Archive — US — techcrunch
Zuckerberg signals Meta won’t open source all of its ‘superintelligence’ AImodels
Meta CEO Mark Zuckerberg announced a shift in strategy regarding the release of AI models, signaling that not all 'superintelligence' models will be open-sourced. He emphasized the potential risks associated with superintelligence while reaffirming a commitment to personal AI, particularly through AR and VR devices. This marks a departure from the company's historical focus on open-source AI amid increased competitive pressure.
Key points
- Zuckerberg emphasizes the concept of 'personal superintelligence' to help users achieve their goals.
- Meta is reassessing its commitment to open-source AI, citing new safety concerns and risks associated with advanced models.
- Historically, open models were Meta's key advantage against competitors, but future models may prioritize closed models for safety and monetization.
- Meta’s shift follows its substantial investment in AI, including acquiring Scale AI and redesigning its AI development strategy.
- Zuckerberg believes personal computing devices, such as AR glasses, will be central to delivering personalized AI experiences.
Why it matters
- Meta’s stance reflects a broader industry trend where companies like OpenAI and Google favor closed models for monetization and controlling advanced technologies.
- In 2024, Meta's competitive pressure increased as it sought to match OpenAI’s advancements with its Llama models.
- As of 2025, the shift to advanced closed models suggests a growing industry emphasis on safety and proprietary technology in AI development.
Zuckerberg's announcement signals a significant strategic shift for Meta as it navigates the complexities of AI development, balancing the benefits of open-source access with the imperative of safety and competitive advantage.