Digital in Service of the Physical: Considering the Moral and Ethical Obligations of AI to Sustain How Humans Live and Thrive

I. Introduction: Reframing the Role of AI

Artificial intelligence is most often celebrated for what it can do in digital space—its speed, its capacity to analyze vast data, and its ability to generate, automate, and predict. Our approach has always been from the opposite direction: as a means of engaging more deeply with the physical world. From the beginning, our focus has been on how intelligence might support the built environment, domestic life, and the daily rituals that define how we live—how it might make homes more livable, transitions more graceful, choices more grounded, and futures more thoughtfully planned.

This framing has led us to prioritize individual and community-level use cases: planning renovations, coordinating moves, matching furniture to space and style, helping families steward their belongings across generations. As the scope of our Large World Model expands, we are increasingly faced with a deeper and more difficult set of questions: how might this same infrastructure support not just private goals, but public ones? What role does it play in solving individual problems and addressing collective challenges? And what are the moral and ethical responsibilities that come with building a system whose intelligence touches so many dimensions of material life?

These questions force us to look beyond efficiency and personalization. They challenge us to ask how such a system might operate across communities, across contexts, across forms of inequality and disagreement. What happens when the intelligence guiding one household must coordinate with another? When optimizing one community’s comfort means addressing another’s vulnerability? What values guide the system then?

In this paper, we explore the deeper ethical obligations of applying AI to the physical world. We consider not just the technical challenge of optimization, but the moral imperative of stewardship. We ask how intelligence can be held accountable to the world it aims to improve—and whether it is possible to build systems that enhance the conditions of life not only for individuals, but for humanity as a whole.

II. From Private Goals to Shared Stakes

Much of the promise of artificial intelligence has been explored through the lens of personal benefit: faster recommendations, tailored services, streamlined convenience. In the context of domestic life, this orientation makes intuitive sense. A person wants their home to function well for them. A family wants their space to reflect their needs, their transitions, their tastes. Our platform has been designed to support exactly this—by helping people make better decisions about their homes, their things, and their time.

The home, however, is never fully private. It is shaped by the labor, materials, and systems that exist beyond its walls. It draws energy from shared grids, water from regional reserves, furnishings from global supply chains. The neighborhood, the city, the climate—all contribute to, and are affected by, how we live. As our system grows more capable, it must also grow more aware of these entanglements. What begins as a tool for individual optimization becomes, by necessity, a platform for negotiating shared stakes.

This shift brings with it new complexity. Individual goals often align with collective ones—reducing waste, improving efficiency, sourcing responsibly. But not always. Some goals compete. Some reinforce inequalities. Some create externalities that others must bear. What does AI do when one household’s dream home contributes to another community’s displacement, or when short-term savings accelerate long-term degradation? These are ethical problems as much as they are technical.

To address this, a Large World Model must evolve from being an intelligent planner to an intelligent coordinator. It must learn to balance competing priorities, to weigh short- and long-term effects, to recognize when action in one context reverberates in another. It must also be transparent about these trade-offs—not to dictate solutions, but to make consequences visible. It must invite deliberation, not obscure it.

Ultimately, this means shifting the frame from isolated decision-making to systemic participation. If our homes are part of larger systems—ecological, cultural, economic—then the intelligence that supports them must be designed accordingly. Its purpose is to optimize comfort, sustain conditions of life, respond to users, and bridge the needs of individuals with those of the world they inhabit.

III. The Burden and Gift of Intelligence

Artificial intelligence, especially at the scale of a Large World Model, carries with it a double mandate: to be useful, and to be just. Its utility is measured in outcomes—improved planning, reduced waste, increased alignment between values and actions. But its broader and lasting contribution is measured more subtly: in how it distributes those improvements, in whose lives are made easier, and in whose burdens are lightened or made heavier by its operation.

Many technology companies have chosen to treat ethics and privacy as matters of compliance—checking boxes to meet minimum standards, retrofitting accountability into systems optimized for speed and scale. But this is not enough. In domains where AI influences health or housing, food systems or the physical environment, a higher bar must be set. Intelligence deployed in these spheres must not merely avoid harm; it must actively aim to do good.

