Roland Pihlakas comments on Why modelling multi-objective homeostasis is essential for AI alignment (and how it helps with AI safety as well)