Lets not countinue the survey topic, we need a rigorous survey to know the communities true opinion?
While i respect your request, the above seems contradictory in a way.
Perhaps we should request staff create a public poll (not public in the way showing how individual people vote) and send it out to everyone as an announcement like the blog?
okayā¦ if it ends up being a program that directs people to comments iāve made, that could potentially be something i could get behind. If itās a program that attempts to rehash what iāve already said, no, iād just never leave comments again. It doesnāt really matter how good it is at this, i do not like the principle.
And this is where i suppose feelings differ from some others on the issue, and reasoning wonāt do anything. Iām not comfortable asking genAI for ID tips, or as a whole really. If you are, thatās none of my business, but can we keep it separate from inat? can we not be explicitly training it on our users without their consent?
iām having trouble understanding the distinction between the following. would you elaborate?
there is no specific implementation at this time. so are you saying that any talk about what the implementation (theoretically) might or should look like should go in a separate thread?
i guess i donāt mind if some stuff needs to be discussed in a separate thread for some reason, but iām having trouble understanding which stuff that is.
The discussion and controversy of all this is large enough maybe seperate forum posts isnt a bad idea. This is already 391 comments.
I believe what many of us are asking is akin to: what was promised to Google or submitted in the grant proposal so iNat would get the 1.5M? Something had to be proposed beyond āwe want to use genAIā so Google could pick among candidates, no? The blog post and FAQ contradicting itself doesnāt help the confusion.
I want to register my disapproval for all the people presenting a sunk-cost defense of this incorporation of genAI and Google tech into iNaturalist. We are all fully aware that genAI and Google products structure our lives and iNat usage intensely. Saying that such existing use should make us okay with any further use of any further related tool is tendentious and fallacious.
Awesome, thank you!
I did make the choice to assume staff mean well. āAssume others mean wellā is, after all, what Tony often gently reminds us to do.
Wait so.. the tool isnāt even gonna be a type of GenerativeAI? I wanted to joke about how it could be made as an internal search engine and then labeled as GenerativeAI, but that would have been more of a mischievous suggestion/joke(which I felt would be tonedeaf in such a serious discussion). And honestly, now Iām even more confused, so it was really just used as a buzzword? It would be a very strange choice of wording if the developed feature will be unrelated. The layer of miscommunication seems to have been doubled. If you can, please elaborate a bit on what you meant there.
So we can hope for more good stuff like the Disagreements filter. If you are looking for me ā¦ I will be there
Good news, but I feel like the time I spent reading this stuff the past two days was a big waste of time then and I donāt appreciate that at all.
Unfortunately it turns out that iNat staff, like many of the rest of us, tend to be nature nerds who are happier hanging out in a bog with frogs than interacting with people, and thus the communication skills may be a little ā¦ lacking ā¦ at times.( I say that with total love as a nature nerd who wants to be in a bog right now). There will be some much better communication incoming soon, and if it helps, they definitely all feel terrible over this entire mess and look like they havenāt slept in several days.
In the blog post, they described a specific feature they wanted to develop, and linked to a blog post from last year that said they wanted to use a Vision Language Model, which is essentially an LLM with some visual processing stuff attached. This isnāt badly worded corporate buzzspeak, they very clearly gave an example of what they wanted to do and have had a plan in place for at least a year now that involves using generative AI.
Personally, the contradictions between what was said in the blog posts, both a year ago and a few days ago, and what has been said on the forums since then are making it hard to feel like I can trust anything the staff now say about this project. It feels like theyāre either wildly backpedalling or have no idea what theyāre talking about when it comes to AI, and if itās the former Iād much prefer for them to just say āweāve listened to the communityās responses and have decided to pivot towards developing something more like this instead of the original plan to use genAIā.
Maybe Iām just incredibly cynical, but I donāt see how saying you want to use a very specific kind of genAI and showcasing a mockup of the feature you want to implement and have apparently been planning for at least a year could be passed off as just ābadly worded corporate buzzspeakā
Thanks for sharing this! This is a great example of a āleast worst way forwardā that I was cautiously hoping for, tbh. I never assumed malice on the part of iNat staff. Just very badly bungled communication. Iām still gonna be guarded, but I am a little more reassured. I also wonder if theyāre taking our feedback a little more seriously than theyāre leading on, but thatās just speculation on my part.