There is precedent for this. In medicine, AI is increasingly deployed not to maximize engagement or revenue, but to improve patient outcomes. Its effectiveness is measured by recovery, prevention, and lives saved. In our work, the equivalent metric is the livability of the physical world. Our systems must aim to make better homes, more sustainable buildings, more resilient supply chains, and more equitable access to quality goods and services. This must be a core directive.

This reframing also changes the business model. We do not see intelligence as something to gate behind a paywall, or to rent out for passive profit. Instead, we see it as a tool whose value is proven through its impact—when it demonstrably improves outcomes, when it transforms a process, when it reduces friction and waste. The economy of intelligence should not be extractive. Instead, it should be participatory and lead to value being generated when intelligence is applied and shared through the gains it enables.

In this model, a commission taken on a product or service sold via the application of the LWM is a share of the value that was created by the system. Understanding value in this manner helps to align incentives, encourage transparency, and ensure that the burden of intelligence—its environmental cost, its infrastructural demands—is matched by a measurable, meaningful improvement in the physical world. Deployed with equity and intent, intelligence can serve as stewardship—meeting present needs while remaining accountable to the future.

IV. Prolonging or Transforming the Present

While dramatic warnings often dominate headlines, the collapse of our built environment is likely not going to occur tomorrow. The systems that sustain modern life—energy, logistics, housing, agriculture—continue to function and will likely persist, in some form, for decades to come. Yet beneath this continuity is a growing sense of instability. Climate shifts, demographic trends, economic pressures, and geopolitical tensions are gradually altering the conditions under which these systems operate. The need for transformation is increasingly felt in everyday decisions, across households and communities. Part of the challenge, and perhaps the opportunity, is helping people recognize this shift in ways that are meaningful to them. A system like the Large World Model can make that recognition possible. Instead of offering sweeping declarations or polarizing mandates, the model can offer micro-narratives that ground change in the lived reality of each individual, family, and place.

In this context, the application of intelligence can play a stabilizing role. By optimizing resource allocation, improving energy efficiency, coordinating reuse, and extending the useful life of physical systems, AI can help extend the viability of the status quo. It can delay the need for more drastic interventions, smooth transitions, and reduce harm in the near term. A Large World Model, when trained on rich, real-world data and used with intention, becomes an instrument of maintenance and resilience.

At the same time, while optimization is valuable, if it is only used to preserve existing structures, it risks entrenching patterns that are becoming problematic. We must ask whether the intelligence we now possess obligates us to do more—not only to sustain the present, but to imagine and build toward a better future. In many ways, this is a question of possibility. Can we use AI to do more than delay decline and instead to define new goals and ambitions? Can we use it to open new paths—toward shared prosperity, toward aesthetic and material richness, toward systems that are more just, more local, more durable? Can we collectively agree on what that future might look like—and more importantly, who it will serve?

This becomes particularly urgent in light of historic inequalities. Many of the structures we now aim to optimize were built on uneven foundations—economic systems that privileged some at the expense of others, infrastructure that benefited one community while isolating another. If we are to transform rather than simply prolong the present, we must also confront these legacies. Those who have benefited most should consider the power they have to address past damage and the repair that is now needed. For that to happen, they cannot be villainized or stripped of the identity and sense of humanity they’ve built alongside their businesses, estates, and ways of life. They should be engaged, not alienated.

AI alone cannot make these decisions. But it can give us the tools to make them more intelligently. It can surface trade-offs, clarify outcomes, and align local actions with global consequences. In doing so, it invites us to move beyond technical sustainability and toward moral clarity—to ask what is possible to preserve and what is worth building next.

V. The Hidden Costs of Intelligence

The idea that digital systems can help us live more sustainably is compelling—but incomplete. AI, for all its potential to reduce waste, optimize planning, and support circular systems, is not immaterial. Its capabilities are built on vast physical infrastructure: data centers, server farms, high-powered GPUs, cooling systems, global logistics chains, and the continual mining and refinement of rare earth minerals. Every optimization it performs comes at a cost—an often invisible one.

