iNaturalist receives grant to improve species suggestions

Well, I guess it could be worse. I think a user-contributed wiki would be better (and more fun!), but I don't make the rules.

Will there be a beta testing phase of this feature, so users (select users? curators, maybe? or those who wish to opt in?) could see how accurate (or inaccurate) it is before it's released to the masses?

Please preserve the ability for community user input for identification tips and text descriptions. (maybe curator only?)

Please be careful not to fall into the trap that Google and others have laid of wanting us to treat AI as if it is a monolith. Image recognition (with suitable caveats) has always been a statistically sensible use case (large, well annotated training sets; fine-grained and testable predictions). LLM-based models, even when propped up with other code to avoid the worst errors, are not the same thing.

Even if I could ignore the fact that generative ML is a cynical bid by Silicon Valley to extract as much value as possible from all of society's data, knowledge and labour and from the environment, I beg you to be very cautious before helping their latest black boxes to become the intermediaries between humans and knowledge. We are living in an age of scientific denial and rejection of evidence. Web search is increasingly unreliable. Public datasets are being taken down. Wikipedia and the Internet Archive are under attack. We have few places where we have direct and relatively unfiltered access to knowledge and expertise. Many see iNaturalist as one of the last remaining good places on the Internet. Don't let Google's interests wreck it.

Please watch for the ways that others profit unidirectionally from your choices. Modern big tech is now comprehensively tainted with bad-faith justifications and disregard for the harm it does to democracy and the planet.

why not let the hard-working users control this feature, instead of implementing a system to scrape their knowledge (along with who-knows-what-else)?

I really wish I had the energy to explain why this is a bad decision

What access will google have to your (our) data? Will user be able to opt out? If not, when will this take place and what are the steps to remove all our photos and data?

I don't like this at all. I don't use anything Google I can avoid for a reason. Many others offered good arguments why this shouldn't happen on iNat. I will add to their concerns by saying that iNat needs to be human driven. AI should have no place on it ever. I understand the value of grants, but they mean nothing if what makes iNat so great is destroyed.

I will add that I too want to opt out of Google having my information.

Once upon a time, the Wikimedia Foundation (the organization that runs Wikipedia) accepted a grant from Google to implement an AI feature called "Computer-aided tagging". The project was hated by the community, gave poor quality results, and was ultimately canceled. All 1.3 million edits made using the tool were reverted. Even worse, the Foundation lost a huge amount of trust and social capital with its community. I hope iNaturalist can avoid a similar fate, perhaps by using part of that $1.5 million to build a parallel human-curated system that will outlive the AI version. iNaturalist's (and Wikipedia's) biggest strength is its army of human volunteers and their freely-given expertise. Leverage and empower that army and you will succeed. Alienate it and nothing good will result.

Something best left in the hands of humans. There's loads of species identification guides out there. If there were a category on species page to add an identification guide, people would do it. And at least we can scrutinize and edit what we humans put out. What's to stop this AI from putting out wrong information and who's going to correct it? This is a bad idea.

Ridiculous solution when a much simpler solution that properly utilizes and engages the community is sitting right there. I suppose accurate information provided directly from experts doesn't pay over a million dollars.

I also much prefer the idea of a user-edited wiki: a collaborative effort by the community as an extension of the current tools. Ethical concerns of generative AI (including environmental concerns) are not unfounded and this is disappointing to hear from iNaturalist that it is jumping on the bandwagon. AI has been shoved centre-stage as the next shiny thing, with some sense of urgency and FOMO - all the more reason to pause, question, and keep a critical eye of its proponents.

I use generative AI to enhance software features in my work and am not inherently opposed to its use to do so, especially for a good cause. However, I do share the concerns of other commenters about whether and how my and community data and effort might be used to train models not controlled wholly by iNaturalist. I have always been excited to improve and expand iNat’s CV model, but I would not want to participate in a project that may be used by Google to enhance its own multimodal LLMs. I hope that the iNat team can share more details about whether iNat data used in any accelerator projects could potentially be incorporated into Google’s datasets for training. The application terms for this year’s accelerator cohort don’t specify that Google will be allowed to do so, but they do require that any applicant must “enter into additional terms with Google (to be provided),” which makes me wary. This blog post doesn’t appear to answer that question.

This is a lot better than I was excpecting, but I still firmly believe this should be at most an opt in feature with heavy disclaimers and limited implementation.

How do I opt out of my photos being used for GenAI?

I'm invested in iNaturalist. I have

• 3500 observations
• 4300 identifications, including for a number of taxa that I am the world expert on
• a dozen scientific publications that use iNaturalist data, including a major on about the impact of community science photos on entomology: https://academic.oup.com/aesa/article/116/5/276/7237936
• presented about iNaturalist at scientific conferences, where I encourage other experts to join and contribute

Generative AI is awful for many reasons, most of which have been made in other comments, so I won't rehash them. If this goes through, it will burn up the good will iNaturalist has with the community that has built up around it. Personally, I am so opposed to it that I will delete all of my observations and my profile entirely and refuse to use iNaturalist data in any future publications. I'm scheduled to give a talk this fall about it at a conference aimed at encouraging other professional entomologists to use it. Instead, I'll use my platform to lay out iNaturalist's missteps and tell them to avoid the platform. I used Facebook and Twitter for over a decade each but deleted both after they enshittified and turned towards platforming fascists. I'm not so invested with iNaturalist that I won't do the same again.

Coudn't agree with zygy more!

This is something nuanced enough that it should really be done by humans. LLMs have a long way to go before it can even come close to doing something like this accurately and even then I would have concerns. I also have privacy concerns and would like to opt-out my observations.

This is a slap to the face to your hard-working, dedicated community who made this app what it is. I do NOT want any of my submissions/photos used by Google. If there is no option to opt out, I will probably delete my account and remove all my submissions. This is so depressing,

See more community comments on the iNat forum
https://forum.inaturalist.org/t/what-is-this-inaturalist-and-generative-ai/66140

This could have been a wiki... 😞

A gen-AI summary is NOT accountable, and cannot be reliably linked to primary sources. This will result in scientific misinformation.

People post poisonous mushrooms on here all the time, for goodness' sake. Are those ID characters really something you want to leave to an AI?!

And if scientists have to validate all these AI suggestions anyway, just LET US WRITE THEM FROM THE START!

Given that generative AI notorious for giving incorrect information, this mostly seems like it'd lead to misinformation being spread via iNaturalist.

