All-In: Meta abandons open-source AI for future superintelligent systems

Meta CEO Mark Zuckerberg has announced that the company’s future superintelligent AI will not be open source, marking a significant reversal from his previous commitment to open AI development. This shift represents a major policy change for one of the tech industry’s most vocal advocates for open-source AI, potentially reshaping how the most advanced AI systems are developed and distributed.

What you should know: Zuckerberg published a manifesto Wednesday declaring that “developing superintelligence is now in sight” but cited safety concerns as the reason for abandoning open-source principles for future advanced AI.

• “We believe the benefits of superintelligence should be shared with the world as broadly as possible. That said, superintelligence will raise novel safety concerns,” Zuckerberg wrote.
• “We’ll need to be rigorous about mitigating these risks and careful about what we choose to open source,” he added.

The big picture: This represents a complete about-face from Zuckerberg’s position just one year ago, when he argued that open-source AI would be safer than closed alternatives.

• In 2024, Zuckerberg wrote: “There is an ongoing debate about the safety of open source AI models, and my view is that open source AI will be safer than the alternatives.”
• The reversal comes as Meta observes its AI systems beginning to show signs of self-improvement, though Zuckerberg notes “the improvement is slow for now, but undeniable.”

Why this matters: Meta’s Llama models have been among the most prominent examples of relatively open AI development, making this policy shift particularly significant for the broader AI ecosystem.

• Critics argue that Llama was never truly open source to begin with, as it requires users to sign Meta’s licensing agreement and imposes various restrictions.
• The Open Source Initiative, which stewards the open-source definition, has noted that Llama’s restrictions don’t align with true open-source principles.

Key restrictions already in place: Meta’s current Llama licensing includes several limitations that deviate from traditional open-source practices.

• Users must obtain additional licensing if their Llama implementation exceeds “700 million monthly active users in a calendar month.”
• The agreement prohibits deploying Llama in critical infrastructure or for regulated controlled substances.
• Meta has tightly restricted Llama’s use in the European Union and hasn’t released the training data used to build the models.

Competitive considerations: The decision to keep superintelligent AI proprietary also reflects Meta’s broader strategy to maintain competitive advantages in the AI race.

• Zuckerberg has been offering substantial deals to recruit top AI researchers from rivals like OpenAI and Apple.
• Meta is investing billions in new data centers designed to develop and run future AI systems.
• In his manifesto, Zuckerberg stated his intent to deliver AI capabilities to “billions of people across our products.”