Critiquing Zuckerberg’s Vision for Open-Source Superintelligence

The Philosophy of Decentralized AI

Mark Zuckerberg’s recent essay, The Future is for Everyone, posits that the path to safe and prosperous AI lies in individual empowerment rather than the centralized control favored by labs like Anthropic. The core argument is that by open-sourcing superintelligence, we avoid the dangers of a "white-collar bloodbath" and prevent the concentration of power in the hands of a few institutions. Zuckerberg frames this as a historical inevitability: technological shifts consistently lead to greater abundance and prosperity, and AI should be treated as a tool for invention rather than mere automation.

The Compute-Capital Paradox

While the vision of democratized intelligence is appealing, the argument contains a fundamental structural flaw: the distinction between the model layer and the compute layer. Zuckerberg suggests that a "dynamic auction mechanism" will ensure everyone gets the lowest price for compute. However, compute is a finite, physical resource. In a world where intelligence is commoditized, the competitive advantage shifts entirely to whoever can afford the most compute. This creates a "permanent underclass" scenario where socioeconomic status at the onset of AGI determines one's long-term ability to leverage that intelligence. If compute is the bottleneck, the "level playing field" is a myth; those with the most capital will always out-compete those with less, regardless of how open the underlying models are.

The Fallacy of the "Super-Lawyer" Thought Experiment

Zuckerberg uses a thought experiment involving super-intelligent lawyers to argue that universal access creates fairness. He claims that if everyone has a super-intelligent lawyer, litigation becomes more efficient and just. This ignores the reality of resource allocation: even if both sides have access to the same model, the side with more capital can afford more compute, allowing their model to "think" longer, perform deeper research, and iterate faster. Thus, the imbalance remains, merely shifted from human skill to machine-compute capacity.

Invention vs. Automation

Zuckerberg correctly identifies that AI will likely create new categories of work—such as one-person product studios and personal biologists—rather than simply eliminating jobs. He draws a parallel to the shift from agrarian economies to the modern era. While this optimism regarding job creation is well-founded, the essay fails to address how the "long tail" of problems will be solved if the compute required to solve them is constantly being outbid by high-ROI corporate applications. The market will naturally direct compute toward the highest financial return, potentially leaving the "long tail" of social or scientific problems under-served despite the theoretical availability of the tools.