Activation Energies and Uncertainty — LessWrong