AI can probably solve "how do we get good shaper implementations into software", provided a good enough spec and test suite are available, but it won't solve "how do we convince stakeholders to value supporting languages spoken by massive numbers of people", most likely