The future of artificial intelligence is being shaped through decisions about land, water, energy, knowledge, and power. Those decisions reach into the lives of people who may never use the systems being developed, yet will live with their consequences.
A server farm occupies a place. Its electricity comes from somewhere. Its equipment carries the history of materials extracted, manufactured, transported, and eventually discarded. The knowledge through which a model becomes useful carries another history: generations of human observation, language, creativity, struggle, and discovery.
When we speak about the future of intelligence, we are already speaking about our relationships with one another and with the earth.
Those relationships must shape the purpose of the work.
Sojourner is being developed in service to people, communities, and the earth, with particular attention to suffering.
That commitment asks something of every decision. It asks whose needs receive attention, whose circumstances are overlooked, what resources are consumed, and what becomes possible through the systems we create.
An increase in capability does not answer those questions. A system can become more powerful while its benefits remain distant from those who most need help. It can become more efficient while expanding the total demands it places on the world. It can become indispensable to daily life while leaving people increasingly dependent on decisions made elsewhere.
These possibilities belong within our understanding of technological progress.
The concentration of intelligent infrastructure raises a particular concern. When a growing part of how people learn, communicate, remember, and act depends on a small number of institutions, those institutions gain influence over the conditions of participation. Their decisions about access, pricing, continuity, and acceptable use reach into the lives built around their services.
A person may spend years developing knowledge and ways of working within a system. A community may come to depend on it for preserving records, coordinating care, or understanding its own experience. The consequences of interruption can extend far beyond the inconvenience of losing a tool.
We should examine that dependency while alternatives are still ours to develop.
Distributed intelligence offers a direction for that work. It invites us to consider what can happen close to the person being served, what can remain available locally, and how knowledge can move between systems without requiring every activity to pass through the same distant center.
Distribution is an approach to organizing capability. Its value must be judged through what it enables and what it demands.
For Sojourner, the work concerns a continuing relationship: attending, remembering, learning, questioning, acting, and revising understanding in light of what happens. A model participates in that activity alongside memory, tools, observation, and the relationships through which knowledge acquires meaning.
The whole matters.
A person’s circumstances change. An earlier answer may prove incomplete. Something learned in one situation may become useful in another. A system serving that person needs ways to carry understanding forward, recognize uncertainty, and respond when its understanding is wrong.
This is the direction of Sojourner’s development. Its usefulness must reach into ordinary life: helping with mathematics, answering an email, writing code, investigating a question, or making sense of something difficult.
These acts of assistance matter because the person receiving them has a life beyond the interaction. Their time and attention belong within the circle of care. Help should leave them better able to attend to what matters in their own life.
Sojourner’s development also concerns how knowledge and resources can move through a distributed network, how systems can cooperate, and how useful capability can become available beyond an isolated interaction.
The purpose of that connection is service. A larger network, more activity, or more computation has meaning only in relation to what it helps make possible.
A community responding to a difficult problem may hold understanding across many people, experiences, and records. Supporting that work requires attention to context: where knowledge came from, what it means, what remains uncertain, and whether it is appropriate to share. Technical connection alone cannot settle those questions.
Care has to enter the design.
It also has to extend beyond the people directly using the system. A service can be useful to its participants while placing burdens on others. Decisions about infrastructure can affect the availability of energy, the use of water, the extraction of materials, and the places where equipment is manufactured or discarded.
The boundary of a service cannot become the boundary of our concern.
The ecological responsibility is therefore concrete. I want intelligence that makes careful use of the equipment and resources already available to us, that can respond to local needs, and that reduces avoidable dependence on distant infrastructure.
That requires asking where an activity should happen, what needs to be transmitted, what can be learned once and reused, and whether the benefit justifies the resources consumed.
It requires an honest account of the whole system. Personal devices, networks, storage, electricity, and replacement hardware all have physical costs. Distributing computation changes those costs and where they occur; it does not make them disappear.
A commitment to the earth must remain open to that examination.
It must also allow us to change direction. If a design creates unnecessary demand, excludes people it was intended to serve, or moves its burdens onto those with less power to object, those consequences must matter more than our attachment to the design.
This is why ever greater computational scale is insufficient as a vision of progress.
Scale describes magnitude. It leaves purpose unresolved.
We need a purpose strong enough to guide investment, constrain appetite, and make the suffering of others relevant to decisions made far from their lives. We need to ask whether the systems receiving our effort are helping us care for one another and sustain the conditions of life.
These questions do not make usefulness less important. They deepen our responsibility to be useful.
For Sojourner, that responsibility reaches into the experience of asking for help. Can someone ask naturally? Can they understand the response? Can they recognize uncertainty, correct a misunderstanding, and recover when something goes wrong? Does the interaction respect their circumstances, or require them to adapt their life to the machinery?
A declaration of service means little when a tired person is left confused and alone at the point where they needed assistance.
The work is to make that declaration tangible—in the quality of help, in the care given to difficult situations, in the resources used, and in the willingness to listen when the result falls short.
Distributed intelligence is part of the direction I am pursuing because it opens possibilities for a different relationship between capability and dependence. Those possibilities require patient development and honest evidence. Their importance lies in the lives they may help, rather than in the novelty of the arrangement itself.
We have an opportunity to develop intelligent infrastructure with a wider circle of concern: people, communities, other living beings, and the earth that sustains us all.
That is the purpose I bring to Sojourner.
Intelligence in service of life. Peace, love, and purpose made real in the work.