Maybe off-topic, but the article says they used a modified machine learning algorithm to nap bat calls to activities and from that developed a translator. With

Maybe off-topic, but the article says they used a modified machine learning algorithm to nap bat calls to activities and from that developed a translator. With the improvements to machine translation without parallelism data and voices-to-voice translation, how stupid is the idea that we might someday be able to take enough bat calls and human speech, and create a true translation system? obviously assumes the bats got something vaguely resembling a language we can translate from.