Meta's GenAI Research Investments and Growth Vectors in the Past Year

Contributing to Open Source AI, Advancing Audio, Visual and 3D Modalities and Ensuring Safety and Alignment

Demand for Generative AI features is growing, as Consumers and Businesses realize the benefits they can bring to their online lives. As a company with products that touch billions of users worldwide every day, Meta is heavily invested in meeting the demand and developing performant ML capabilities and engaging AI-powered features across its entire user base, across Consumer and Business audiences.

Through its social, messaging and ads products, spanning multiple platforms including Facebook, Instagram, and WhatsApp, Meta has the potential to fundamentally transform and improve the way that Consumers create and share content with each other, and the way that Businesses create and push their ad content across all modalities, from text to 3D.

AI-powered features are not just an opportunity for Meta, but a business imperative. As a primarily Consumer-focused company with an ad business model, Meta’s success depends on its ability to deliver generative AI features that grow and maintain engagement for its products across all of its touchpoints.

And Meta is well on its way to do so. Last month, at the Connect event, the company announced that more than 400 million people are already using Meta AI monthly, with 185 million doing so each week.

This post is a look at the areas that Meta is investing in when it comes to AI research and how it engages with the open source developer ecosystem in order to contribute to technical and alignment advances.

Meta’s GenAI investment strategy includes three major directions:

• Contributing to open source AI to promote exploration, experimentation and testing of its models and frameworks for as many use cases as possible

• Advancing all modalities, from text, to audio, visual and 3D generation, as well as multi-modality support in a single model,

• Improving AI alignment with human intent and guardrails for safe content, as well as AI model attack protection

Read below about Meta’s model and framework catalogue and how its releases this past year fit into these strategic directions.