As an identifier it can already be hard enough trying to correct misidentifications when people don't listen to your explanations of why something can't be identified to species or is actually a different species. I can only imagine how much worse the situation would be if lots of people who don't know much about the species they've posted are being presented with an AI description full of incorrect information that they end up trusting because from their perspective iNaturalist seems like a reliable source.

I can only imagine how many people, who are new to documenting and identifying the wildlife around them and therefore don't yet understand the complexities of identifying many species, are likely to take these descriptions completely at face value, misinformation and all. This is likely to be an annoyance at best, and outright dangerous at worst. Many people are unaware of just how often AI generates incorrect information and aren't going to think to fact check what a website dedicated to wildlife has told them about that wildlife.

I'm appalled at the complete lack of foresight and integrity required to think that implementing this is in any way a good idea.

Will you allow users to opt out of having observation photos used to train Googles AI? If not, I may be done with iNat.

My questions:

1. Like others, I want to know what data, algorithms, and/or other benefits Google gets out of this.

2. What will be the environmental costs of developing and using the system? The blog post about the text search only briefly touched on this, but it sounds like this model would be unique and therefore much more resource-intensive.

3. How many users are still regularly using the image text search system? I, like I think a lot of people, briefly tried it out, found it mildly entertaining, and then went back to searching for observations the same way I always had. I understand that that was a different use case, but unless you make the generated hints opt-in, the image text search traffic seems like the closest way to estimate potential voluntary use.

4. While this might, in theory, be doable for taxa that are already completely represented in the CV model, we already have a big problem with people submitting excessively specific identifications of the one species in the CV model from a large group of similar-looking species or being overly willing to commit to a species ID in groups that can only be IDed to species at certain stages (as adults, when flowering, etc.). How are you going to keep the new system from providing confident-sounding explanations of why something is Species X that really describe entire groups?

• For example, Archips semiferanus currently drives eastern US caterpillar identifiers bonkers because it's one of the few tortricids (out of >700 species in the region) with caterpillar photos in the CV. Providing a statement that "it's A. semiferanus because it's a small caterpillar with a dark head and a pale green body" (just like hundreds of other tortricid caterpillars) seems like it would make things even worse. I think that figuring out a way to build missingness into the model and/or the ability to force it to provide above-species IDs for taxa that cannot be photographically IDed to species would be necessary prerequisites to generate reliable descriptions. These are things that identifiers have been asking about for quite some time.

Whether this should be done is an important question that partially depends on the answers to the questions I and others have asked. But as currently described, I'm not at all convinced that this can be done well.

As other commenters have stated, I believe only comments from users who opt into the project should be used. I make a lot of identifications with ID tips and have made a quick guide in a journal post to help people give better general IDs on caterpillars. I'm not sure yet what I will do with all those if this project goes forward in its current form, but I can link all those tips back to a published source or another experienced user who can do that. If I provide over-simplified or incorrect reasoning for an ID, I can track those IDs and comments down and correct them. Unless there would a way for this model to link its information back to the original source, I don't see how we could trust it on a scientific/knowledge level, let alone the issues with privacy, etc. raised by other commentors.

This seems to violate iNaturalist’s goal of community building and protecting the environment. If you do not provide the ability to opt-out, I fear I may have to leave the platform, and that breaks my heart. I know I am just one of many users, I don’t mean to sound self important, but I do want to add my voice to the conversation as well. I get that you need grant funding to operate a platform of this size, but this wasn’t the way to do it. Please listen to the vast majority of commenters who are concerned, upset, and frustrated with this move.

This is so disappointing. I've been thinking about this all day and I'm still having a hard time putting together words for how disappointing this is.

As an artist I have seen genAI seek to exploit and replace the skilled work of myself and my peers. I've seen companies and communities that were supposed to be for artists turn towards catering towards genAI instead of supporting the artists who have had their art scraped for other purposes.

It's deeply disappointing to see iNat falling into the trap of genAI as a replacement for uplifting and enabling scientists and forging connections between the users. But as others have said, I guess it's hard to turn away from the allure of 1.5 million dollars.

I do not trust genAI. When I've reached out to it in desperation to find information after exhausting my own searches, it either gives me nothing new or makes up and misconstrues information, confidently and with "sources" that don't actually back up its claims.

I would much rather put structures into place that empower users to help others learn how to identify organisms. Something that collates and links discussion threads for organisms would be really cool (imagine if iNat suggested you a few buteos, and then linked you to observations with comments discussing how to differentiate between the buteos offered?). User-curated wikis sound amazing (and again, what if when iNat offered you the suggestion of valley oak vs blue oak, if it could link you to appropriate user-created guides or pages?). What if we used the tools available to us to uplift and connect people with knowledge, instead of scraping it, pulverizing it, and reshaping it into something that may or may not have the accuracy or nuance necessary?

Human knowledge is not infallible, but people seem so willing to believe in everything that genAI says. at least when I learn something incorrect from a person, I am still engaging with the scientific community. I don't know.

While I understand the reason, I think the level of the knee-jerk reactions in the comments is not proper of a platform like iNaturalist.
I too am deeply tired of LLMs being shoved down our throats.
I am, also, a person in the exact field where I'm being literally asked to work towards replacing my own position with an LLM. I don't like LLMs. I also thought "Ah sh*t" when I first saw the forum post's title.

With that said, I ask you all to pause for a second and consider this before posting:

-Lots of comments are about "how about a wiki made from humans", about letting the platform's do the work. All taxa have an "about" section that links directly to the taxa's wikipedia article, alongside links to other relevant platforms with info on the taxa. If no wiki article exists, there is a link to create one. Did you edit the wiki articles for every single taxa you offered relevant comments on before to ensure the info is there? Is there any instance of an "This is XYZ because of these reasons" comment that you left without ensuring that same data was added to wikipedia or other relevant sources of ID info for the public? Do you think everyone else also ensures the About tab has all those details every time they comment? Do you realise this project aims to do that work that you, me and millions of others didn't do?

-What do you think your reaction would be if Computer Vision didn't exist yet and it was announced today? Do you think Computer Vision should be removed entirely, since it also tends to give out wrong IDs? Should Seek be binned as well?

-Did you take all possible actions on your side to ensure iNat doesn't have to resort to taking a $1.5M grant from a megacorp? Did you chose the copyright option that allows everything to be uploaded to GBIF? Do you donate regularly? Do you campaign for others to choose CC0 and donate?

-...If you did allow GBIF to use your data, do you realise that that action has already enabled Google to use it without compensation, permission or attribution as well?

