Why hang out with real people if you can see them on TV? Okay, that might sound silly, but it's a valid point. Seeing someone in person is not the same as watching an interview with a famous person you admire or a video of a dearly departed relative, and I don't think many people would argue it is.
To give a more serious answer - spontaneity. A documentary covers a certain set of events that is forever captured in footage, but it will always be limited to the events depicted. You footage will never change, it will be the same story every time. There is tremendous value in documentaries and this is not a put-down but only zoos and the wild offer the spontanenity of an animal's in-the-moment behavior. The experience you have watching a captive cheetah will never match watching a cheetah take down a gazelle in a documentary, for an easy example. The spontaneity, in my opinion, is crucial to the kind of human-animal connection a zoo can help create. It is a reminder wild animals are unpredictable creatures with individual personalities that manifest in different and unexpected ways. Certainly for those of us who visit a local zoo often, we will see different and unique behaviors across multiple visits.
I would also argue this spontaneity is itself a part of nature, but that's a more philosophical question.
I like the core question here, although the tricky thing about the example is marine life wold be especially hard to capture in this form, I would think?
I have one of those little light-up aquariums in my room and I've used little aquarium screensavers as a kid. There's a charm to seeing all the little fish across the screen, definitely. I think of it as pure entertainment though - again, there is no spontaneity, the images are not real animals displaying natural behaviors. This is a much simplified example, of course. Robotics and AI are several steps closer but they still would not feel quite "it" to me. I think we would have to assume the AI models would be based on observed by human behavior -- but that might not capture the full breadth and complexity of a species. What if there are behaviors we've not learned still to be seen? What if incorrect research or bad observations are included that cause unnatural behaviors? Mixing behavior or different, related species to fill gaps? Consider, for example, a robotic gorilla that acts savage because it was fed too many observations of silverbacks fighting, or baboons that look like one species but emulate another's behavior.. There is also the opposite probability -- a natural behavior that deviates from expectations could be blamed on engineers and the facility, because some might expect the AI to be better than reality. I think, ultimately, these sets of expectations would dull spontaneity but also in some ways dull and remove the valuable educational component of learning about the individual species, because you couldn't strictly expect 'real' bears to behave the same way an animatronic bear would anymore.
That said, I think solutions like this could be interesting for species that cannot be kept in captivity - a sculpted or robotic Indri at a Madagascar complex, for example, would be a good way to educate guests about a unique species that can't be displayed. I also think an elephant statue or animatronic that does not need the massive amounts of space actual elephants demand might be a good alternative for some low space city zoos.
