According to a Wall Street Journal exclusive, Meta, Facebook and Instagram’s parent company, is developing a more powerful and open-source artificial intelligence (AI) that will rival OpenAI’s most advanced model.
WSJ reported that individuals who know the situation stated that Meta intends for its new AI model to be “several times” more potent than the Llama 2 model released this year.
According to WSJ sources, Meta’s current plan for the new system will be open-source, allowing other companies to develop AI tools that generate high-level text, analysis, and output forms.
In addition to constructing the data centers required to create such a high-level system, the company has been acquiring more of Nvidia’s H100 semiconductor chips – the most powerful and coveted chips on the market today.
Llama was trained on 70 billion parameters, and while OpenAI has not disclosed GPT-4’s parameters, they are estimated to be approximately 1.5 trillion.
According to the sources, Meta anticipates that training for the large language model (LLM) will commence in early 2024 and be completed by the end of the year. It will be released after Google’s anticipated LLM Gemini.
Microsoft is a significant contributor to OpenAI and collaborated with Meta to make Llama 2 accessible on Azure, its cloud-computing platform. However, according to sources, Meta intends to educate its future model on its infrastructure.
This development arrives when major tech companies and governments compete to develop, deploy, and control advanced AI systems.
The United Kingdom government recently announced that it intends to spend $130 million on high-powered chips to create AI systems.
China’s new legislation on artificial intelligence recently went into effect globally. Since then, according to the CEO of Baidu, a significant Chinese technology company, over 70 AI models have been released in the country.