How complex, precise and “accurate” could you be using descriptors?
I’m also talking about spatial intelligence as well.
I.e. let’s assume I’m describing an object:

There’s a large square-like base that’s four by four meters in length and width and length, forty centimeters in height and concave at each angle. The base it holding a giant panel that’s flat on each side with a depression from the front side; its measurements are forty centimeters in width, thirty five in length, and fifty in height. On top of it, there’s longer lid-like almost curved box-like “ram” that’s forty centimeters in length, ten in height, similar in width, but it’s curved at the end like a tridimensional semi-circle. A lever works the mechanism inside the machine, which keeps the five pumps inside it going.
At the back of the panel, five centimeters off the base, there’s a generator which holds the energy. It’s separated by a thin wall where all the wires and cords are connected, leading into another fake panel beneath the panel, into the base, where it leads through a few even and circular holes into the pumps.
There’s another lever that’s connected to the generator.
Outside the panel there are also round spherical turning wheels which are being worked on by a long band that seem to bear protruding pipes which go through the machine. They’re raised at a level that it unreachable by the pumps, separating the inside of a panel into two sections.

Can language be used this way and produce any kind of result?