CONCEPT
The Gold-Foil Method
Rutherford’s technique for learning the hidden structure of an opaque object by bombarding it and reading the pattern of what comes back—the deepest method available for understanding a thing you cannot open, and the right frame for interpretability research on sealed neural networks.
In 1909, Hans Geiger and Ernest Marsden, working under
Rutherford’s direction at Manchester, fired a beam of alpha particles at a thin sheet of gold foil and counted where the particles went by watching faint flashes on a fluorescent screen. The prevailing model of the atom predicted that the foil would barely disturb the beam. Instead, roughly one in eight thousand particles bounced back toward the source—deflected through more than ninety degrees. From those rare violent rebounds, Rutherford inferred the nucleus: a concentration of charge and mass packed into a core thousands of times smaller than the atom around it, with nearly all the rest of the atom empty space. He had not seen the nucleus. He had inferred its existence, its approximate size, and its charge purely from the statistics of how a beam scattered when fired at something he could not open. This is the move: perturb an opaque