Machines should feel the field
AI that respects physics needs fewer sensors.
Most AI models learn from data alone. Show them enough examples and they find the pattern. But they do not know that heat flows from hot to cold, or that mass cannot vanish.
I think the next useful models will be field-aware. They will carry the rules of physics, chemistry and biology inside them.
What that looks like
A field is a quantity with a value at every point in space: the temperature in a room, the pressure in a pipe, the pull around a magnet. A field-aware model predicts the whole field from a few measurements.
This is called virtual sensing. A handful of real sensors anchor the model. The model fills in every point between them.
The honest limit
You cannot remove every sensor. A model with no real measurements drifts away from reality. A few sensors must stay, to anchor it and to check it.
Where I want to use it
Indoor farms, where every sensor adds cost. Chemical plants, where some points are too hot to measure. Robots, which must feel their surroundings to move well.
The marks behind this page are a small nod to the idea. Move your cursor. It acts like a magnet, and the field turns to follow it.