Meta is building an AI model as powerful as GPT-4, report says

• Wall Street Journal report says Meta is developing an AI model designed to compete with GPT-4.
• It's expected to be much more powerful than Llama 2, the open-source AI that Meta recently released.
• The move signals that the AI arms race is heating up as Meta attempts to catch up with rivals.

Meta is developing an AI that can go toe-to-toe with OpenAI's most advanced model.

The Wall Street Journal reported the news, citing unnamed sources, who said the tech giant was working on a new AI system that aimed to be as powerful as GPT-4, the latest version of OpenAI's large language model.

The news comes as the company seeks to catch up with its rivals, with Big Tech companies scrambling to develop their own AI models.

The Journal reported that Meta could release the model, which is expected to be significantly more powerful than the company's recently released Llama 2 AI, at some point next year.

The AI will reportedly be primarily aimed at businesses as a tool for producing text and analysis. This is similar to ChatGPT enterprise, the business tier of OpenAI's highly successful chatbot, which launched last month.

In a departure from similar models launched by other big tech companies, such as ChatGPT and Google's Bard, unnamed sources told the Journal that Meta boss Mark Zuckerberg was pushing for the upcoming AI to be open source.

Open-source models are free to access and allow users to modify and use the AI however they want, although Meta's models have restrictions and users must follow the company's terms of service.

As flexible as open-source models are, there are fears they could be used for malicious purposes, such as spreading misinformation or infringing on copyright. The Journal reported that Meta's lawyers had raised concerns about potential misuses of the company's AI model.

Meta has increasingly shifted its focus towards AI as interest in the burgeoning technology has skyrocketed, with Zuckerberg describing it as a key theme for the company in a recent earnings call.

The company faces a battle to catch up to the likes of OpenAI and Microsoft, however, with internal memos from last year obtained by Reuters admitting that the company had a "significant gap" in AI research and investment.

Since then, the Facebook owner has been rapidly purchasing the precious H100 chips that are needed to train large language models and is expected to rank 2nd in the world next year in regards to its number of H100s, researchers said.

Meta has bet on open-sourcing its AI models to cut the lead built up by its rivals. The company announced it would open source its Llama 2 model earlier this year, making it free for research and commercial use.

Meta did not immediately respond to Insider's request for comment made outside of normal working hours.