-When you say that this will erode the community, have you considered that this use would appear a step before the community gets involved, in the same step that CV appears? Do you prefer for non-experts to have less information at that step, in case some of that information was taken from an user that was wrong before, because you think that iNat users (the same ones that were scraped for the data) should be the ones giving out this information? Isn't that circular reasoning? What prevents the future IDer from being "more right" than the previous IDer whose info was scraped?

-If you think threatening with leaving if this is implemented... What does this achieve? Does removing your contributions to science improve the science? Is "not making my hard work available to Google" (it already was) a good reason for making that hard work unavailable to researchers? Is making LESS good data available a good way reducing the enshittification of data?

I've often felt that it would be great if iNaturalist had a built-in framework for users to create, curate, and update keys for clades that have them (yes, there's the About link to Wikipedia, but Wikipedia itself isn't always an appropriate place to have an ID key, let alone putting one at the top of the article where it would be most useful to someone browsing iNaturalist). I'm glad to see that many, many others have also been wishing for such a framework. I think it would be extremely helpful.

As for this project - first, I need to emphasize: LLMs don't know things. They are trained on and operate on words, not reality; they don't have an underlying model of the real world that they are expressing with those words. They have no knowledge, and nothing they say has any connection to reality beyond statistical connections to words they've been exposed to.

(This is not about "AI" broadly - there are plenty of useful and reasonable uses of AI, especially in analysis and interpretation, such as with computer vision. One could imagine using AI to search existing user comments for discussions that help to distinguish species and linking to those discussions, for example! My complaint is about machine learning that aims to produce novel content that purports to relate to the real world, when the machines are not trained on real-world data.)

So, given that LLMs don't know things, are known to confidently claim falsehoods, and are often used to justify massive scraping operations and dehumanization of organizational structures, I don't trust the outputs of this project or the organizational outcomes that may come from it. I thank zygy for mentioning the Wikimedia case, which should be closely examined and considered.

I'd also warn about the material risks if the AI outputs incorrect guidance on the identification of venomous animals, toxic plants and fungi, or other hazards; people may see an explanation once and then never engage with the platform again, taking that misinformation with them even if it is later corrected.

Finally, I would like to make sure that Google (and by extension the MAGA regime in the United States) does not have closer, easier, or deeper access to my data than they already would through traditional scraping of public content. Opt-out options would be great.

Lots of comments are about "how about a wiki made from humans", about letting the platform's do the work. All taxa have an "about" section that links directly to the taxa's wikipedia article, alongside links to other relevant platforms with info on the taxa. If no wiki article exists, there is a link to create one. Did you edit the wiki articles for every single taxa you offered relevant comments on before to ensure the info is there? Is there any instance of an "This is XYZ because of these reasons" comment that you left without ensuring that same data was added to wikipedia or other relevant sources of ID info for the public?

@naelin - There are two problems with using Wikipedia as our repository for identification information: (1) Wikipedia requires its content to be cited to published sources. You can't just go onto Wikipedia and write "The Rainforest Rocket Frog typically has a white stripe extending from the eye to the groin." even if you are the world's foremost expert on rocket frogs. Unless the statement is cited to a published journal article, it will probably be deleted. (2) Wikipedia is an encyclopedia, not a field guide. You can't add things like identification keys and schematic diagrams showing people how to ID organisms. It's not what Wikipedia is for and such content will get removed.

Do not want. The strength of iNat and the reason I contribute is PEOPLE. The environmental impacts of generative AI are obscene. Making this move is only going to alienate the very people who have made this place what it is. Please reconsider.

I took the liberty of putting this thread in ClaudeAI (ChatGPT said it was too long to analyze) and asked it how many positive comments it could find. Only two could be considered positive (out of 47) and one of those was lukewarm. Since then several more comments have been added, all strongly negative.

I will add my voice to the negative. If iNaturalist goes through with this deal, it will lose it's soul. That is not an exaggeration; all integrity, all trust, all credibility will be gone. And for what? A measly $1.5 million? That wouldn't buy you a bungalow in San Francisco. I doubt it would keep iNat's servers going for all that long. If iNaturalist is determined to honk off it's user base, it should at least ask for a decent price tag.

GenAI is pure poison. It corrupts everything it touches. It's so-called "hallucinations" (when human's do it, they're called mistakes) are a problem that isn't going away. I dread to think what sort of rubbish will get produced (and then copied uncritically across the web and even in scientific journals) if this project goes through.

I am not a scientist, but I do spend a lot of time photographing things I find in nature (ok, mostly birds). For the things I am not as knowledgeable of, I do find the image based machine-learning algorithms to be useful in getting close to what it might be, and then hoping an actual human expert in the field can narrow it down.

LLMs on the other hand, are not specific machine learning algorithms designed to identify species from images, they're designed to work with Language (it's in the name) and they're not particularly efficient or good for the environment when they do so.

I have absolutely no interest in engaging with LLMs or the hype-train around it. I've enjoyed my time on iNaturalist, and hoped I was helping researchers with their work, but if iNat goes down the LLM route, I fear my time here, and my field observations here may be at an end.

Learn from the DuoLingo fiasco, people don't want LLMs in these fields.

Destroying the natural world with AI while teaching people to appreciate the natural world? No thank you. And I'll be cancelling my monthly donation

I strongly oppose this project on both ethical and practical grounds.
In the walks I lead and the classes I teach, I often promote iNat as an exemplary piece of tech that really does help connect us to the natural world and to each other. In this respect, it is almost unique. Instead of singing its praises, I fear I may soon be warning of its sloppification, its unreliability, its dehumanization, its cooptation by deeply troubling corporate forces, and its abandonment of exactly those things that make it work - community, trust, and human expertise + accountability.
iNat is precious to me. This is a disappointing and frustrating step in the wrong direction. I urge the staff to listen to its userbase. 1.5 million is, all things considered, a really low sum to sell out for.

Why not just use AI tools (or some other search function) for gathering possibly useful comments, and have humans do the synthesizing and decision making? Users are clearly willing to work on these identification notes. GenAI feels out of place and unnecessary here.

I might add one other point. Of the three LLM families I've been exposed to (ClaudeAI, ChatGPT, and Google Gemini) Google is by far the worst. Compared to the other two, Gemini is a sad joke. INat didn't even pick the right devil with which to do a deal. To quote Sir Humphrey Appleby in Yes, Prime Minister: "If you're going to do this damn silly thing, don't do it this damn silly way."

