That's the funny thing, MWI works perfectly fine without conscious observers. It's predictions match other QM interpretations. The problem comes from introducing ourselves as a conscious observer and trying to understand what it means to be consciously aware of a single outcome.