Potential Shutdown of Future Meta Models Due to Superintelligence Concerns
Meta Shifts Focus to Developing Closed AI Model: Introducing "Personal Superintelligence"
In a significant move, Meta, the tech giant behind Facebook, has announced a strategic adjustment in its approach to Artificial Intelligence (AI) development. On Wednesday, CEO Mark Zuckerberg unveiled his vision for "personal superintelligence," marking a departure from competitors who are focused on "automating all valuable work."
The company's new focus has led to the establishment of the "Meta Superintelligence Labs," a specialized division within Meta. This unit, consisting of four teams, will focus on large models, AI products, infrastructure, and fundamental research to accelerate innovation in AI.
In June 2025, Meta embarked on a focused effort to develop artificial general intelligence (AGI), investing $14.3 billion in Scale AI and acquiring Scale's founder and CEO. This investment is a testament to Meta's commitment to advancements in AI technology.
Previously, Meta's Llama family of open models was a key differentiator for the company. However, with the shift towards developing a closed model, Meta may no longer release everything it develops historically. The spokesperson affirmed the company's ongoing commitment to open-source AI, but also indicated that Meta anticipates training closed-source models in the future.
Zuckerberg's vision for "personal superintelligence" represents AI designed to help individuals achieve their unique goals. This AI will be delivered through Meta's own products, particularly augmented reality glasses and virtual reality headsets.
However, the development of such advanced AI will raise novel safety concerns, according to Zuckerberg. He emphasized the importance of sharing the benefits of superintelligence with the world as broadly as possible, while also being rigorous about mitigating these risks. Meta will need to be careful about what they choose to open source.
Meta's competitors maintain closed models primarily for economic reasons, as they believe that closed models provide greater control over the monetization of their products. In contrast, Zuckerberg has emphasized the importance of open-source AI, acknowledging the possibility of restricting access to certain models in the past.
Testing on Meta's latest Llama model, known as Behemoth, has been paused. The company has also built new data centers to support its AI development initiatives. Despite the shift towards a closed model, Meta plans to continue releasing leading open source models, just as it did with its previous Llama models, which were a key differentiator for the company.
The announcement represents a potential departure from Meta's previous strategy, but it is a strategic move aimed at surpassing the capabilities of its competitors, particularly OpenAI's GPT-4 model, during the development of Llama 3. If a qualitative change occurs in a model's capabilities, and it's not responsible to open source it, Meta may choose not to do so, according to Zuckerberg.
In a 2024 letter, Zuckerberg stated that future Llama models would become the most advanced in the industry. With this new focus on developing a closed model, Meta aims to deliver on this promise and establish itself as a leader in the field of AI. Superintelligence will indeed raise novel safety concerns, but Meta is prepared to face these challenges head-on.
