Meta is reportedly developing an open-source AI system that is expected to be as powerful and advanced as OpenAI’s GPT-4 model, according to recent reporting from The Wall Street Journal. It will likely be trained using Nvidia’s H100 chips and could potentially start training in 2024. The company is currently building up the data centers necessary for the job and acquiring more H100s.
This new AI system will allow companies to develop services that produce sophisticated outputs including text and analysis. This development follows other recent releases from Meta, including SeamlessM4T, an AI model that works to facilitate communication across languages, and Code Llama, a large language model that uses text prompts to generate and discuss code.
