One Intelligence, Many Recipes: Why Planetary AI Still Needs Cultural Filters

Black female AI researcher guiding one luminous planetary intelligence through multiple cultural lenses — One intelligence can serve humanity through many culturally informed lenses. Original MaryChuks.com illustration.

The global argument about “local AI” may be asking the wrong question. Countries do not necessarily need completely separate forms of intelligence. What they need are systems capable of understanding the cultural lenses through which people interpret family, authority, risk, respect, community, law and everyday life.

My framework is simple: one planetary intelligence, many cultural recipes.

Think of food. The planet produces common ingredients—grain, vegetables, protein, water and heat—but communities transform them into different meals. The ingredients may overlap; the recipes carry memory, identity, climate, belief and social meaning. Intelligence could work in a similar way. The underlying reasoning system may be shared, while the filters governing interpretation and delivery must be sensitive to the people using it.

Local AI is often a demand for cultural fit

When a country says, “We want our own AI,” part of that demand is about sovereignty: data location, infrastructure, security and control. Those are real concerns. But beneath them is another demand that is easier to miss: we want an intelligence that recognises who we are.

An assistant trained mainly around American institutional assumptions may give a technically reasonable answer that feels socially wrong in Lagos, London, Seoul or Dubai. The failure may not be factual. It may be contextual. It may misunderstand extended family obligations, forms of politeness, community authority, religious practice, local humour, historic mistrust or the gap between formal policy and lived reality.

That is why localisation cannot stop at translating English into Igbo, Yoruba, Arabic or Korean. Language is one filter, not the entire cultural system.

What belongs inside a cultural AI filter?

Language and register: not only vocabulary, but formality, proverbs, humour, indirect speech and code-switching.
Social relationships: how age, family roles, community and professional hierarchy shape acceptable advice.
Law and institutions: answers must reflect the jurisdiction actually governing the user.
History and power: colonial history, conflict and institutional trust influence how recommendations are received.
Religion and moral pluralism: systems must recognise meaningful differences without declaring one community morally superior.
Domain expertise: a medical, political or psychological question needs a different evidence filter from an entertainment request.
Individual preference: no national filter should erase personal identity or force every citizen into a stereotype.

The important word is filter, not cage. A cultural layer should improve relevance while preserving the user’s ability to see alternative interpretations.

Some foundations must remain planetary

OpenAI states that its mission is to ensure artificial general intelligence benefits all humanity. UNESCO’s global AI ethics framework similarly combines universal principles—including human dignity, fairness, inclusion, human oversight and accountability—with attention to diverse societies.

This suggests a useful architecture: global foundations with locally governed lenses. Evidence standards, security, privacy, traceability and a basic human-rights floor should not disappear because a harmful practice is culturally familiar. Cultural sensitivity cannot become cultural relativism.

There is also a political danger. A government could label censorship a “local value.” A dominant ethnic or religious group could present its own worldview as the national culture. A technology company could reduce millions of people to a few market-research clichés. The answer is not to abandon cultural filters; it is to make them plural, visible, contestable and auditable.

A better planetary AI design

A shared core model with clear evidence and safety standards.
Country and regional knowledge layers maintained with local experts.
Community participation from groups that are usually underrepresented.
User-selectable lenses rather than one compulsory national personality.
Explanations when culture materially changes an answer.
Audit trails showing which legal, cultural or professional layer was applied.
A visible route for correcting stereotypes and outdated assumptions.

This would turn general intelligence into contextual intelligence. Instead of silently producing one culturally loaded response, the system could say: “Under UK law, the answer is this. In a Nigerian extended-family context, the social considerations may differ. Here is what remains evidence-based across both.”

The business opportunity is in the recipes

The most valuable AI companies may not all be the firms training the largest foundational models. Many opportunities will belong to builders who create trusted cultural, professional and organisational layers: an education lens for a school system, a clinical lens for supervised practice, a civic lens for a local authority, or a customer-service lens that understands a community’s language and expectations.

That is not a smaller ambition than building a model. It is the work that makes a planetary system usable by actual people.

The MaryChuks principle: one intelligence, many recipes

A planetary AI should not arrive in every country wearing the same cultural clothes. Nor should humanity fragment intelligence into sealed national machines that cannot learn from one another. The stronger future is a common intelligence capable of passing through many carefully designed lenses—while showing us which lens is active and allowing us to question it.

That is how AI can be global without becoming culturally flat, and local without becoming intellectually isolated.