The post goes into this in some detail. In human languages, the embeddings for specific concepts seem to follow very similar geometric relationships no matter what language. If that holds across species, then you could use geometric similarity to map animal embeddings to human concepts.
Also, since chatbots just output "meaningful" strings of tokens based on a string of input tokens, you can build a chatbot that can "talk" to animals without us necessarily being able to understand it. But sure how useful that would be, but interesting nonetheless.
Also, since chatbots just output "meaningful" strings of tokens based on a string of input tokens, you can build a chatbot that can "talk" to animals without us necessarily being able to understand it. But sure how useful that would be, but interesting nonetheless.