Meta is building data centers to create an open-source artificial intelligence model that could launch as early as next year.
Meta put its ambitious metaverse plans on ice this year, and now we know why. The Big Tech giant is developing an artificial intelligence system that could rival the model that OpenAI built to power ChatGPT, the Wall Street Journal reported.
This new development comes after Meta announced the release of Llama 2 alongside Microsoft in July. Llama 2 is the tech giant’s open-source large language AI model, free for research and commercial use. The Facebook and Instagram parent company wants to build something even more powerful than that. WSJ reports that Meta is developing an open-source AI system that could launch as early as next year. The new tech would allow companies to build AI tools that could produce various content, such as high-level text and analyses.
WSJ got the scoop from a couple of sources close to Meta, who also said the tech firm is building data centers that can handle the significant amount of computing power it will need to create a new AI system. The data centers will be equipped with Nvidia’s H100 semiconductor chips, AI-enabled chips tailored for generative AI outputs due to their quick processing power. Training on Meta’s new large language model could start in early 2024, ahead of its release. The AI system was built by a special core team that Meta CEO Mark Zuckerberg put together earlier this year.
Despite partnering with Microsoft to develop Llama 2 — which seems a bit conflicting since Microsoft is also heavily invested in OpenAI — Meta plans to develop and push out the new AI system on its own. Meta is hoping to build a system as powerful as OpenAI’s ChatGPT-4, but by the time it releases next year, we can guess that OpenAI will drop a few more new models. This doesn’t necessarily mean Meta is behind the wave, but it definitely has some competition in the generative AI arena.