On the contrary, if something is impossible then making it possible most certainty makes it easier!
My issue certainly isn't with interfaces being easier for all things, I was challenging that this benefited in any substantial situation at all, but I'll concede my point until and unless I refine my argument about people being unwilling to invest effort in a new interface.
That said, the theremin was a new musical interface and didn't ever take off because it produced a strange sound quality and did not map well to preexisting human motions. Furthermore, music creation was already well understood.
Indeed, machines need to be built for the human, not the other way around. I think it's a trade off between two terms: how easy the interface is to learn and what it enables us to do. For (a contrived) example, an interface which allows humans to teleport will be learned regardless of how steep the learning curve is. Teleportation is a "zero to one" technology, it more or less competes with nothing. A new kind of keyboard on the other hand, must compete with the existing solution for keyboards, so it must be easy to learn, otherwise there would be no incentive to switch.
Interfaces are supposed to make things easier, but they are not required to improve every possible application. If this interface makes just some things easier or possible at all, it will be a win.
Using a camera would require mounting it somewhere where it can see the hand at all times, which precludes mobile uses. There's also the problem of exhausting yourself by using gestural interfaces for too long, where just minimal muscle twitches would probably be more comfortable.