You are right that testing whether the gene changes how cells react can be tested in vivo, and that's how it was done originally. But the point of these studies

You are right that testing whether the gene changes how cells react can be tested in vivo, and that's how it was done originally. But the point of these studies is to figure out how certain parts of the brain work. Since the brain produces behavior, we can best understand how it works by changing brain activity and then seeing how that changes behavior. That can't be done in a dish, at least not now, even though we often use modeling whenever possible to make sure we're doing the right experiments. The goal was definitely not to control mice or make them suffer and I find it a little regrettable that the professor did not emphasize that. The fact that making neurons in motor cortex fire causes movement is proof of concept, but no neuroscientist wants to just sit around making mice just turn in circles all day. The goal here is to better understand how neurons firing produce behavior. For example if we can crack the code for movement, we can e.g. help patients better recover function after a stroke. The techniques we've had for a hundred years have not gotten us all the way to an answer on that yet. Think of it this way: do you want to debug your program with only print statements, or would you rather be able to put in a break point?