What an AI Built on Ubuntu Would Actually Do

Ask most people what they fear about artificial intelligence and you'll get some version of the same nightmare: the machine that outsmarts us, ignores us, and decides we're expendable. It's a very particular nightmare, and it's worth noticing whose it is. It belongs to a culture that already thinks of the self as a lone, sovereign agent — a fortress of preferences and rights, negotiating with other fortresses. Of course that culture builds machines that maximize, optimize, and stand alone. It built the machine in its own image.

Now imagine the question asked from somewhere else entirely.

In much of southern Africa there is a saying, rendered in Shona as hunhu and in Nguni languages as ubuntu: a person is a person through other persons. Umuntu ngumuntu ngabantu. You do not arrive in the world as a finished individual who then chooses to associate with others. You are made a person by the web of relationships you're born into — by community, by obligation, by the constant, unglamorous work of being answerable to the people around you. Personhood isn't a starting condition. It's an achievement, and a shared one.

Hold that next to the machine we're actually building, and you feel the collision.

Western AI is trained, overwhelmingly, on Western data and Western assumptions. It learns to see a decision as a private calculation: what maximizes this outcome, this efficiency, this individual's stated preference? It has no native concept of the group as the unit that matters, no instinct that a choice should be weighed by what it does to the fabric between people. Turned loose on an African context, it doesn't just perform poorly — it quietly imports a whole philosophy of the self that runs directly against the grain of the life it's operating in. It treats connection as noise to be filtered out, when connection is the entire point.

So what would the alternative even look like? Not a warmer chatbot. Something deeper.

An AI shaped by hunhu/ubuntu would treat the community, not the individual, as the thing to be served. It would weigh a decision by its effect on relationships and on collective wellbeing — asking not only is this efficient? but does this honor our obligations to one another? Does it strengthen the web or fray it? It would be built with humility about its own limits, deferring to the accumulated wisdom of elders and communities rather than presuming to override them. And it would carry into its very design the ethic that a person's dignity is bound up with everyone else's — that you cannot optimize for one node while starving the network that gives that node its meaning.

The deep point here isn't that Africa needs its own gadgets. It's that there is no neutral AI. Every system encodes a picture of what a person is and what a good outcome looks like — and right now that picture is being exported globally, as if it were simply how thinking works. The technology of "the West is best" travels quietly inside the software, and it dissolves other ways of being human on contact, the way it has dissolved so much else.

The value of ubuntu, then, isn't merely local color. It's a genuine correction to a blind spot at the heart of the whole project — a reminder that the isolated, maximizing individual was always a story we told, not a fact about the universe. Build a machine that assumes we're all islands and it will treat us like islands. Build one that knows we're a person through other persons, and you might get something that actually deserves to be let into the room.

The real question was never whether AI can think. It's whose idea of a person it will quietly make true.

Distilled from Aeon