I think I have a source for this claim below:
A fundamental principle of control engineering is that a controller benefits from a model of the item it controls.
And John Wentworth proved that under certain conditions, optimal regulators are required to use a model to control something:https://www.lesswrong.com/posts/Dx9LoqsEh3gHNJMDk/fixing-the-good-regulator-theorem#Making_The_Notion_Of__Model__A_Lot_Less_Silly
So the claim is actually justifiable, under certain conditions.
I think I have a source for this claim below:
And John Wentworth proved that under certain conditions, optimal regulators are required to use a model to control something:
https://www.lesswrong.com/posts/Dx9LoqsEh3gHNJMDk/fixing-the-good-regulator-theorem#Making_The_Notion_Of__Model__A_Lot_Less_Silly
So the claim is actually justifiable, under certain conditions.