Why’s it hard to imagine? More training data will solve whatever language lapses it has. The next miracle is that TTS is perfect now, so they don’t need to be able to read.

You can convey abstract concepts as alternate abstractions, explain like I’m five but on turbosteroids. It’s the ultimate teaching tool and it’s about to be ubiquitous.

What training data? Many of these languages have very little digitized literature. Even if we assume they have sizeable extant corpuses (e.g. Tibetic/Bhoti), that's not enough. LLMs are still pretty garbage at English prose, for example.