I think it would be awesome if an observation is marked unknown the computer algorithm would try to ID it even if its wrong its better than it never being seen again in the unknown cluster. I’m sure many people could fix em better, plus maybe the algorithm will get better.
it would be interesting to run the algorithm on everything, albeit it would probably be a computationally very intensive process. But i think if so the IDs would have to be tracked differently, perhaps you could sort by ID but they shouldn’t have the same weight as a human observer at this point.
Yes, you are right.
Writing from South Africa I find that the “automatic ID” feature gives me too many suggestions which could not be from ZA. The accuracy of automatic ID could be greatly improved by filtering for location.
Which that should also be a feature.
There is further discussion on this topic here: https://forum.inaturalist.org/t/species-suggestions-for-the-wrong-continent/789