Meta Aims to Rival ChatGPT with New AI System Set for 2024

Meta Races to Develop AI Rival to ChatGPT, Targeting 2024 Launch

Facebook parent Meta is quietly developing a new artificial intelligence system that it hopes will be on par with ChatGPT, the viral chatbot created by Microsoft-backed startup OpenAI. According to a recent Wall Street Journal report, Meta's new AI model could be ready as soon as 2024 and would aim to produce sophisticated text, analysis and other outputs that can be used by companies to build services and products.

The planned system is part of an accelerated push by Meta and CEO Mark Zuckerberg to establish the company as a major force in AI, after falling behind rivals like OpenAI and Google in recent years. The competition to develop and deploy the most advanced AI has intensified sharply in just the last year, sparking divergent views on business models, ethics and regulation.

Meta's new model will be spearheaded by a group Zuckerberg formed earlier this year specifically to focus on generative AI — technology that can generate human-like, conversational text and images. The aim is for Meta's model to be several times more powerful than LLaMA, the natural language AI system Meta unveiled just two months ago.

According to the WSJ's reporting, Meta plans to start training the new model in early 2024, building up the data centers and acquiring the advanced chips needed to handle the computing demands of training such a powerful system. While LLaMA was developed in partnership with Microsoft and runs on Microsoft's Azure cloud platform, Meta intends to train this latest model completely in-house.

The push for an ultra-advanced natural language AI model reflects Zuckerberg's ambition to reassert Meta's position at the forefront of artificial intelligence research. The company was caught off guard by the explosion in popularity of ChatGPT and other AI chatbots in recent months. After pouring resources into developing VR and metaverse technologies, Meta appeared to miss how quickly text-based AI could go mainstream.

Now Zuckerberg is pressing Meta's AI team to close the gap with OpenAI and other rivals. Meta is hiring more engineers and researchers with AI expertise and forming partnerships with academics to advance its AI capabilities. Developing a more powerful successor to LLaMA is seen as a way for Meta to showcase its progress and engineering chops.

As with LLaMA, Meta plans to make the new model open source and freely available. This aligns with Zuckerberg's vision for Meta's role in AI — enabling any developer or company to build services and products powered by its systems. Meta would benefit in turn from expanded use of its AI across the internet.

This approach differs from rivals like Google, which has so far kept its most advanced AI models closed to outside developers. It also contrasts with OpenAI's decision to commercialize access to ChatGPT through a subscription pricing model. Meta is betting that AI generative models can be monetized indirectly, fueling engagement and usage of its apps and services.

But a more powerful AI model also presents risks if released into the open. Researchers have raised concerns about potential biases and harm from large language models, if they lack sufficient guardrails. Meta will need to implement stringent controls and safeguards to prevent misuse and abuse of such a powerful system.

The race to develop ever-more capable AI has entered a new phase of urgency and competition. With its new model slated to come online in 2024, Meta aims to prove it still has the resources and talent to push the boundaries of artificial intelligence alongside OpenAI, Google and others. But the company will also need to proactively address ethical dilemmas and safety issues to ensure its technology has a positive impact.