@silversea_starsong Don't kid yourself. This particular project may seem rather inane, but it's the camel's nose in the tent. It wouldn't surprise me if this isn't stage 1 of Google turning iNaturalist into "iNaturalist, an Alphabet Company."

I am an insect taxonomist and like to lend my area of expertise to iNat as well as post new geographic records I find. if you go through with this I am going to delete my account. I will spend any time or effort or contribute my photos to a braindead plagiarism machine.

The group that I specialize on, scolytine beetles, is already constantly misidentified by the algorithm on here. Adding AI text explaining its wrong answer is only going to increase confusion.

iNaturalist is one of the last remaining places on the internet where people's actual expertise and learning is celebrated.
Using LLM tech completely undermines the whole point of iNaturalist.
iNaturalist has such amazing people power and community. LLM's take the people out. it's the entire point of LLM's.
Microsoft also just released a study sayign that chat gpt is turning our brains to slop so...
https://www.microsoft.com/en-us/research/wp-content/uploads/2025/01/lee_2025_ai_critical_thinking_survey.pdf
iNaturalist does amazing in getting people involved, while LLM's encourage people to stop thinking and stop valuing.

I work with prestigious gardens in my area, and local ecology groups, and local First Nations groups. I won't be using iNaturalist if this goes ahead.

People's IP deserves to be celebrated by community, with real valuable interactions. Not by data severs.

In trying to think of why iNat's leadership would make this kind of choice, I imagine the possibility of grant money is an important factor. Volunteers do great work and would compile an excellent ID wiki - the team has to know that. But volunteer hours won't fund the technical creation and maintenance of the wiki by staff. iNaturalist does accept donations, but perhaps that's not enough to sustain feature development.

I wonder if perhaps, as an alternative to fundraising with these sorts of partnerships, there might be better ways to encourage donations? Maybe by helping users see the effect of their donations? I don't have any perfect ideas; maybe if there were funding targets tied to deliverables ("We're at 120K of our 250K goal for funding a community ID wiki!"), that could help generate more excitement for donation drives, as well as provide opportunities for outside partners to donate to things that are specifically useful to them from the site. Crowdfunding campaigns do this a lot, to seemingly good effect. Or perhaps iNat could have a few distinct pools for people to donate to that fund different aspects of the platform.

I'm trying to figure out why this deal made sense to the iNaturalist team/leadership (as opposed to many other proposed or in-progress feature changes that seem to see little movement), and what alternatives there are to meet the goals the iNaturalist decision-makers are pursuing here. I think community upset is important to express, but I fear that it won't amount to anything if the decision-makers' motivations are substantially different to the community's.

@eddieroosenmaallen to be clear, I am completely, 100% in favour of changing the default content of the "about" section to at the very least include a user-driver identification guide/resource recommendations/etc, (though I am also quite sure it would be way less useful to folk outside the USA than inside).
I also understand Wikipedia it is not a field ID guide and not always a good place to put some info, but it IS good for a lot of relevant information, iNat can be and is used as a source (primary sources aren't even allowed there anyway), and more importantly it is what is implemented in the "about" section at the moment, yet iNat users don't usually fill in even the most basic data in in most cases.

What my point is is that a lot of comments are about "Let people do the work!" ignoring that there is absolutely no way for "people" to do this work at the scale needed, because we just WON'T. The reason why iNat is aiming at getting the info from comments on IDs is merely the strength of Cunningham's Law: They know they will get much better and much more info from people correcting others than from asking people to proactively answer questions.

Asking the users to manually input field ID data is eons worse at scale than taking the existing user data, process it automatically and then asking users to correct what's wrong. Cunningham's Law again.

@quillipede why do you think aggregating user comments is "bypassing the community"? No community, no data to aggregate. Hell, iNat is literally a data-aggregation platform. If the data processed is aggregated from iNat, it is community data.

I am strongly against the incorporation of an LLM into what has been my favorite website for nearly 6 years. As a budding taxonomist and "power user," I would be happy to lend my expertise to the iNat wiki if developed. I see very little benefit to this partnership, and I do not think that it will have the positive impact on identifying that iNat seems to think that it will.

Edit: If this is incorporated, against much user input, at least give curators a way to manage misinformation.

@naelin I think it's unfair to say that people "just WON'T" do the work when we haven't even been given a chance to.

A user lead wiki would be a fantastic way to do this! Its also something users have been asking for for years!! I know there are forum requests about exactly this! I think using an LLM to do this instead of our amazing community disparages the hard work our many wonderful IDers do. Why have an unaccountable robot do the job our thousands of IDers would do better and build our community even stronger. This is a massive mistake and I really hope the staff understands how much many of us really dislike this decision. I feel like not only was this a spit in the face to our community, but shows a disregard for how we as a collective of academics and regular people are important in the operation of iNat as a whole!

Yes iNat uses a learning algorithm to teach it's "AI" to give ID suggestions for what people see. This is EXTREMELY different from a generative AI (LLM) which are not peer reviewed and are known to take shortcuts in information. If the differentiation in ID for a springtail species pair (a group I ID) is 2 setae vs 3 how can we guarantee the AI can take these notes into account properly. There are thousands of obs of the Gray/Copes tree frog pair and MANY obs say cannot be IDed without a recording. Will the AI know not to suggest this on an audio observation? I think its just too unreliable and will create an even bigger mess of work than already can stem from species entering the CV which maybe shouldn't. Among springtail IDers the CV is notoriously bad at IDing many springtails due to the high quantity of low quality photos feeding it (Not an inherent issue on anyone specifics part springtails are tiny and those using mobile devices can only do so much). With just this however it leads the CV to call anything vaguely low res and gray/brown a springtail. I fear that incorporating generative AI to this would only compound the headache we get from just this. I love the CV, but it has its flaws. Generative AI has many more and for a more scientific setting where data accuracy is vital for our partners trust in us (ie GBIF) it is imperative we shouldn't just let an LLM overconfidently contribute to ID processes. I have seen long and vitriolic threads on bugguide where experts outright hate iNat because it feels too untrustworthy in the accuracy of its IDs from the CV and less knowledgeable users who are overconfident in their abillity to ID. Using an LLM to supplement ID where a user led Wiki would be far superior will only serve to alienate these experts further which will deprive iNat of other more valuable ID resources!

Additionally I agree with the many other commenters that it should be opt out from the get go and only be available for users to opt in. I do not like the idea of google having a free backdoor to all my data. If their own AI's can access iNat data what will they be able to see? My home, areas I frequent. Will the locations of rare species susceptible to poaching be accessible to their LLM to give poachers a backdoor to obscured species? How much will they see?

