Meta, the parent company of popular social media platforms Facebook and Instagram, is reportedly working on a cutting-edge artificial intelligence (AI) model that aims to rival OpenAI’s most advanced offering, according to exclusive reports from The Wall Street Journal.

Sources familiar with the matter have revealed that Meta’s objective is to create an AI model significantly more powerful than its Llama 2 model, which was launched earlier this year. The company is setting its sights on developing an open-source system that would enable other companies to harness AI tools for generating high-level text, analysis, and various other outputs.

Meta has been diligently establishing the necessary data centers for this ambitious project while securing additional Nvidia H100 semiconductor chips, known for being the most potent and sought-after chips currently available in the market.

Llama, Meta’s previous AI model, was trained on a staggering 70 billion parameters. While OpenAI has not disclosed the parameters for its GPT-4 model, estimates suggest they may reach around 1.5 trillion.

Insiders reveal that Meta expects to commence training for its large language model (LLM) in early 2024, with a target release date sometime in the following year. It is likely to debut after Google’s forthcoming LLM, known as Gemini.

Microsoft, a primary supporter of OpenAI, has also collaborated with Meta to make Llama 2 accessible on Azure, its cloud computing platform. However, Meta intends to train its upcoming model using its proprietary infrastructure.

This development unfolds in the midst of a global race among major tech companies and governments to create, deploy, and control high-level AI systems. For instance, the United Kingdom government has allocated $130 million for acquiring high-powered chips to build advanced AI systems. Meanwhile, in China, new AI legislation has taken effect, with numerous AI models being introduced, underlining the growing significance of AI on the global stage.

By Impact Lab