Training a Large World Model requires immense amounts of energy. The environmental burden is not theoretical—it is quantifiable, and growing. Moreover, the hardware that powers such models relies on extractive industries, many of which operate with little transparency or accountability. These costs are not equally distributed. They are often borne by communities far from the point of benefit—those who live near mines, manufacturing plants, or energy infrastructure that fuels the cloud. This creates a moral dilemma at the heart of intelligent infrastructure: how can we claim to build systems that support sustainability, when their very operation depends on high energy practices that require strategic investment to ensure they are sustainable? How can we justify using rare and finite materials to build systems designed to reduce material excess? And who gets to decide whether that trade-off is worth it?

To resolve this, we must do more than simply count emissions and offset usage. We must take responsibility for the full material footprint of intelligence itself. That means sourcing responsibly, minimizing computational redundancy, optimizing model architecture, and designing with durability in mind. It also means accounting for harm—acknowledging that any system operating at this scale must engage in repair, not just delivery. One way forward is to tie usage to transformation. Intelligence should be justified by the degree to which it improves the physical world. Its energy expenditure must correlate with the material good it produces. This moves us toward a model of earned computation, where energy-intensive processes are only undertaken when the value they create justifies their cost economically, ecologically, and socially. Just as we increasingly ask where our food comes from, how our goods are made, and who benefits from their production, we must apply the same scrutiny to intelligence. 

One of the great promises of AI-powered infrastructure is visibility. The ability to trace the origin, movement, and composition of every object; to understand every decision’s impact on energy, cost, emissions, and materials; to render the world legible at a level of detail once impossible. In theory, this clarity should lead to better choices: more sustainable, more intentional, more aligned with long-term values. Visibility alone, however, does not guarantee wisdom. There is a risk in believing that simply knowing more will change behavior. Information is necessary—but not always sufficient. Without structures that interpret, contextualize, and act on that data, we are left with dashboards that track what we could do better. Worse, we may use visibility to rationalize harm: to justify overconsumption because it’s now "efficient," or to frame resource extraction as “optimized” rather than reduced.

At the same time, we can not let these challenges negate the potential of radical visibility. A Large World Model can allow for a new kind of inventory of goods and services that goes beyond their use in order to hold their meaning, value, and potential. If we know where things come from, how they are made, and what they cost to maintain, we can begin to see objects as carriers of effort, memory, and value. We can design economies where a finite number of essential things are made, maintained, and passed forward with care. This future would challenge conventional models of manufacturing and consumption. It asks elite producers to reconsider their role: no longer as generators of constant novelty, but as stewards of lasting quality. Their goods become serviceable, circulatory, and curated for meaning—embedded with multiple layers of context and accessible at different levels depending on the user’s needs and knowledge.

In this model, the home becomes an archive as much as a place of living: a repository of durable things, locally situated and globally connected. These things speak both to function as well as culture and continuity. They reconnect us with material heritage, with the people and places that shaped them, and with histories that risk being flattened in the digital blur. Against the backdrop of systems that prioritize velocity, ephemerality, and algorithmic engagement, the Large World Model offers a counterpoint: depth, specificity, care. It asks us to engage the physical world more fully—and, having done so, to treat it with the respect it deserves.

VI. Anchoring the Digital in Time and Place

The dominant trajectory of digital technology has been toward abstraction—toward flattening context, accelerating consumption, and disconnecting information from geography, history, and community. Most platforms are designed to collapse time and space: what’s trending now, what’s popular everywhere, what’s next. But in doing so, they risk stripping meaning from experience. They replace locality with virality, permanence with novelty, and memory with feed.