Hopefully iNat are learning a very difficult lesson about communication and PR. 
Awesome, thanks for posting this! Iād still love a user-based wiki thoughā¦
I mourn the observations, identifications, annotations, other collected efforts, and users lost among the confusion.
iNaturalist bled data and community.
Thanks @graysquirrel for spending time today and helping talk through how we got here.
I want to assure everyone that WE ARE NOT CHANGING THE DIRECTION OF iNATURALIST. As I say every chance I get (and really mean it) iNat is great because the community is great. iNat is the community. I can assure you that I and the entire iNat team hates the AI slop thatās taking over the internet as much as you do. Thank you for trusting us as stewards of this awesome people-powered community and website
On top of that, I want to reiterate that:
Some quick reassurances:
This grant is very standard alongside the kinds of philanthropic grants weāve depended on to keep the lights on here. Whether itās Microsoft AI for Earth or Amazon Open Data Program or Google Earth Outreach, we donāt always love that weāre so dependent on these kinds of grants or credits, but applying for these grants is how weāve managed to keep iNaturalist online and non-commercial for going on 18 years. These are grants, they get nothing in return except for some of the glow of supporting awesome social initiatives like iNat. We get money or credits.
Google.org isnāt expecting anything from this grant except for us to explore Gen AI technology in the context of our work to come up with solutions for how to better surface and organize expertise on the site.
Our grant deliverable is to make a demo. Many of you saw the Vision-Language demo we made as part of a grant with Microsoft AI for Good program to help run our servers. We didnāt end up integrating this vision-language technology into the site, but we did learn a lot and it directly led to the Photo Similarity feature we released that we love. Gen AI means so many things to different people, but to us it probably means that the demo we will make will leverage interacting with an LLM in some way. We are still figuring out specifics of how LLMs might play a role in this demo, and honestly donāt know more than that at this point, but we are listening to your feedback and concerns (e.g. making it opt-out) as we proceed.
This grant will fund a lot more than just that demo, and I think weāll learn a lot about how the tech landscape is evolving and what that means for iNat. If the demo is not helping it wonāt get integrated into the site. And let me reiterate: thereās no way weāre going to unleash AI generated slop onto the site. iNaturalist is about human connection and expertise and using technology to help elevate and support that.
We are working really hard to do a lot of things at the same time: stay focused on the iNat community, surface all of your expertise and hard work, hustle to fund our small non-commercial nonprofit, stay up to date on the (very rapidly evolving) tech landscape, and more. We will have more for you about this grant and project after the weekend. But I hope writing this will help reassure people going into the weekend. Again Iām very sorry for the miscommunication on our side.
Thanks for doing this - itās good to hear. On one hand, Iām hopeful that this reflects the fact that the clarity we desperately need is coming.
On the other hand, so long as the blog post (and its update) remains the latest official public announcement of the project, my deepest worries remain. The announcement (and the Google blog post) still suggests an implementation of genAI to explain species IDs that is unworkable, untrustworthy, dangerous, and undesirable, among many other risks and downsides that need no repeating.
What does bear repeating is that the blog post did not draw a clear distinction between iNatās existing machine learning tools (the CV and geomodel), and its suggested implementation of AI text generation. I hope that, moving forward, the iNat team make this important and relevant distinction as clearly as they can. AI is not a monolith. Image recognition is categorically distinct from a genAI model that scrapes ID remarks to produce natural language sentences. It is a mistake to treat them as the same kind of solution - a mistake many continue to make.
In addition, I would still like to see broader, more open, more visible avenue of community feedback and discussion on this topic. The discussion on this forum thread isnāt easy to find.
Yeah, I come out of this even less assured than before. So either they are now back paddling and lying to us here about not having planned to use any āGenAIā technology at any point - or there was some equally playing fast and loose with the truth towards Google when applying for that grant.
Not that I care much for Googleās feeling, but Iām not sure I like that approach any better.
The bleeding is likely a drop, it will be replaced within days anyway. Judging by their stats - a steady flow of users and data, almost Google Bait - they can definitely afford (no pun intended) losing a few hotheaded identifiers, observers, even donors.
edit: the most unkindest cut of all might very well be the slight but enduring erosion of trust and enthusiasm from those members who are still here, see msgs belowā¦
