What is the difference between using a machine and building one?

Using

The machine exists. The questions are about what it does next: what it perceives, which action it chooses, how well it performs a task, how it is deployed and operated.

This is where most of the attention in physical AI goes. That is DatumProof’s observation, not a measured claim.

Building

The machine does not exist yet, or is about to be changed. The questions are about what it is:

• which states it can be in;
• in what order it moves;
• under what conditions the interlocks engage;
• which combinations must never occur — two axes in the same zone, a transfer moving while a door is open, no way to recover after an emergency stop.

These are decided by the people who design the machine — mechanical, electrical, control, software, safety — each correct within their own scope. Problems appear when the machine is assembled, as a behaviour written into nobody’s design.

Why the building side is its own problem

On the using side a wrong decision can be observed and corrected on a machine that runs. On the building side there is nothing to run: the design usually lives in natural language and in the heads of senior engineers, in a form where the question “is this right?” cannot even be posed.

That is the gap DatumProof works on. It designs automation equipment, and builds the platform it designs on: one model of the machine, from which its tables and procedures come and against which its correctness can be asked about — before the machine reaches the floor.

Human-readable versions: about · What comes after “design with models”

Back to the question list.