It is reported that Meta is developing a new conversational AI targeting OpenAI's GPT-4



Meta, which operates Instagram, Facebook, etc., is purchasing high-end GPUs for AI training and building a data center in order to create a conversational AI comparable to OpenAI's large-scale language model 'GPT-4' It has been reported that.

Meta is reportedly working on a new AI model to rival GPT-4 - The Verge

https://www.theverge.com/2023/9/10/23867323/meta-new-ai-model-gpt-4-openai-chatbot-google-apple



According to the Wall Street Journal, Meta has purchased a large number of high-end GPUs for AI training, NVIDIA H100 , to strengthen its AI development infrastructure. Therefore, it has been reported that Meta does not need to use Microsoft's cloud computing service ` `Azure '' for conversational AI training.

In February 2023, Meta CEO Mark Zuckerberg announced the creation of an AI development team for Instagram, WhatsApp, and Messenger, as well as AI tools across apps and services, as well as a variety of ways to engage users. He revealed that he would be developing a long-term AI 'persona' designed to assist him.

Meta CEO Mark Zuckerberg announces establishment of AI development team for Instagram, WhatsApp, and Messenger - GIGAZINE



Meta has already released large-scale language models such as `` Llama 2 '', but according to the Wall Street Journal, Meta plans to start training newer large-scale language models from early 2024. It has also been suggested that Meta aims to develop a large-scale language model comparable to OpenAI's GPT-4 to create new and more powerful conversational AI.

Meta releases commercially available large-scale language model 'Llama 2' for free, and collaborates with Microsoft and Qualcomm to optimize for smartphones and PCs - GIGAZINE



The overseas media The Verge said, ``Mr. Zuckerberg seems to be encouraging the development team to provide the AI model free of charge so that external companies can create AI tools using the AI model developed by Meta.'' states.

in Software, Posted by log1r_ut