Fast Cat seems pretty good - with occasional aberrations such as calling a red-necked grebe a little grebe.
Is there any way of requesting FastCat to look at a particular observation? I’ve had no feed-back so far on this one, which is admittedly difficult.
It does look a bit redpollish but there is just not the detail in the image due to being against the light so I suspect AI would not be able to help. It was always (my) intention to use AI on older observations both that do and that don’t have agreements or ID’s. However there are a couple of problems, first don’t want to further annoy the community about AI and second with fastcat at least there was a capacity issue I think although that might have been resolved.
If there were a separate way to do this so that it did not annoy the community and potentially could also give an additional agreement where it agrees with an ID that has already been given by human but there is no agreement then that might be helpful too. Infact just finding these observations might be useful i.e. the AI looks through all old observations and for those without an agreement sees if it agrees with the ID that has been given and if it does then make a list of the observations for the community to have a look at so that humans can give an agreement.
Personally, I have no problem with FastCat. I think it’s the plant one that has caused all the aggro because it is often so way off the mark.
I would be all in favour of using FastCat to check birds that don’t have any agreements - there are dozens, possibly hundreds, in the South African community. But I don’t know whether FC is programmed for non-UK species. I imagine that relevant data is variable, particularly for parts of the world where there are fewer bird-watchers/ornithologists.
I thought I’d try it on iNaturalist. It came up with lesser redpoll.