The company's substantial investment—tens of billions of dollars—in AI, augmented reality, and other Metaverse technologies projects a record-high capital expenditure forecast for 2024, potentially reaching $37 to $40 billion.
Furthermore, Meta's open-source approach to its AI models, freely available to developers, is strategically designed to accelerate the development of enhanced tools for its platforms.
The company projects that this research will facilitate the creation of fully embodied agents within the Metaverse, resulting in more realistic non-player characters, wider access to character animation technology, and innovative immersive experiences.
Meta Motivo aims to resolve body control issues often encountered with digital avatars, allowing them to execute movements in a more realistic, human-like fashion, according to the company.
Furthermore, Meta is introducing a new training model for language processing called the Large Concept Model (LCM), which seeks to "decouple reasoning from language representation."
"The LCM represents a significant shift from traditional LLMs. Instead of predicting the next token, the LCM is trained to anticipate the next concept or high-level idea, represented by a complete sentence in a multimodal and multilingual embedding space," the company explained.
Among other AI tools released by Meta is the Video Seal, which embeds an invisible watermark into videos, making it undetectable to the naked eye yet traceable.