The problem with Nick Cammarata's interpretability results is that they're all subjective: I can look at a neuron and imagine it's representing all sorts of things that it may or may not be. But never fear, objective interpretability results are coming.