Today’s editorial: We propose a Post-Necessity Institute to test how AI abundance can expand agency, purpose, and access before economic disruption arrives.

More than 200 economists, AI researchers, and technology leaders, including at least 16 Nobel laureates and figures from OpenAI, Anthropic, and Google, have signed a new statement titled “We Must Act Now.”

Its warning is concise: AI could transform the economy on a scale larger than the Industrial Revolution, while giving society far less time to adapt. It could produce major gains in living standards, but also large-scale job displacement. The signatories call for deeper research and for the incentives, guardrails, and institutions needed to make AI complement humans and benefit society.

Institutions… in my opinion, this part is one of the most challenging.

What exactly should we build? What should those institutions protect? What should they make possible? And what can we test now, before economic transformation turns an institutional design problem into an emergency?

“Steam, electricity, and computers each gave societies decades to adapt. AI may give us only a few years,” economist Anton Korinek told Reuters. Waiting for certainty, he argued, means arriving too late.

So let’s look into the future.

Abundance. It has become one of the favorite words for describing AI’s promised future. One word, and yet a mouthful. What does it actually mean? And why might abundance be as frightening as scarcity?

Its Latin roots are closer to overflowing than simply having a lot of stuff: abundantia and abundare, meaning to overflow, linked to unda, a wave.

A river spilling over its banks can nourish a valley or destroy a village. If intelligence, labor, and productive capacity begin overflowing the economic containers built around scarcity, the question becomes: what channels do we build for them?

AI abundance could mean that analysis, software, education, scientific research, coordination, and eventually some forms of physical labor become dramatically cheaper. But productive capacity can overflow while access remains dammed behind ownership, subscriptions, infrastructure, and political power. An economy can produce more than ever while much of its population experiences less security and bargaining power.

Preparing for abundance therefore requires more than planning for unemployment. Work currently provides income, but it also organizes time, status, competence, discipline, social belonging, and contribution. If paid employment weakens as the default structure of life, cash transfers alone will not replace everything it was doing.

That is the transition I mean by post-necessity: a society in which survival depends less heavily on continuous paid employment. People may still work intensely. They may build companies, conduct research, raise children, restore forests, make films, care for their communities, or spend years mastering difficult subjects. The difference is that necessity no longer dictates every choice.

We have very little institutional preparation for that world.

Here is my intentionally uncomfortable phrase for what we should explore: Aristocracy for All. I mentioned it a few times before, and more I think about it, the clearer it becomes:

Historically, aristocrats were among the few people who did not organize life around survival labor. They had time, tutors, secretaries, household staff, land, networks, rituals, obligations, and inherited roles.

Aristocracy was profoundly unequal and often grotesque. It is also one of the closest historical archives we have for observing what people do when necessity no longer controls their time.

The results were mixed. Some used their freedom to support art, science, politics, education, and public institutions. Others converted it into status games, boredom, decadence, and domination.

What I think is important in this history is that freedom from necessity does not automatically produce human flourishing. It creates room for flourishing, dependence, curiosity, passivity, contribution, and resentment. The surrounding culture and institutions influence which of those paths becomes attractive.

A democratic version would distribute the supporting capacity while rejecting inherited rank. Ordinary people would gain access to some of the time, education, networks, tools, and assistance historically reserved for elites. Society would also need meaningful routes through which that freedom could become contribution.

This is where a new institution could begin.

The Stanford statement asks leaders to start building institutions before the transformation arrives. One such institution could be a Post-Necessity Institute.

Its purpose would be explicit: prepare society for AI abundance while that future is still uncertain enough to shape.

It would combine historical research, speculative design, public infrastructure, and long-running regional experiments. Its work would be organized around four connected programs.

Demis Hassabis has described a successful post-AGI future as one of “radical abundance.” He has also said that we need philosophers, economists, and social scientists thinking about purpose and meaning before we get there.

Science fiction gives that work a testing ground. Iain M. Banks’s Culture novels are particularly useful because they do not present abundance as a clean utopia. The Culture has freedom, near-limitless material resources, superintelligent Minds, and the persistent question of whether humans remain meaningful agents or become cherished dependents.

This program would combine such imagined futures with historical evidence from aristocracy, retirement, religious orders, welfare states, universal-income experiments, deindustrialized regions, and communities built around strong public commons.

Philosophers, historians, psychologists, artists, mathematicians, educators, and researchers of human development would work alongside the people running the Institute’s practical programs. Their role would be to identify the assumptions buried inside different versions of abundance and translate them into institutions that can be tested.

