Most people have been introduced to artificial intelligence through the wrong lens and the wrong language.
We hear about job losses, threats, monsters, god-machines, arms races, giant data centres and billion-dollar companies. That language makes ordinary people feel small. It makes AI sound distant, dangerous and owned by someone else.
Instead of treating AI only as a giant machine in the sky, what if we treated it more like farming?
AI models are not all the same. Some are big paddock crops. Some are delicate herbs. Some are mushrooms in a box. Some are useful little seedlings that do one job well and do not need to become a giant rainforest. Like crops, they need to be chosen for the soil, the season and the job.
Seen this way, AI is no longer only a black box inside a distant data centre. It becomes something that can be cultivated close to home, close to work and close to the problem that actually needs solving.
The point of “AI as farming” is not to make artificial intelligence cute. It is to make it usable.
This matters especially in healthcare.
A GP practice or rural hospital may not need a giant AI system trying to replace everything. It may need a small local AI tool that helps staff search policies, summarise guidelines, draft routine documents, support training, translate instructions, organise rosters or find the right information faster.
Local problems
In many settings, the goal is not to build the biggest AI in the world. The goal is to build useful AI that works safely for a local problem.
The farming analogy points to a natural category: small-scale AI farming.
Can it be done? Yes. Free and open tools already exist. Downloadable models, open-weight models and local AI platforms are becoming easier to use. There is still a learning curve, and the hard yakka can be surprisingly rewarding.
You no longer need to begin with a subscription to a large AI company or a constant internet connection for every AI task. You can begin with a computer, a practical problem and a willingness to learn. Over time, you can adapt models, workflows and local knowledge to suit your own needs.
Small-scale AI farming is not anti-cloud, anti-industry or anti-progress. It is the missing middle between trillion-dollar data centres and people with real problems.
At one end are frontier AI models and huge industrial systems. At the other end are hospitals, schools, farms, councils, small businesses and community organisations with existing hardware, local knowledge and practical work to do. Between them sits a large opportunity: community-scale grassroots AI.
The first step is simple. Look at the hardware already around you. Reuse if possible. Buy only what is needed. Gaming PCs, old workstations, office servers, laptops, mini PCs and small embedded devices can all play a role depending on the problem. That is the soil.
Then choose the seed: a language model, vision model, embedding model, speech model or tiny machine-learning model.
Then choose the crop: summarising documents, searching policies, drafting letters, assisting training, helping with stock control, supporting translation, organising rosters, detecting problems or enabling privacy-sensitive workflows.
The recipe is simple enough to begin.
Start slowly
Start with one small plot. Pick one repetitive problem. Run a local model. Add your own documents. Test it on real examples. Keep a human in the loop. Record where it fails. Improve the prompt, search, workflow, interface and safety rules. Add the equivalent of fences, irrigation and weed control. Then expand slowly.
This is not magic. It is practice. You learn the soil by working it.
The detailed technical structure of an AI model farm is beyond the scope of this article. Hardware choices, model selection, local document connection, safety testing, privacy design, cultural governance, maintenance and monitoring all need careful work. That practical framework is being developed separately through Ol’ MacDonald’s AI Farm as a manual: a simple, usable model for helping individuals, organisations and communities start small, learn safely and grow local AI capability over time.
The limits must also be stated clearly.
A local AI model under your desk is not wisdom in a box. It is a tool, and every tool can cut the wrong way. Local AI models are not the same as the largest frontier models. They may be slower, smaller and less fluent. They can hallucinate, misread context or produce confident nonsense if used carelessly. They still need maintenance, updates, testing, monitoring and governance.
Running a model locally does not automatically solve privacy, bias, safety or cultural risk. Some configurations can still be resource-intensive. Some uses should not proceed without specialist oversight. Some information should never be fed into a system at all.
Small-scale AI
But limits do not kill the idea. They define the design.
A backyard vegetable patch does not replace global agriculture. It does not claim to. Yet it gives a household freshness, skill, resilience, independence and a different relationship with food. Small-scale AI can do something similar for intelligence work.
This also has business potential. There is room for local AI farm kits, model selection services, privacy-aware appliances, school and council deployments, hospital pilots, maintenance contracts, community training, rural edge intelligence, small-business automation and cooperative AI infrastructure.
The value is not only technical. It is economic and civic. Like local produce, local AI can keep more skill, capability and decision-making inside the community. It can create new kinds of work around installation, support, education, governance and adaptation. These jobs will be interdisciplinary, combining domain expertise with technical know-how. That is not a side effect. It is part of the market.
There is another important layer. AI is not culturally neutral just because it is technical. A model built far away may not understand local context, community priorities, Aboriginal and Torres Strait Islander knowledge systems, minority experiences, regional realities or the obligations that come with Country, kinship, language and cultural authority.
Small-scale AI farming allows local values to be built into the system from the beginning. This is not decorative ethics. It is the fences, gates, water lines and firebreaks of the AI farm, built with the community rather than imposed on it.
Serve the people
The model must fit the Country. The tool must serve the people.
That is the strength of the farming metaphor. It reminds us that local conditions matter.
This idea is urgent now because every front page seems to carry another story about mega data centres: more power, more water, more land, more money and more scale. Those data centres will not disappear, and some problems really do need massive infrastructure. But if the only story we tell about AI is the story of bigness, most people will remain spectators inside someone else’s machine.
Small-scale AI farming offers another path. Let people become participants in the AI economy, not just customers of it.
Local AI is not exempt from safety, ethics or oversight. It still needs boundaries, testing and responsible use. But those boundaries should be shaped with input from people who understand the local problem, not only by distant institutions.When communities can participate in AI development outside purely corporate stacks and cloud-connected hyperscalers, the discussion about regulation becomes more informed.
If people can immediately see how to use an idea, adapt it, argue with it and carry it elsewhere, then it has life. The idea germinates because it is practical. That is the crux of AI farming.
Voltaire ended *Candide* with a sentence that still cuts through centuries of noise:
“We must cultivate our garden.”
Perhaps the same is now true of artificial intelligence.
Dr Balaji Bikshandi is a Canberra-based intensive care specialist and healthcare innovator interested in how artificial intelligence can strengthen, not shrink, the role of clinicians, individuals and communities. His work explores practical ways to expand professional autonomy, empower practitioners, promote health equity, improve health system capability and preserve clinical judgement at the centre of medicine while thoughtfully adapting the profession to changing times.
Add Comment