I have been an avid donator to iNaturalist since 2019, but if iNat is going to use LLMs to further drain environmental resources (IIs this not antithetical to iNats purpose and goals to begin with??) I will likely stop contributing to the upkeep of the site. I may even look for another database to house my photos and data. (SOMETHING I REALLY DO NOT WANT TO DO)

PLEASE RECONSIDER THIS DECISION!

@blazeclaw I agree. This decision will drive away experts and prevent others from becoming identifiers here, even more than has already occurred due to lesser data quality management tools. Completely agree with your last paragraph, as well.

@edanko Fantastically written and all more great points! Please iNat staff JUST LET US MAKE A WIKI!

I hope you intend to put out frequent environmental statements detailing the energy use and environmental conservation of these systems. You also need to opt all users data out by default. I think this is a terrible idea that's going to at best lead nowhere and iNaturalist should be ashamed they allowed themselves to be bought and for such a small amount of money. These systems are not very effective and have an enormous environmental impact and Google is monopolist that has abandoned all its environmental commitments to pursue these technologies. This is such a disappointment.

This is so disappointing. You have been my one shining beacon of hope as US fed-funded conservation projects (like mine!) are defunded. I thought we had your support to continue science, not throw us into Google's AI woodchipper. Is this really what the spirit of our CC licenses are? You're taking steps that will result in past observations being deleted and future observations being untrustworthy.

This website depends on the labor of experts. You're saying our job now is to babysitting AI that will always have random hallucinations and failures. Why are you making our volunteer job harder?

Quite frankly, inviting gen ai imagery into this website is celebrating the downfall of inat: false observations. If you welcome genai images here, you are welcoming genai """""observations"""""". Once we lose the ability to trust observations, inat loses its entire reason for existence.

Respect your contributors' labor, respect the core spirit of iNat. This is not it.

Who asked for this?

I'm optimistic, simply because I have seen for years that iNat staff like @loarie value nature, thoughtfully ID observations, and build up taxonomic knowledge with our community. I am hopeful that they, like us, have high standards and will make decisions based on the same values. In other words, I assume that the iNat staff have already taken into account all the problems that have been brought up above, and plan to use the AI in a role where these problems are irrelevant. When there have been trade-offs in the past, as with the rise of unreliable CV suggestions, I think these trade-offs were still made based on the overall preferences of iNat users?

Some notes on making a wiki, that are also relevant for an AI that has a wiki-like output:
When it comes to explanations of how to separate species, citations are critical to demonstrate credibility. If I share a diagnosis, it's because I've either seen on my own that it works or I've put my trust in another source. Since I can't simply trust AI or a wiki, the credibility of AI or a wiki is contingent on the quality of their citations. If the AI or wiki employs citations, this also allows others to reproduce information shared by the AI or wiki and give the credit to the source cited by the AI or wiki.

When I give someone a set of characters to use for ID, I am saying "@edanko looked at literature and/or specimens, and the bottom line is that @edanko found adequate evidence that these characters are reliable." For example, let's say I once noted that a certain species was the only species in a certain group in North America with orange legs. Perhaps I was not the first to discover the leg color, but I was the first to point out that the leg color was a reliable way to distinguish this species in live individuals that have not discolored from being preserved. If so, the 'citation' for this character may simply be "unpublished work by @edanko".

I'm also a little confused--are the iNat staff talking about an LLM that reads iNat comments/text and acts like chatGPT, basically? This seems to be what others have assumed above.

Alternatively, are we talking about actually improving the CV interface so that it can speak for itself, so that instead of the CV saying "We're pretty sure this is in the superfamily Dendrobatoidea" it says "We're pretty sure this is in the superfamily Dendrobatoidea because body-part-so-and-so is shaped so-and-so"?

Pasted from forum and edited. I dont have the energy to explain in huge detail, but i stand with what seems like the consensus of my fellow inatters like @zdanko @blazeclaw @humanbyweight @zygy and so many more, (i recognize so many names).

AI should not be entrusted to give detailed explanations on how to ID species, (this is different from expanding the CV to explain itself. Although it is pretty much just an image matcher). 80% of Chironomids haven’t even been IDed on INaturalist. How on earth can it be expected to generate anything close to accurate when it may only have access to 5-10% of information for species of the group?

I fear this will just push many large identifiers away making issues around data quality, misidentifications, correct identifications, and community health worse.

Edit. I can try to have faith implementation is good. But its just hard. In large part becuase the CV itself still has so many issues and creates many challenges already for certain identifiers.

@edanko - The computer vision models don't know why they recommend an identification. Human input is the way to get such answers.

I am completely opposed to the incorporation of generative AI in any aspect of iNat.

As this is a user based platform, let's keep things user based and not AI based.

Using generative AI for creating species descriptions is an absolutely terrible idea given how badly this type of AI hallucinates and creates outright false information, even when making 'summaries'.

To my mind this move is diametrically opposite of the entire point of iNat, which is to foster more human engagement with nature and learning about nature.

From my post in the forum:
So rather than take a step back and look at the overwhelming majority of iNat users explaining their opposition and concerns to this, iNat has decided to double down on it. I feel like absolutely no questioning or research was done on this. It doesn't take much looking to see what the average user of iNat is and how said users feel on gen AI. We're not Facebook or Twitter users who suck up to it, we deeply oppose it.

The stated reasoning for using this AI is something humans have done for centuries: Identification guides. Why not have a section on species pages where we humans can add that ourselves? Human writing can be vetted and easily edited. Most people would rather no information than unconfirmed data that comes from gen AI.

It's the same reason it's dumb that iNat pushes heavily to get an observation down to the species level. Many observations simply have bad photos and therefore bad data to get there. If a human can't ID it, AI definitely can't either. And this comes back to the whole identification guides. What happens if a species only has single digit observations? Gonna make an assumption on that?

Like I said earlier, species guides exist already that date back centuries. I'm sure people would make guides on iNat if you added the page for it on species. That way it'd be done by people who are actually knowledgeable in the species or have proper data. I imagine it'd be a much easier thing to program than AI.

And finally, it's frankly disappointing that this wasn't brought up with the community or funders. It seems that once an organization reaches a certain size, transparency deceases and communication slows down. I was seriously hoping iNat would be different since it was built and run by the community for the longest time.

Please, just abandon this project like people are begging you to. Many people are deleting their accounts over this, which harms data even further. Nobody wants gen AI or Google associated with iNat or our observations. Read the room, back down.

