TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Foundations · AGE 16—18 · FACTUAL · EVOLVING

Where the model runs

A model's location decides whether your drawing leaves your hands.

The idea

A model runs somewhere physical: on a server far away (the cloud), on the machine in front of you (the device), or split between the two, a small model on the device doing what it can and passing the rest up. The largest models need more memory and chips than a laptop has, so most powerful tools run in the cloud; smaller ones run on phones. The location decides whether your input leaves your hands, how fast the answer returns, what each use costs, whether it works without a network. Before you upload a drawing, know where it is going.

Why it matters

"Where does it run?" is the first privacy question, the first cost question, and the first site-visit question. Will this work where there is no signal?

See it in the studio

You photograph a client's half-built house to ask a tool about a crack. If the model runs in the cloud, a photograph of someone's private property has now left your phone. A tool running on the device would have sent it nowhere. Same question, different consequence.

Watch for this

Assuming "on my laptop" because the app is installed on your laptop. Many installed apps are thin windows onto a cloud model. Check the settings, the pricing page or the network activity, not the icon.

Try it

Install a local model runner (Ollama is one) and pull a small model. Switch off Wi-Fi. Ask it something. Then switch Wi-Fi back on and ask the same of a cloud tool. Note the speed, the quality, and what each could have seen.

Prove it

Explain the difference between a model running in the cloud, one running on your device, and a split between the two, and name what changes for privacy, speed and cost.

How it works

Cloud inference means your input is sent to a provider's data centre, processed there, and the result sent back; the provider's terms govern what is kept. On-device inference keeps the data on the machine, limited by the model size the machine can hold. Hybrids now exist. Apple's Private Cloud Compute, for one, runs small tasks on the device and larger ones on servers the company says are designed so it cannot read the data. These arrangements differ by vendor and change yearly, which is why this card is EVOLVING. The Studio Practice strand turns this into a rule about what may be uploaded.

What this idea builds on

What this idea opens up

Sources

Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.

Age grows from 11 at the centre to 22 at the edge, and six sectors show the learning strands. Tab into the map and the arrow keys step from idea to idea, following the links where there is one. Enter opens the idea under the cursor, and E reads out its links and the reason recorded on each. Press slash for Search, question mark for the full key list, and Escape to leave. Open Ideas for the complete readable list, including what each idea builds on and what it opens up.

LOGIKA · RBDS AI LAB INDIA
ON-RAMP · AGE 11 · FIRST ENCOUNTERS, NOT GATES — IDEAS · — DEPENDENCIES
DONE
OPENS NEXT
SOLID — STANDS ON