, and will be available in beta within months.
Looking ahead, key areas for AI include:
- computer vision;
- natural language processing; and
- building systems that can reason and learn.
The introduction of PyTorch 1.0 will allow developers to use computer vision advances such as DensePose, which puts polygonal mesh overlays over bodies moving through a scene.
With recent breakthrough research, Facebook managed to train an image recognition system to match a set of 3.5 billion public photos with hashtags and score 85.4% accuracy.
Such advances allow researchers to scale their work quickly, and Facebook has leveraged this work in production to improve its ability to identify content that violates its policies.
AI will be used to understand nuance in text speech. PyTorch 1.0 is being used to perform nearly 6 billion text translations per day for 48 commonly used languages on Facebook.
AI technology will also train robots in real life, and teach computers.
Re-rendering the real world
As part of its efforts to improve connections between people around the world, Facebook has been developing AR and VR to create better shared experiences.
Facebook has developed a prototype system for 3D reconstructions of physical space.
An ability to re-render the real world will be a real game-changer, both in terms of the surroundings and the avatars moving through them. We can immerse ourselves in 3D.
AR can be critical in building projects, and state-of-the-art research has been used to develop lip synch systems in different languages, and photo-realistic avatars.
Source: Facebook Engineering
Reaching the world
At F8 2018 it was noted that while many innovative technological leaps are being made, they’ll only be relevant and accessible for people who have a strong internet connection.
There is more than half of the world’s population, 3.8 billion people, who currently do not have access to the Internet. Facebook is working hard to increase global connectivity.
It is developing next-generation technologies and a number of projects to lower costs, reach the unconnected, and improve internet access in many countries around the world.
Facebook may have had a rough couple of months, with some scandals and setbacks, but the mood at F8 2018 was positive and enthusiastic. The world’s biggest social media network looks set to keep leading the way in the connection between people and technology.