I certainly like the idea of uploading fossils and I do understand that is not the intended usage of iNat. Given that this would not encompass a very big swath of observations and that it is consistent with the spirit of iNat in the sense that it allows us to better learn about the biological world around us, I don’t understand why this is a problem.
Would it really feel that wrong if, hypothetically, there were people who came to iNaturalist exclusively to ID fossils?
I posted some fossils that I found in pea gravel when I first started iNat, including this blastoid https://www.inaturalist.org/observations/103961867 although there is presently no blastoid taxon so it’s just stuck at echinoderms.
iNat has consistently discouraged the inclusion/uploading of fossils, so I don’t think there’s any chance this will be encouraged. This includes having a DQA field (Recent evidence) that is pretty targeted towards fossils and official guidance for curators not to add long extinct/fossil taxa in most cases.
Yes, I tried to state that I understood that. I comprehend that it is not encouraged but merely tolerated and only allowed out of courtesy. I’m trying to understand why it is that way.
It makes sense to discourage people from posting museum specimens and adding long lists of extinct species that will never be used, but I’m not sure that is the majority of what is going on https://www.inaturalist.org/projects/fossils-of-the-world
One reason I’ve heard is that fossil taxa can mess up the CV model. I could imagine it could be harder to bin taxa in appropriate groups if there’s lots of intermediate steps filled in between what are now distinct clades. Though maybe the bigger issue with that is just that all fossils will generally look alike (i.e., grey rock or amber, I presume) to the computer vision regardless of whether they’re a dinosaur or an insect, since it’s not really looking for synapomorphies. At any rate, I could believe it adds non-random noise to the model.
iNat’s about encouraging the mapping of, appreciation of, and conservation of extant species. It’s simply not designed or intended for fossils, just like it’s not designed or intended for geology, weather, and other natural phenomena. All are important and interesting, but not within the direct scope of iNat.
I’d be interested to know where you heard that. As far as I know, we don’t train on taxa marked extinct.
The only way I could see it affecting it (other than an implementation bug) is:
1.) It could affect the level that the model is ‘pretty sure’ at if curators create otherwise unnecessary infrataxa to support the fossil taxa (though, I think that would always make the ‘pretty sure’ at a more specific taxa than it would otherwise be)
2.) If fossils are uploaded ID’d to a non-extinct taxa with extinct descendants (i.e., the fossil can only be ID’d to one of two extinct genera in a family that also has 3 extant genera) then the fossil observations could actually end up in the model, though I would guess only at some marginal level and only if the model has trained on a non-species node (i.e., it would likely only affect non-observose or very cryptic taxa).
No idea, it’s been a while. Probably uninformed speculation on a thread about the CV here on the forum. And it might not have been specifically about training on taxa marked extinct, but rather if people include photos of fossils under the miscellaneous parent taxon (since we often don’t really know where to graft a particular fossil lineage into the phylogeny) possibly affecting how confident the model wants to settle on that higher level taxon. At any rate, I don’t think the CV tail should wag the iNat dog, so this isn’t a huge deal for me.