Mark Zuckerberg is steering Meta towards a future defined by "personal superintelligence," shifting away from solely focusing on open-source AI models. This new AI strategy emphasizes personalized AI assistants designed to understand individual user's emotions, anticipate their needs, and adapt to their routines. Meta envisions these AI systems as being deeply integrated into users' daily lives, accessible across platforms like WhatsApp, Messenger, Instagram, and even within AR/VR experiences on Quest headsets.
The core of this strategy revolves around creating AI that empowers individuals by helping them achieve their goals, enhance their creativity, improve social connections, and foster personal growth. Zuckerberg believes that personal superintelligence will have a more meaningful impact on people's lives than AI focused solely on automating workflows. Meta AI can provide tailored answers, advice, and inspiration, while also offering fun ways to edit videos. The Meta AI app remembers user preferences and interests to provide relevant and helpful responses.
Meta is making substantial investments in AI infrastructure and talent to support this vision. This includes significant capital expenditures, projected to be between $66 billion and $72 billion in 2025, and similarly large investments are expected for 2026. These investments are geared towards expanding compute power, model training capacity, and scaling systems. Meta has also been actively recruiting AI experts, including key hires from companies like OpenAI and Apple. Alexandr Wang, formerly CEO of Scale AI, has joined Meta as Chief AI Officer.
Zuckerberg acknowledges the "novel safety concerns" associated with superintelligence. Meta is emphasizing a responsible approach to AI development, balancing innovation with ethical deployment. Zuckerberg stated that Meta needs to be rigorous about mitigating risks and careful about what the company chooses to open source. This stance reflects a more cautious approach to openness compared to Meta's previous emphasis on open-source AI. Meta is implementing safety measures to address context-specific risks, with mitigations layered across different intervention points.
Meta is also incorporating privacy safeguards into its AI features to protect user information. Users have control over their data and can delete information shared in chats with AI. Meta does not use the content of private messages with friends and family to train its AIs. The company provides transparency resources, including a Generative AI Privacy Guide, to help users understand how its AI models are built and how the features work.
Meta is also using AI to automate a significant portion of its internal privacy and safety checks. AI tools will handle most updates, including changes to core algorithms, safety tools, and content-sharing policies, with human experts focusing on "novel or complex" cases. This shift aims to improve efficiency and speed up product rollouts, although some concerns have been raised about the potential for increased risks due to less human oversight.
Smart glasses are expected to play a key role in Meta's AI strategy. Zuckerberg envisions these glasses as a gateway to personal superintelligence, enabling AI to "see what we see, hear what we hear," and provide users with a cognitive edge. Meta has been developing smart glasses in partnership with EssilorLuxottica, with recent models powered by Meta's Llama AI models. These glasses can respond to questions, take videos, and make hands-free calls and messages.