Seems i'm going to regret having explained to community members about how i'm making more precise identifications on observations as the "Google.org Accelerator" text read "iNaturalist enhances biodiversity data by converting thousands of identification remarks into natural language explanations."

That said, i'll initially go against the trend above to state that i'm very glad to hear a core of this initiative may revolve around drive to present the "WHY" with any proposed identification (and therefore with any name potentially link to a wealth of data from that). If grand initiatives like EOL had not had the typical boom and bust cycle, then maybe this system could have looked there as once tried, but iNat must continue the push in its own direction and increasingly shows us a leading role.

The comments about a user driven wiki based approach seem to be something that's really worth investigation, at worst as a parallel source of ideally high quality user generated information that the exploitative AI can use as one of its training sources. I don't see how any the existing core wiki sites provide scope for what seems desirable to many above, yet a parallel iNat linked initiative could. Another past scheme that existed a few years ago was "Species-ID" (then on species-id.net/wiki) but now i see only parts through Archive, e.g. Wayback machine. Plenty of things since have developed about extracting out core text on diagnoses and identification from open access taxonomy works. There's much scope to now actually develop something unify that information, but i'd hate to see another thing being thrown in that again lacks much human intervention - especially as iNat already has that giant, engaged and often highly informed community. How many replies above? The interest is there.

Anyway, the lack of on the "WHY" for identifications is perhaps the biggest thing i see lacking in iNaturalist to date, hence i welcome any steps towards that. Many users want to learn about their observations, so improve their knowledge of nature. However, in terms of the role of AI, then I'm another who is still very skeptical and even negative - on team luddite. As someone who has tried to share my knowledge from my own study and education (gathered through significant time and monetary investment), then i see how this can be viewed as another development that exploits the once gifted collective expertise across the community. Educating others is of course about knowledge sharing, but i'm not keen on exploitative AI doing that (still unpaid) job for me, and worse, it producing terrible results (as generative AI overview does on anything taxonomic from that system's search engine results). Of course the modern world is still in its infancy of what AI can do, but if the next steps on iNaturalist are going to be led by scraping existing content, then i'm expecting low quality results for a fair while and alienate many. I fear that will only generate distrust from many users and bias what was previously freely said. Worse though is my concern how my own words will be used to produce awful output, for example, I've often used simplified approximate wording to avoiding precise and complex terminology that many casual users often seem intimidated by (in hopes to be more understandable to the widest audience). E.g. a hypothetical case from the specialist taxonomic literature of "Retrolateral margin of tergite IV with parallel rows of pearlescent nodules" might become (as my comment under an observation on iNat) as "there's a special arrangement of little bumps on the sides of the abdomen". Even then i'm doubtful some readers will know what an abdomen is! The point of these final words is for the informatics staff to think about the quality of scraped information that might be pulled by automated systems and fed into any potential AI, the standard problem of Garbage in, garbage out (GIGO). The AI system could present users with a reformed AI approximation of my poorly public "special little bumps" or it could help users to determine accurately what "pearlescent nodules" are and where they are on the creature. It's your call on which way to go on this iNaturalist!

@naelin I appreciate your intent to talk nuance right now, but I don't think this is being taken as urgently serious as it should be. Nuance needs to be worked out before hand, not after the project is announced. People aren't waiting for nuance and clarity to delete their accounts. They're just doing it. This is blowing up on Bsky right now and I'm horrified how quickly things are falling apart.

That's cool. I hope you're ready to see users mass-deleting their observations and shutting down their accounts en masse as a result of moving forward with this.

Google gave up on their "Don't be evil" motto in 2018, and this year in 2025 they removed their pledge not to use AI for weapons. They are collaborating with fascism, with war, genocide, and occupation.

I will delete my data and my account before I will have my data used in any way by anything associated with Google.

We want human-made wiki from experts. AI can never and will never (atleast not any time soon) replace human skill and knowledge.

The amount of comments already existing makes it a moot point to rehash just how bad of an idea this is. But I wanted to throw in my two cents, ever so briefly, on one particular fact. It seems rather counterintuitive, even antithetical, to support the usage of an LLM-based program, one that would need to be so incredibly specialized that it'd no doubt require insanely intensive resources, when iNaturalist purportedly seeks to "[...] advance biodiversity science and conservation." This is not beneficial to science. This is not going to advance conservation.

In addition, the implementation of an LLM to "[...] support the hard work done by folks making IDs[.]" is directly in conflict with one of the primary stated characteristics of the site, which directly from the About section, states, "iNaturalist is an online social network of people sharing biodiversity information to help each other learn about nature." The keyword is people, here. People. Not an LLM spoonfeeding information that is, more likely than not, going to be wrong. iNaturalist is a community at its core, not a piece of technology. If we are so inclined to resort to the LLM hype, then I believe it'd warrant the removal of that statement. Not only does it demean the hard work of identifiers by taking hard-learned knowledge and watering it down, but also, there's no realistic way for this to be implemented and not go horribly wrong. Not everything can be identified to species, but just like CV, an LLM will have the goal to do just that. I worry that this misplaced confidence in identification by an LLM will only serve to poison data even further.

Guys, NO AI. I work in data. It’s such a bad idea to use AI. Please, don’t use that slop.

Respectfully, this is completely unnecessary and I hope you reverse this. If we want to know the reasoning behind an ID, we can easily just ask an actual human being. iNat has tons of professionals that will gladly offer much better explanations if asked without the price of our data. The cons strongly outweigh the pros, anyway.

iNaturalist’s vision model to identify flora and fauna is incredible.

Google’s attempts at generative AI have so far been misleading and dangerous garbage. Suggestions we use glue to hold cheese on pizza — an actual result from Google — are easy to spot.

Generative AI will potentially lead to more dangerous results with unacceptable consequences.

I would like to opt out.

This strongly needs further clarification. My opinions and thoughts change on what is actually happening and is "the plan".

Are user comments being used to train an llm? Is the CV getting an upgrade so it can talk and explain why it thinks its this species?

What is the plan?

This is fracturing the community, we need to have clarification.

I kinda surprised to see such comments on site which main feature is a user-trained computer vision system(virtually an AI suggestion already?)...
Could anyone explain how it is different from generative AI?

