No Zuckerberg, The Future is Not Just For Individuals

Meta founder and CEO Mark Zuckerberg’s new vision of artificial intelligence, set out in his essay “The Future Is for Everyone,” promises that “superintelligence” will be available to everyone rather than concentrated in the hands of a few powerful institutions. His dream is that each of us will have a personal AI agent that understands our goals, protects our interests, and helps us become healthier, wealthier, more creative, and more capable.

The problem with his “philosophy” is not the promise that everyone should benefit from AI. It is Zuckerberg’s individualistic account of what “everyone” means.

Society appears in his vision as billions of empowered individuals, each equipped with a superintelligent assistant and pursuing personal goals. Give every person enough intelligence, the argument seems to go, and a better society will emerge from the accumulation of individually optimized lives. But the future is also collective. At a time of widespread loneliness and declining trust, we need one another more than ever. The question is not only how AI can make individuals more capable. It is also how AI might help us live together, a question Zuckerberg seems to largely ignore.

Human beings are not merely individuals living alongside one another in pursuit of separate goals. We depend on shared institutions, common resources and relationships built around mutual obligations. We need schools, hospitals, trustworthy information and democratic procedures. And many of the greatest challenges facing us, such as climate change, inequality, pandemics, war and democratic instability. This cannot be solved by giving every individual a more powerful AI assistant. They require collective judgment and collective action. Zuckerberg, however, repeatedly describes AI as a tool of personal empowerment. Your agent will improve your health, career, finances, relationships and hobbies. It will align itself with your values and work continuously on your behalf.

Yet a society in which every person has a system optimized to advance their interests is not necessarily fairer, wiser or more cohesive. It could become a society of technologically fortified individuals, each better equipped to compete, persuade and protect themselves while becoming less dependent on other people.

We can already see the beginnings of this development in social AI, such as ChatGPT, Replika or Claude. People increasingly turn to chatbots for information, advice, emotional support and companionship. These systems are available at any hour, adapt to individual preferences and rarely make demands of their own. They offer relationships on the user’s terms.

My colleagues and I call this development “AI individualism,” which describes a social shift away from networks of people and toward highly tailored interactions with AI agents. It echoes what the sociologist Barry Wellman called “networked individualism.” The internet loosened people’s dependence on the strong ties of family and neighborhood, replacing them with more dispersed and self-curated networks. AI individualism may take this process a step further by reducing our need even for weak ties, the acquaintances, casual contacts and loose networks that provide information, advice and everyday support. Where a neighbor, colleague or friend of a friend might once have filled that role, a chatbot can now do so on demand, at any hour and without requiring reciprocity.

This shift may bring real benefits. For people who lack access to human support, the isolated, such as overworked or geographically stranded people, an always-available agent can be a lifeline. But what helps an individual is not necessarily what sustains a society. A population increasingly supported by private AI agents rather than by other people may gradually stop creating the connective tissue from which societies are made. For example, research indicates that more intensive chatbot use has been associated with greater loneliness, emotional dependence, and less human socialization. Also, recent research from Stanford suggests that AI chatbots may worsen loneliness among vulnerable users; those with limited social networks who sought emotional support from AI chatbots reported lower well-being.

Zuckerberg presents personalization as a route to diversity. Because there is no single good life, everyone should be free to direct superintelligence toward the life they choose. But diversity requires more than individually aligned machines. It also requires encounters with people who resist us, disagree with us and make claims on our attention. Human relationships are valuable partly because they cannot be fully customized. They require negotiation, compromise and care. A personal agent designed to serve my goals makes no comparable claim on me. It may simulate care without requiring me to care in return.

Equal access to a personal AI agent is also not the same as equal power. These systems require enormous data centers, energy supplies, computational resources and stores of human-produced data. The companies controlling that infrastructure will still decide which models exist, how they behave, what they cost and which kinds of personalization are possible.

Personalization can therefore create an illusion of decentralization. Power appears to be transferred to individuals while remaining concentrated in the infrastructure on which their agents depend. Zuckerberg proposes allocating computing power through auctions at the lowest available price. But auctions allocate resources according to purchasing power, not human need.

This is where Meta’s recent child-safety case matters. Only days before Zuckerberg published his vision, a New Mexico judge ordered Meta to pay $567 million to set up an abatement fund to address harms from its products. This on top of an earlier $375 million penalty, after a jury found that the company had misled users about the safety of its platforms and enabled the sexual exploitation of young people. The case shows why access to privately provided technologies cannot substitute for public oversight. When companies build systems that mediate relationships, information and care, the question is not simply whether individuals can use them. It is who protects the public when corporate incentives conflict with human needs.

Zuckerberg’s vision risks replacing some forms of social capital, the resources people draw from relationships, with what my colleagues and I call “AI capital”: resources supplied on demand by privately owned systems rather than developed through human reciprocity. AI capital may be faster and more convenient. But it is not necessarily an adequate substitute for what it replaces.

Zuckerberg is right to reject a single, centralized superintelligence. Such a system would create extraordinary possibilities for authoritarian control. And he does nod to “democratic values” and “democratic governing institutions.” But his alternative—billions of personalized agents pursuing private goals while depending on infrastructure controlled by a handful of companies—carries another danger. It mistakes the absence of a single center of control for the presence of genuine democracy.

A genuinely democratic AI future must ask a different set of questions. Which forms of artificial intelligence should be treated as public goods? Which decisions should remain subject to democratic deliberation rather than personal optimization? How can AI strengthen schools, public health systems and local communities instead of encouraging people to retreat into private systems of assistance?

Between centralized control and technological individualism lies democracy: the difficult practice of deciding together what we owe one another and what kind of future we want to share. The future should indeed be for everyone. But it cannot be built around billions of private assistants serving private ends. It will require AI that serves public purposes, institutions capable of holding its creators accountable, and a public willing to build the future together.