I work at NIST and took a tour of this machine a few months ago. A few bonus facts:<p>- The spatial distribution of the local gravity field is a significant part of the uncertainty of these measurements. The weights extend into the basement of the lab, and the gravity in the basement is less than that above the surface. They produce gravity maps by dropping things in a vacuum to get a handle on it.<p>- The drift in the dead weight standard was mostly caused by the individual masses welding themselves together under the immense pressure of the weight stack. The interface has been designed to reduce this effect.<p>- The same group is also working to count the number of atoms in the kilogram, so that the mass of the dead weight stack will not be 500,000 times the mass of a piece of platinum in Paris, but will be 500,000 times the mass of 6.XX E23 silicon atoms.