Meta, the company formerly known as Facebook, is working on a new artificial intelligence model that is more powerful than ChatGPT 4, the latest conversational AI system from Microsoft and OpenAI. Meta does not want to rely on Microsoft for training its new AI model, and is developing its own infrastructure and data sources.
About ChatGPT 4
ChatGPT 4 was released in July 2023, and claims to have improved performance and coherence over its predecessor, ChatGPT 3. It has 175 billion parameters, which is the same as ChatGPT 3, but uses a more efficient architecture and training method.
Meta’s Next-Generation AI Model
Meta, however, is not satisfied with ChatGPT 4, and wants to create its own AI model that can surpass it in terms of quality and diversity. Meta’s CEO Mark Zuckerberg announced in August 2023 that Meta is working on a new AI model that will have 250 billion parameters (which is way more than ChatGPT 4), and will be able to handle multiple languages, domains, and tasks.
Zuckerberg said that Meta’s new AI model will be able to power various applications and services in the metaverse, the virtual reality platform that Meta is building.
Meta does not want to rely on Microsoft Azure Cloud platform for training its new AI model, because it believes that Microsoft’s data and infrastructure are not suitable for Meta’s vision and goals.
A report also reveals that Meta is buying more Nvidia H100 AI-training chips (or Graphics Card) to improve their AI infrastructure so that they can train their new Generative AI model inside their own house.
Meta expects to launch its new AI model by the end of 2023 or early 2024, and hopes to challenge Microsoft’s dominance (Chat GPT 4) in the conversational AI field. Meta also plans to make its new model accessible to developers and researchers, who can use it to create innovative and immersive experiences in the metaverse.
Hi there, I’m Anindya Naskar, Data Science Engineer. I created this website to show you what I believe is the best possible way to get your start in the field of Data Science.