Or everybody just hates Google and LLM and that is the trigger?

there is very clearly a transparency issue here. I recognize a lot of names in this comment section. a LOT of names. a lot of power users expressing their distaste for these changes, vowing to pull back from the platform, or even delete their accounts entirely. this is not great! I also see a lot of misunderstandings of what is actually going to be implemented, here and in the forums, but also especially on bluesky. I won't delude myself into thinking that it's a majority of users that will be deleting their accounts in protest - most likely, most users won't ever have a clue this is even happening. but very vocal, very involved users have some STRONG feelings about this, and I think a follow-up announcement with a LOT of elaboration is warranted. this isn't irredeemable. but this is a bit of a bungle, y'all, and putting in the work to regain trust would go a long way.

basically, iNat need to do damage control, and fast. you need to make extremely clear what, exactly, the plan is, and how, exactly, the data are going to be used, who will have access to it, and for what purpose.

or just carry on and pretend nothing is amiss. which might be convenient enough, and it might "work". but it would make me and others really sad and disappointed.

full disclosure, I'm not the biggest fan of this news, but I am not doomsdaying, nor am I planning to pull back from the platform. I understand the implementation proposed here is pretty "light". but cmon, y'all. read these comments. read the forums. people are freaking out, and given the state of, well, everything lately, I cannot blame them. giving some reassurance and at least acknowledging people's fears here would go a long way, I think.

TL;DR maybe read the room? or not, I guess. 🤷🏻‍♀️

I really dislike the thought of this feature. I'd greatly prefer a wiki instead.

I think if this is indeed being implemented in a way where Gen AI adds photography tips in conjunction with CV akin to 'CV is confident this is X genus, which are typically then identifiable to species by Y feature(s), try making sure those features are visible the next time you photograph this organism'. Or 'The distinctive X feature is visible, which indicates Y species is likely.' this is probably a good feature that compliments the existing CV, which is not perfect, but is a good baseline for those who don't have much experience identifying what they are photographing. One wonders if using Gen AI is actually the best method for implementing such a feature though.

I think iNat needs to be really cautious with how they announce and utilize Gen AI solutions and make it very clear that an AI will not be making authoritative statements on what is criteria for species identification. These are not popular tech solutions that are prone to inaccuracy and will make many people upset when implemented. I definitely had a very strong negative reaction initially, but I think if the scope is limited to just compliment CV this COULD be OK. I agree with the majority of the concerned posts here as well, lots of potential for this to be a bad move, and probably far more elegant solutions available to implement this feature than reliance on Gen AI.

I'd like to ask iNaturalist to update its Terms of Use (https://www.inaturalist.org/pages/terms) to more completely address user contributed content. At present, the Terms grant iNaturalist a license to utilize "Content" in certain ways related to "Your observations and journal," but they appear to be silent on Comments and Identifications as forms of user contributed content, and the ways in which the platform may "reproduce, modify, adapt, and publish" them. I think given this planned new direction, an update is in order.

wow. i really used to like inat.

time to find something else, i guess.

This is so disappointing. The way you used non-generative AI in the past made total sense. I don't want any generative AI involved in any of this! You shouldn't trust google to have your best interests at heart. They definitely don't have OUR best interests at heart.

It's interesting that inat wants to help make ID's more reliable but at the same time an admin censored my post and cut off comments after only 5 comments (btw from some of the most prolific posters and observers and knowledgeable people in the field) one of which was there are too many bad photos and another too many incorrect ID's on a huge amount of moth posts. Certainly a slap in the face to some of the most prolific contributors here. One wonders if admins have any standards themelves or can any one of arbitrarily stop comments and/or block or suspend accounts without any due process. Quite concerning. This is absolute censosrship. I am unaware of any violation involved. Quite the contrary it was what we call constructive criticism. I guess that is not allowed here. Should I be worried that my account will be closed becaue some random admin didn't like my comment? It's happened once with me already and the guy is still an admin. I'm surprised you have not cut off the comments here as most are quite negative yet you cut off mine and other after only 5. What hypocrisy. lol

@john_ibis the majority of the comments on this blog post are criticism. Infact some are not even constructive and are destructive criticism.

I think people threatening to or actively deleting accounts, esp. those with tons of contributions on the IDs made end are kind of overly hasty. I know the whole AI buzzword has kind of become a bogeyman, especially on certain ends of social media but like. lol iunno. That being said? This just kind of feels like a nuisance feature at best that will increase the amount of wrong IDs made via CV, confirmation bias from a text snippet and all, especially for taxa that the CV struggles with (and with these groups in particular it seems that experts on them often tend to disregard inat as a worthwhile source of data and thus the site remains a poor resource on all fronts for them without anyone even attempting to correct ids). Plus if this feature is going to shake out the way that it sounds like it will, you're going to need curators to go through each description for all but the most visually apparent and well sampled of taxa with such a fine toothed comb that you may as well just have it be user made in the first place

Flippant comment (but with intention to help things keep focused on identification issues): If only biological taxonomists had been publishing diagnostic keys for identifying creatures for several hundred years! Imagine what a resource that might be?!?

I really don't have time to read through the mass of responses this blog has generated, but I do have a technical question:

@carrieseltzer Q: At this point in time, does staff have some sense of what the manifestation at the UI from all the GenAI text crunching will look like? Will this show up as supplements on the "About" tabs on taxon pages? Will it pop up when someone (initially or secondarily) suggests an ID for an observation?

Selfishly, I'd like to hear more about how the proposed new integration of GenAI will affect my user experience at the UI. That will give me more context about whether this might be a thumbs-up or thumbs-down proposal.

Please do not do this.

NOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
i think you as fellow scientists and naturalists are well aware of how bad AI is for the environment. Its upsetting that this being even considered.

Not read all the comments, but so many / most seem to be conflating issues ( or non-issues imo ) with LLMs and broader societal gripes which are irrelevant to this use-case ...and are just causing some sort of bias in their logic. It would be good to have this conversation on the forum instead where people can really debate points.

People concerned about information and sources from their comments being harvested seems particularly counter-productive and needless to me - if I harvested info and sources from your comments manually, would people be equally annoyed and threatening to erase all their IDs ?

Similarly people raising environmental cost concerns - I see nobody raising this already with regard to existing model training. The misinformation about LLM environmental costs is so widespread and largely misplaced...

Anyway, in general I think this sounds great to explore at least, provided users can feedback directly to fix incorrect summaries and sources as outlined in the post. I see no reason not to use generative AI for this - it may or may not be more or less effective / problematic than other options. I agree that a solely human wiki might be effective...but for example, if the wiki's were all generated by AI before going live on obs... then signed off by humans to activate...this could end up just a quicker way to consolidate existing information or even to scrape / analyse images to find differences which aren't even currently noted in keys. Overall, it really depends how it's implemented... but given the community concerns and the possibility for misinformation, it should obviously be done with due care.

