TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · The Brief · AGE 18—21 · POSITIONAL · HELD
Synthesis is not collage
A blend you cannot account for is a collage you happen to like.
The dilemma
You have three references that each carry something you want: the spatial pressure of one, the proportional discipline of another, the constructional clarity of a third. Some current image-reasoning models can blend several references into one new thing rather than pasting them side by side. The question is what you ask for, and what you accept when it comes back.
The choices
Ask for "a combination of these three" and accept what arrives. Or name what each reference is there to contribute, ask for a synthesis, and then read the result for which ancestor became too loud — and redirect.
The consequence
The first path gives you collage: three recognisable quotations in one frame, each borrowed, none yours. The second gives you synthesis — one new thing that carries what it inherited without becoming a muddle — and only if you did the curating, named what each reference contributes, and judged the blend. The difference is not in the model. It is in whether an author directed the blend and can say what each part owes to whom.
The case
A student brings four public-domain plates — Piranesi, Wagner, Viollet-le-Duc, Fergusson — and asks for a section of a colonial-era Indian courtyard in ruin. First pass: seventy per cent there, the Piranesi darkness swallowing the Fergusson clarity. The student's job at that moment is not to praise the image. It is to say "Piranesi up, Fergusson preserved," and to know why those are the right words. Four references that share a deep logic, and the synthesis starts to hold. Four that do not, and no instruction will save it.
Our position
The mood board has always been the prompt. AI has made the synthesis visible. The reference set is now brief input, the curation is the authorship moment, and a blend you cannot account for is a collage you happen to like.
Why we hold it
Architects have always taken crosshatch density from one drawing and proportion from another and called it inspiration. Calling it synthesis, and insisting that the author can account for it, is what separates an authored plate from an averaged one. It is also the honest answer to the question of whose work is in the image.
The strongest objection
A synthesis is still made of other people's work, and naming the ancestors does not settle the ethics. Public-domain plates are one thing; a living architect's photographs are another, and "directed synthesis" can be a respectable name for derivative work with better curation. The claim of authorship may be bigger than the provenance can support.
What would make us revise it
If the profession, or the courts, settle that blended reference output is derivative regardless of direction — or if our own students' synthesised work turns out, on inspection, to be one source in costume — we would retreat to "synthesis is a technique; authorship is a separate question," and send the reader to the provenance strand. Reviewed each edition.
Prove it
Combine three references into one synthesised image and write beside it what each ancestor contributed. Then show one case where the model collaged instead, and how you redirected it.
Take it to crit
Can the student tell directed synthesis from collage in their own output, name what each ancestor contributed, and defend the result as authored rather than borrowed?
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.