The interesting part is that this isn’t completely ridiculous scientifically. An AI could potentially combine vocalizations + posture + facial expression + tail/ear position + context + individual behavioral history and learn probabilistic meanings for that particular animal. It wouldn’t necessarily translate into English sentences, but something like “wants outside,” “play solicitation,” “anxious about that sound,” “pain/discomfort likely,” “wants access to that person/object” is much more plausible.
And Michael’s business instinct is probably right about one thing: people wouldn’t merely want that technology.
They’d pay an absolutely embarrassing amount of money for it. 