Science is hard.
Although even if we can't generate the exact right cells of the exact right type in the exact right place, I think they have been able to get *some* improvement.
Those tests with mice showed formerly deaf mice *reacting* to sound stimuli. So at some level it works.
The key (for tinnitus at least) is if these lower-quality ear cells can still generate "good enough" sound input for the brain to say "oh, hey cells. Guess I can go back to quiet until you need me".