When we think of artificial intelligence (AI), we often think of robots, self-driving cars or virtual assistants. But AI spans across a wide range of technologies that go beyond the futuristic devices we read about and see on TV. In fact, AI powers many of the software and hardware products at Meta. From personalized News Feeds and automatic language translations to the integrity work that keeps Meta technologies safe, people rely on these AI innovations to stay connected.
AI has come a long way, but there’s a lot more to do when it comes to advancing the field and developing systems with human level intelligence. This is the focus of Meta’s Applied AI Research team. AI researchers apply cutting-edge algorithms to a wide range of media—from photos and video to audio—to understand and address challenges across Meta technologies. By actively collaborating with the broader research community, the team is accelerating research breakthroughs to develop state-of-the-art AI that will have a positive impact on people around the world.
The Applied AI Research team works within 12 areas of Engineering: Speech & Audio, Multimodal Video Understanding, Conversational AI, Language & Translation Technologies, Personalization, Experiences, Computer Vision, Reality Labs, XR People, AI Infra, Integrity and Responsible AI.
Researchers and engineers come from diverse professional and academic backgrounds, and they work from offices across the US and EMEA, including Menlo Park, Seattle, Redmond, Bellevue, New York, Pittsburgh, Boston, London, Zurich and Tel Aviv. Read on to learn more about five of the research areas within Applied AI, the work teams are doing and the impact they have on the products we use every day.
Computer Vision: Developing models to better understand visual content
Within Applied AI Research, the Computer Vision (CV) teams are focused on understanding visual content. Researchers build models that are used in relevance and ranking applications, business applications that help small businesses sell products easily at scale, new augmented and virtual reality experiences, and integrity solutions that minimize deep fakes and misinformation.
Whether it’s empowering more immersive interactions with Portal’s Smart Camera, using AI to proactively detect and remove policy-violating photos, or better connecting people to the content they care about, research scientists and engineers on the CV team are pushing the frontier of what’s possible with computer vision for Meta products.
Meta built and deployed GrokNet, a universal computer vision system designed for shopping. It can identify fine-grained product attributes across billions of photos.
Language & translation technologies (LATTE): Connecting people no matter their language preference
The mission of the Language and Translation Technologies Team (LATTE) is to understand all languages. They envision a world where everyone, no matter which language they speak or understand, can connect with others and enjoy a positive Facebook experience.
Researchers and engineers on LATTE work on several aspects of language understanding, including machine translation, pretrained representations, cross-lingual understanding, semantic parsing, question answering and information extraction. They work cross-functionally with product teams across Meta to build Natural Language Processing (NLP) solutions that address problems like identifying hate speech, developing information ranking and providing recommendations to people across our apps and services.
Diagram from Meta’s XLM paper showing the schematic of the cross-lingual language model.
Personalization: Providing value by connecting people to what’s meaningful for them
The Personalization team builds cutting-edge AI to understand and connect people to what’s most meaningful for them, whether it’s communities, products, articles or other people. They work on problems related to ranking, recommendations, sequence modeling, graph learning, representation learning, reinforcement learning, large-scale training and machine learning. The Personalization team powers many popular Meta technologies that billions of people use and love, from Instagram’s Explore page to Facebook’s News Feed and recommendations.
Unlike product teams, Personalization researchers take a long-term, research-based approach to creating technology breakthroughs that foster positive experiences for people around the world.
Team members gather together to take a break, connect and converse in one of the many common areas at Meta offices.
AI Infra: Enabling and scaling artificial intelligence and its use at Meta
The AI Infra team is the backbone of the Applied AI research team. AI Infra team members focus on optimization, performance, efficiency, scalability and interpretability of Meta’s infrastructure. Their work powers Meta’s personalization and content understanding technologies to help bring the world closer together. They do this by building server and mobile frameworks, pipelines and platforms that encompass the complete lifecycle of machine learning research and production.
The team also works on election integrity and fighting misinformation, while making sure Meta’s services protect the privacy and security of the people who use our products.
Meta’s data center in Altoona, IA.
Reality Labs: Bridging distances and building new ways for people to connect
Reality Labs (RL) is exploring new ways to break down barriers and bridge distances for people around the world to connect through immersive experiences. Developing world-class hardware and software, Reality Labs is developing the technologies needed to enable virtual ready headsets and augmented reality glasses that will transform the way we interact with our surroundings. The team has global departments dedicated to AR/VR research, computer vision, haptics and social interaction, and team members are committed to driving the state of the industry forward through relentless innovation.
Meta Quest 2 is the latest VR headset from Meta.
The Applied AI Research team is working at the intersection of research and innovation to develop models that are poised to pave a path for the future of artificial intelligence. Interested in joining the team?
This post, originally published on October 1, 2020, was updated on November 2, 2022, to reflect our shift to Meta and new details about team members, roles and responsibilities.