I have been thinking about this quite a lot and if we take Dwarkesh's essay seriously its undeniable that current model development is exactly equivalent to Gain Of Function
and if thats your framework the conversation around all AI development questions changes dramatically