I agree, transformer models trained solely on animal language may be able to predict the next "token" and generate coherent animal "sentences", even if we don't understand what any of those sounds/vocalizations mean.

However, I do think there is a path to some sort of translation to human languages. An English dictionary may seem pointless because it defines every word using only other English words. But the meaning is contained in the _relationship_ between the words.

The _relationship_ between certain animal-language tokens may look close enough to the _relationship_ between certain human-language tokens to bridge the gap, even if both languages are entirely disconnected and we've never seen a direct translation between them.