- A really clear, obvious disclaimer that it is AI and may be incorrect. It would have to be very visually obvious - not a little disclaimer in normal sized font or fine print.
- Any user can report a problem with the AI (it’s inaccurate, inappropriate, whatever). And there would need to be a plan in place for correcting it (or removing/flagging problematic AI content), even if it’s a seemingly minor inaccuracy.
- Users can opt out.
4 Likes
To be clear, this is not the case and has never been the case. Google has given us a grant for us to try and surface and perhaps synthesize helpful ID comments, and make a demo for it. We’re not giving data to Google, and Google isn’t storing data for us. Google has no more access to iNat data than other sites do.
15 Likes
I think I understand. For someone who used punch cards in the 1970s, what do you mean by “try and surface”? Thanks!
1 Like
Something quicker and easier than - I know I saw that explanation, somewhere. Or the pile of bookmarked obs that I can no longer sort thru. Or this project- a good idea but not so practical to search.
1 Like
This is a recent use of the verb form of surface, meaning to bring to the surface, unbury, discover, bring to light, reveal, etc.
7 Likes
https://forum.inaturalist.org/t/taxon-photos-are-editable-by-many-people-leading-to-abuse/67614
https://forum.inaturalist.org/t/reversing-mass-incorrect-ids/69366
…there are tons of more stuff in the forum like this.
I would be much more tolerant to AI if it can help managing people (e.g. enforcing policy compliance, flag incompliant content, assess and recommend user training and status, slow-down uncontrolled mass identification, etc etc etc…)
That’s the bottleneck, and not the proposal of giving a specific ID or similar related things.
ID assignment is a cooperative effort and you obviously need to learn to manage people and process do ID here in a good way. If gen AI can’t do this, it shout like a good robot stay out of the way and should help sort out other, more pressing and inconvenient tasks.
1 Like
I am hesitant to put this suggestion under a thread with this title, since I’m not sure it makes anything tolerable, but what if iNat goes in a completely different direction, and uses the grant to develop a ChatBot that says “I see you’re trying to create an observation with no identification. Can I help you to assign a preliminary ID to your observation?” and then asks a series of interactive questions to guide the user to a high level preliminary ID? This way, we’d be reducing the Unknowns, and not having AI mucking about at the species level.
2 Likes
Doesn’t the CV basically already do this?
1 Like
Interactive questions to get us to an ID?
No. We get a list, which we must pick over and see if one fits. Or none of the above.
It won’t work on upload, because the yellow label projects take time to populate. But you can tackle your preferred slice of biodiversity via
I’m not sure what you mean - the same thing that triggers the dialog box can trigger a chatbot - no project need be involved.
2 Likes
Can - yes. But the projects exist already - if there is a taxon that interests you. Beetles? (there is a blue and white one parked across the road ;~))
I’m not following you at all - I’m not suggesting beginning user engagement with projects - that would require more knowledge of iNat than I suspect many new users have. What I’m proposing is chatbot handholding for attempts to upload observations with no ID attached to prevent unknowns. By the time an observation is in a yellow label project, we’ve already lost and are just cleaning up the mess.
3 Likes
I’m not sure what category this falls into but I think AI could be used for searching observations for habitat indicators. This isn’t generative AI but rather a more robust and refined image search capability.
Let me give an example.
The science center where I work has a planted garden which has attracted a lot of insects. There has been a lot of pollinator activity recently which I’ve been documenting. While I could also document the plants, I am less inclined to do so because they are captive and cultivated. And even if I were to do so once, do I really want to clutter the servers with multiple observations of the same cultivated marigold on different days just because a different pollinator is on it?
I think a better solution for those interested in studying urban wildlife that depends on captive and cultivated organisms is to be able to search the observations looking for habitat indicators.
4 Likes
If this topic is still open, I heartily second a standardized way to add a warning on species that cannot be ID’d with current knowledge or fairly easily obtained equipment. I believe that current practice and AI/CV is building in errors and misleading ordinary users on the messy ways in which nature and evolution work, as well as our lack of knowledge and understanding.
4 Likes
I think the chatbot idea or similar could make a real difference. In two projects I curate, doing my best to ID with limited knowledge, I have had good responses when I took the time to comment, explaining to folks who may have been new some of the reasons why IDs are difficult, things to take into consideration, what to look for, why they should be cautious about picking from a suggested list, etc.
2 Likes
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.