The output should shape curricula, public spaces, civic roles, AI systems, and regional experiments. We need more than books and roundtables. We need philosophy connected to implementation.

Aristocrats had staff. A democratic version of abundance could give ordinary people access to a baseline of non-human assistance.

Libraries, schools, community colleges, legal-aid centers, rural hospitals, small businesses, and local governments could become access points for advanced AI capability. A library card might provide both AI tools and trained human guidance for learning a subject, understanding government bureaucracy, translating documents, preparing forms, researching a medical question, starting a project, or organizing a community initiative.

The goal is practical capacity. Can people use AI to obtain services, exercise rights, build things, and participate more fully in economic and civic life?

Public AI infrastructure would require privacy protection, appeal mechanisms, human support, interoperability, and independence from any single vendor. Otherwise, “universal access” may simply create universal dependency on a handful of private platforms.

This is also where the meaning of complementing humans becomes concrete. Complementarity should be measured by what people become able to do, not merely by whether an employer keeps them on the payroll.

The Institute should select several urban and rural regions and fund community-designed experiments for long enough to observe real behavioral and institutional change.

Different testbeds could combine cash support, universal services, public AI access, lifelong education, civic fellowships, care networks, public laboratories, maker spaces, small-business support, local science, and cultural production.

The combinations should vary. Some regions might emphasize entrepreneurship and education. Others might invest more heavily in care, civic participation, or environmental restoration. Communities should help design the experiments, and independent researchers should evaluate them.

The central question would be agency.

Which combinations improve health, trust, learning, economic mobility, and contribution? Does AI access help people begin ambitious projects or gradually weaken their skills? Does cash support expand freedom, or is it captured through rising rents and prices? Do civic fellowships create genuine belonging, or another layer of administrative theater? Which programs produce independence, and which quietly manufacture dependence?

Failed experiments would be as valuable as successful ones. The point is to discover these failure modes while they remain local, reversible, and measurable.

If work becomes less reliable as society’s main source of identity and belonging, people will still need ways to be useful.

Markets already underfund many socially valuable activities: tutoring, elder care, environmental restoration, public archives, civic mediation, local journalism, open-source maintenance, community science, cultural preservation, and neighborhood design.

A post-necessity society could recognize these as legitimate forms of contribution. Some would be paid, some supported by fellowships or public stipends, and some voluntary. They would require training, standards, social recognition, and clear evidence that the work benefits other people.

Survival and contribution could become less tightly coupled without making contribution disappear.

The difficult question is no longer simply, “What jobs will remain?” It becomes, “How will people know that they are needed?”

Credibility would require independence from any single AI company, vendor, or government. Governance should include communities participating in the testbeds, labor organizations, public institutions, economists, technologists, educators, historians, and researchers.

Technology companies could help fund the work through arms-length pools without controlling research questions or suppressing uncomfortable results. Data and methodologies should be public wherever privacy allows. Negative findings must be published.

During its first year, the Institute could establish its governance, launch the Culture and Human Purpose program, and begin public AI pilots through existing institutions.

Over the following four years, it could open regional testbeds and produce the first comparative evidence about what strengthens agency and what creates dependency. The next five years would be used to scale the strongest models through libraries, universities, cities, hospitals, schools, community colleges, and civic organizations.

After that, successful programs should be handed to society. The Institute should prototype the transition, not become the landlord of abundance.

AI abundance remains uncertain. So does the comforting assumption that existing labor markets and institutions will absorb whatever arrives.

The value of these experiments does not depend on the most dramatic AI forecasts coming true. If abundance never arrives, people still gain better access to education, healthcare navigation, legal assistance, tools for small businesses, and ways to contribute locally.

If it does arrive, we will have rehearsed the harder transition. We will know more about what gives people agency, what creates dependence, and what humans do when necessity no longer tells them who to be. I think we are a little bit lost on all these topics.

The Stanford statement is right: waiting for certainty means arriving too late. Acting now should mean building small, reversible, measurable versions of the future while we still have time to learn from them.

We need to create the channels before the river overflows. And AI will help us with that.

If any of those thoughts resonate with you – share them across your social networks. Let’s keep the conversation going.

📹 In this episode of Attention Span, we draw a practical map of CPU, GPU, TPU, NPU, APU, IPU, LPU, FPGA… That’s a video update to our super popular article “CPU vs GPU vs TPU vs NPU: AI Hardware Processing Units Explained” Check it out →

Follow us on

Trends we see looking at every paper related to AI and ML published last week:

⬅️ FOD 156: What is the harder human-capital problem beneath token capital?