From Metaverse to AI: Why Meta's Focus is Shifting Towards Artificial Intelligence for a More Connected Future

Facebook, now known as Meta, has been one of the leading companies in developing and promoting the concept of the metaverse, a virtual environment where people can interact with each other and digital content in immersive and realistic ways. However, in recent months, Meta has also shown a renewed focus on artificial intelligence (AI), the technology that powers many of the features and applications that make the metaverse possible. Facebook has been investing heavily in generative AI.

Generative AI is a sub-field of machine learning that involves generating new data or content based on a given set of input data. This can include generating text, images, code, or any other type of data. Generative AI systems use generative models such as large language models to produce data based on the training data set that was used to create them.

Facebook has been developing and deploying generative AI models for various purposes, such as improving its content moderation, personalizing its ads and recommendations, enhancing its user experience, and creating new products and services. Recently Facebook announced its plan to revolutionize advertising using generative AI. Some other examples of Facebook's generative AI projects are:

Wit.ai: A natural language processing platform that allows developers to build chatbots and voice assistants using generative AI models. Wit.ai can understand user queries and generate appropriate responses based on the context and intent of the conversation.

BlenderBot: A chatbot that can engage in open-domain dialogue with humans using generative AI models. BlenderBot can handle multiple topics, express emotions and personality, and use knowledge from external sources to enrich the conversation.

DINO: A computer vision model that can generate high-resolution images from low-resolution inputs using generative AI models. DINO can also perform image segmentation, object detection, and semantic labeling tasks.

SEER: A self-supervised computer vision model that can learn from unlabeled images using generative AI models. SEER can recognize objects and scenes in images without any human annotation or supervision.

Barlow Twins: A self-supervised machine learning model that can learn from unlabeled data using generative AI models. Barlow Twins can extract useful features from data without any predefined task or objective.

These are just some of the examples of how Facebook is using generative AI to enhance its core business and explore new opportunities. According to Facebook's chief AI scientist Yann LeCun, generative AI is one of the key technologies that will enable the metaverse, as it will allow users to create and share their own content in realistic and creative ways.

In February 2022, Meta hosted an event called "Meta AI: Inside the Lab", where it showcased some of the latest breakthroughs and innovations in AI research and development. Some of the highlights included:

BuilderBot, a tool that allows users to create or import objects and scenes into a virtual world using voice commands. For example, a user can say "show me a castle" or "add a dragon" and see their request fulfilled in real time. BuilderBot uses natural language understanding, computer vision, and generative modeling to understand and execute user commands.

Inclusive translation systems, which aim to improve the quality and diversity of language translation for different dialects, accents, and speech styles. Meta's AI researchers have developed techniques to leverage large-scale data and self-supervised learning to train more robust and accurate translation models that can handle various forms of speech and text input.

A new AI model for chatting with virtual assistants, which can generate more natural and engaging responses than existing models. The model, called BlenderBot 2.0, can remember previous conversations, search for relevant information on the web, and personalize its responses based on user preferences and feedback.

Meta's AI efforts are not only driven by its vision for the metaverse, but also by its mission to connect people to what matters and to help keep communities safe. Meta's AI team works on various projects and challenges that span across different domains and applications, such as:

Computer vision, which enables machines to understand and manipulate visual information, such as images, videos, and 3D scenes. Meta's computer vision research covers topics such as face recognition, object detection, scene understanding, video understanding, augmented reality, virtual reality, and more.

Natural language processing, which enables machines to understand and generate natural language, such as speech and text. Meta's natural language processing research covers topics such as dialogue systems, machine translation, natural language understanding, natural language generation, question answering, summarization, sentiment analysis, and more.

Machine learning systems, which enable machines to learn from data and improve their performance over time. Meta's machine learning systems research covers topics such as deep learning, reinforcement learning, self-supervised learning, federated learning, optimization, distributed systems, hardware acceleration, and more.

Meta's investment in AI is not only reflected in its research output, but also in its infrastructure and resources. In November 2021, Meta announced that it had designed and built the AI Research SuperCluster (RSC), which it claimed was among the fastest AI supercomputers running at that time and would be the fastest AI supercomputer in the world when it was fully built out in mid-2022. The RSC is designed to support large-scale AI training and experimentation for Meta's researchers and engineers.

Meta's shift from metaverse to AI does not mean that it has abandoned its original goal of creating a virtual world where people can connect and create. Rather, it means that it has recognized that AI is a key enabler and driver for the metaverse, and that it needs to invest heavily in AI research and development to make the metaverse a reality.