TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Foundations · AGE 16—19 · POSITIONAL · HELD
Parametric is not AI
Parametric logic can be followed step by step, and a generative model's cannot be printed.
Our position
Parametric is not AI by default. In a Grasshopper definition a person wrote every rule, and the logic can be followed, taught, and defended step by step. In a generative model nobody wrote the behaviour, and its logic cannot be printed. Calling both "AI" hides the question that matters most in professional work: can the reasoning behind this output be shown? The Lab keeps the two apart so that students ask that question of every tool they touch.
Why we hold it
The two demand opposite habits. Parametric work is verified by reading the definition; generative work is verified by testing outputs. Students who mix them up either distrust honest algorithms or over-trust opaque ones. Both habits are expensive in practice.
The strongest objection
Transparency is not the boundary. A parametric workflow can hold black-box solvers, evolutionary optimisers, stochastic processes and plug-ins whose insides the architect cannot explain; some AI systems are symbolic, rule-based or interpretable by design. "Parametric" names a way of structuring relationships; "AI" names a broad family of methods, and one system can be both at once. "Can I trace and defend the reasoning?" is a governance test, not the line between two categories — so the categories may be the wrong frame, and the Lab may be teaching a boundary the tools do not respect.
What would make us revise it
When mainstream parametric workflows carry opaque learned solvers as defaults rather than as plug-ins, and "which parts can show their working?" has no single answer for the tools a student meets, the clean split has become a spectrum and this card is rewritten around that question instead of around two categories. We check this every edition against what the tools actually ship.
Try it
Open any Grasshopper definition and trace one output back to its inputs. Then try the same for one generated image. Write one sentence on the difference.
Take it to crit
Show one piece of work where you can name which parts of the toolchain could show their reasoning and which could not, and how that changed what you checked.
What this idea builds on
What this idea opens up
- Nothing yet names this as a foundation.
Sources
Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.