Intermediate Technology

In the AI Story

The criteria Schumacher specified were practical, not sentimental. A hand loom is intermediate technology. A bicycle is intermediate technology. A small-scale irrigation system is intermediate technology. In each case, the tool enhances human capability without requiring surrender of control. The user understands how the tool works, can repair it when it breaks, and does not depend on a distant corporation for continued access. The relationship between user and tool is reciprocal — the tool serves the user's creativity, the user's creativity improves the tool.

Applied to AI, the framework produces a split verdict. Claude Code satisfies the first criterion (cheap: $100/month is accessible by any reasonable measure) and the second (small-scale: the solo builder works at human scale, directing the work with personal judgment). It fails the third criterion in one critical dimension: the tool is opaque to its user. The builder can direct the tool but cannot examine the process that produced the output, cannot modify the model, cannot repair it when it fails, cannot reproduce it if the corporation withdraws access. The reciprocal relationship Schumacher envisioned is replaced by a one-way dependency.

Berry and Stockman's 2024 paper 'Intermediate Artificial Intelligence' proposed extending Schumacher's framework to AI through open-source, locally deployable models subject to community governance. The proposal addresses the opacity problem by restoring the user's capacity to examine, modify, and operate independently of centralized infrastructure. It does not claim parity with frontier models; it claims the intermediate niche — adequate capability, genuine sovereignty, reciprocal relationship — that Schumacher argued was the scale at which technology serves humans rather than subordinating them.

The contemporary appropriate-technology movement, continuing Schumacher's institutional legacy through Practical Action and the Schumacher Center for a New Economics, has begun treating the AI transition as the next frontier for the framework. The work is difficult — open-source models lag frontier capability, the expertise to modify them exceeds the expertise to use them, and the market rewards frontier labs more reliably than cooperative infrastructure — but the analytical case is clear: the tools that will serve humans well are the tools their users can understand, govern, and replace.

Origin

Schumacher developed the concept during his work with U Nu's government in Burma in the 1950s and formalized it in a 1962 paper for the International Seminar on Planning in Hyderabad. The Intermediate Technology Development Group was founded in 1966 with support from Oxfam and other development institutions. The framework reached wide public audiences with the publication of Small Is Beautiful in 1973.

The concept has been refined and contested over sixty years. Critics argued 'intermediate' was patronizing, implying that developing nations needed lesser technology. Schumacher's response was that the criterion was not 'lesser' but 'appropriate' — that industrial technology was no more appropriate for Manchester than for Mumbai when it subordinated the worker to a process the worker could not control. Practical Action later adopted 'appropriate technology' as a less loaded term.

Key Ideas

Three criteria. Cheap and accessible, suitable for small-scale application, compatible with human creativity — the conjunction matters; any two without the third produces a technology that fails Schumacher's test.

Reciprocal relationship. The user understands the tool; the tool responds to the user's creative judgment; modification flows in both directions — and this reciprocity is what distinguishes intermediate technology from the industrial kind.

Not a hierarchy of capability. Intermediate is not inferior; it is appropriate. The question is not 'how powerful?' but 'how well does the tool serve the user's sovereignty and development?'

Applied to AI: current frontier tools satisfy two criteria and fail the third, because their opacity makes the reciprocal relationship impossible — suggesting the intermediate AI layer (open, local, governed) is where the framework most urgently needs institutional work.

Institutional requirement. Appropriate technology does not emerge from market forces; it requires deliberate development, public support, and communities of practice — which is why Schumacher founded an organization rather than merely writing a book.

Debates & Critiques

The sharpest contemporary debate concerns whether 'intermediate AI' is achievable or utopian. Skeptics argue that frontier capability requires resources only concentrated institutions can provide, so the intermediate layer will inevitably lag and the builders who depend on it will accept contingent sovereignty rather than genuine independence. Advocates counter that the argument was identical fifty years ago about industrial technology, that the intermediate layer delivered real value despite the capability gap, and that the absence of an intermediate AI layer guarantees the consolidation of control in a handful of corporations whose interests do not align with the public's.