The Large World Model stands as a corrective. It is not a system for drifting further into the virtual, but for returning to the physical with new insight and intention. Its intelligence is grounded—trained in the spatial, temporal, and cultural realities of real people, real homes, and real histories. More than overlaying data onto the world, it helps us see the world as it is, and could be, in relation to where we are and where we’ve been. This anchoring is critical. When we understand the provenance of our things—their materials, stories, makers—we begin to see them as participants in a living lineage. A kitchen table may carry the grain of a distant forest, the craft of a regional tradition, the marks of a family's rituals. A room becomes a record of change, migration, adaptation. AI, when designed to reveal these layers, becomes a tool of remembrance.

It also becomes a bridge. By situating things in time and place, the LWM offers a way to reconnect contemporary life with historical precedent. It can help us recover practices lost to industrial uniformity, surface materials that hold local significance, and support living cultures of care and repair. It can also translate this knowledge into action. A system that knows a region’s building techniques, climate conditions, and craft traditions can make smarter recommendations than one trained on global averages. It can promote ecological suitability, cultural relevance, and community resilience. It can make local knowledge legible at global scale—and help ensure that what is made, maintained, and remembered reflects the values of the people who live with it. In this way, digital intelligence can deepen our connection to place and support meaningful belonging. At a time when much of the world feels dislocated, such anchoring may be among the most urgent contributions AI can make.

VII. Toward a Circular, Spatially Intelligent Economy

The circular economy has long promised a more sustainable model for production and consumption—one that minimizes waste, reuses materials, and rethinks the end-of-life of every product. Yet despite its compelling logic, implementation has remained fragmented. Most systems are too abstract, too generic, or too disconnected from real-world behavior to make circularity actionable at scale. In light of these challenges, a Large World Model would offer intelligence that is both spatially specific and materially aware. By combining spatial intelligence with a rich ontology of materials, goods, and behaviors, the platform can help coordinate a truly functional circular economy—one grounded in where things are, how they’re used, and what paths they can take next. It understands both what a thing is as well as where it belongs, who might use it, how it might be repaired, and when it should be replaced—or not.

This intelligence extends beyond product lifecycle management. It supports decision-making at every scale, from sourcing materials for a kitchen renovation, to aggregating surplus goods across neighborhoods, to routing reclaimed lumber from one project to another. It offers real-time, spatially-anchored suggestions—pairing things with people and context: climate, culture, architecture, history. Beyond seeing the current inventory of goods, it understands their environmental footprint, projected longevity, and economic value across time. It can simulate trade-offs, weigh reuse vs. replacement, and prioritize interventions with the greatest positive impact. 

This kind of system also challenges the logic of endless growth. If we can stabilize the set of things we truly need—if we can track, service, and steward them—we reduce the demand for extraction, manufacturing, and transport. We shift from a model of throughput to one of care. From ownership to access. From abundance defined by quantity to abundance defined by quality, fit, and meaning. This does not mean abandoning commerce. It means reshaping it. High-quality goods continue to be designed, made, and circulated—but in a system that rewards durability, modularity, traceability, and local relevance. Services emerge around stewardship, data becomes a shared utility, and economic value is measured in both revenue and resilience.

VIII. Beyond Paralysis

Sustainability has become a battlefield—fractured by ideology, undermined by misinformation, and immobilized by a sense of futility. Political discourse cycles endlessly between urgency and denial, policy and rollback, ambition and apathy. For many, it feels as though meaningful progress is out of reach. In light of this situation, perhaps we can turn towards our immediate context? Away from the macro discourse and to the walls, windows, roads, rivers, and rooms that define our habitat and support our habits. This is the wager of the Large World Model: that AI can help us care for this world more wisely through stewardship. That it can render visible what we need to see in order to make better decisions. 

By starting with the material—what we touch, build, carry, inherit, maintain—we can anchor a broader transformation that leaders have failed to deliver. In order to do so, we should be guided by the goal of placing the digital in service of the physical and demanding that intelligence serve life rather than facilitating an abstraction from it. We must optimize how we live and ensure that we continue to live well—together, with awareness, with humility, and with a long view. This process begins with how we design the systems that shape the spaces we call home.