No. iNaturalist doesn't need this, and your users don't want this. Google's Generative AI is ecologically destructive and unethically sourced. Don't tie yourself to it.

$1.5 million is barely the cost of web services for half a year. Stupid to do this for chump change. See for yourself from 2023: https://static.inaturalist.org/wiki_page_attachments/4009-original.pdf

I support the idea of implementing a user-maintained wiki for having identification and other info easy to see to all people on the site, that perhaps could be tied to the taxon pages themselves.

There were already 'Guides' but the feature is no longer supported/being developed. It's not well integrated or well visible on the site either which is a big issue.
https://www.inaturalist.org/guides
Wikipedia wouldn't be great for this either, as it has a completely different purpose and writing style.

@sbushes Just curious, are you familiar with who Andy Masley is and what the organization he represents does? It is deeply embedded with the AI tech crowd, to say the least.
At any rate, here is a detailed takedown of his article and why it is completely false: https://portal.mozz.us/gemini/gemlog.xgranade.com/2025/05/16.gmi?

Please rethink this model. Use a community wiki or use a platform that keep you as data owners (there are several options to keep owning your own data while using ready made models, check databricks). About the money, crowd found it through this community

Google took the thing they were known for (websearch) and made it bad. I've been telling friends for months not to rely on Google Lens for IDs because iNaturalist is so much more accurate, and that accuracy comes from a well-vetted dataset and an army of volunteers willing to help with corrections. Adding Google to iNaturalist may improve Google. It definitely won't improve iNaturalist. I want my sightings to benefit scientists, not megacorporations. I certainly won't spend more time reviewing sightings if Google is involved. They can pay for that labor if they want to use it.

@danieavilao INaturalist has used a form of AI since 2017. That is the Computer Vision.

At this point I think staff also need to do a Q&A.

geothlypis philadelphia

@thebeachcomber I agree. As upset as I am, I'm not going to be deleting anything until I am sure there is absolutely no other option. I hope others will do the same. With how the world is right now, it often feels like every worst-case scenario is inevitable, but it's important to remember that things can turn for the better, sometimes.

Emotions are high amd it is understandable.
But @jonvalliere, let's please refrain from the harsh expletives. I'm happy to say i haven't seen any truly explict guideline violations (like people insulting one another), let's keep it this way.

Many of us are indeed stressed, the @thebeachcomber has a very good comment above about taking a break to let the situation settle and play out. We will likely hear more from staff at some point soon.

@thebeachcomber

Many of the comments here and on social media, eg BlueSky, are wildly speculative or outright completely wrong, and that is only serving to inflame the situation.

Yes, ironic that so many are reacting to the prospect of genAI involvement on iNaturalist by mimicking its worst attributes.

How hilarious to suggest you protect users photos with copyright then just sell all of them to Google ai. I expected this obviously but I dared to hope you wouldn't sell all the inat users out like this.

As a mega user of iNat from the under-resourced part of the world, I fully welcome this initiative, and look forward to seeing how it evolves. Even if it does have major faults it is a move in the right direction, and will be an improvement on what we have, even if it is imperfect for the next few years.
My request though is to please include subspecies and varieties into the AI. That the AII stops at the species rank does not help conserve and understand the huge biodiversity and biogeographical variation that exists below the species level.

I would like to share my concerns about this, not re-iterating all the problems of AI that have been already stated. I acknowledge the positive things that can arise from AI, but this whole thing seems sketchy. Which data does Google get access to? Will it use the data to enrich its own models or do the data and algorithms stay strictly within the iNat infrastructure ? ( I suspect the former and I strongly disagree with that). Is it really necessary to use these technologies that are very bad for the environment? I know a lot of people that have already decided to leave the platform just based on the announcement, but I hope that you will at least go for an opt-in model so we can all reconsider leaving and deleting our observations.

Yeah, this seems like a waste of time and developer energy that will cause more problems than it can ever solve. Setting aside my reservations about the emissions produced in the creation and use of such a feature, reliability and accuracy will be a major issue. The CV model is impressive, but it still makes frequent mistakes, both major and minor. This proposed feature seems like it will only supercharge the current issues with CV accuracy - if this feature is going to be trained on a dataset of existing observations, it will absolutely exacerbate CV misidentification. This is of particular concern when it comes to poisonous and venomous species where misidentification, and providing misleading identification instructions in particular, could cause real harm to users. On top of the risk of misinforming users, giving feedback to try to correct bad identification information will create an extra workload for staff and dedicated users. It sounds like a lot more trouble than it's worth.
There's pretty major community opposition to this proposal for many reasons, and I hope iNaturalist staff will take these concerns on board.

If iNat is seriously keep pushing this terrible idea despite almost every users are against this... It should be opt-in, not opt-out. if users want to feed their data into generative AI and use this feature, they could opt in.
There's so many negative impacts of generative AI, not just only energy consumption.
I hope iNat will not ignore the devoted community.

On top of the comments I've already left in the forums, we need some transparency on what kind of data you're sending to Google as part of the deal. We also need to ensure that this is a program that users MUST opt-in by choice. Retroactively opting everything in and requiring users to opt out is a breach of consent.

When we signed up, we consented to iNat using our data to train your CV to help biodiversity mapping, we never consented to contributing more towards Google's database of stolen content. Just because Google has done it behind our backs anyway, doesn't mean we liked or agreed to it, we were just powerless to stop them.

Knowing iNat is going to enable Google's disregard of consent is heartbreaking to me, considering I'm a visual artist with author and musician friends who have already seen firsthand how their creations were stolen, butchered, and regurgitated into frankeinsteined Gen AI outputs without their consent.

As a power user with 11k observations and recurring donor, I'm genuinely upset by this vague announcement. I know you're probably still exploring your options at this stage but I hope you understand our concerns.

Given the sheer amount of people here, many of which are experts their respective fields, who would be willing to contribute to something like a community wiki, I cannot see any reason why anyone at iNaturalist can defend this. LLMs, Gen AI, whatever name you wish to give it. They’re all very dubious in legality and not something many communities want to put up with. And with something like the identification of various forms of life, the fact that Gen AI hallucinates so much on top of everything else seems incredibly counterproductive to being “accurate.”

I’ve used iNaturalist for a few years now and I’ve been using it to try and document the various insects that are in my area. Being in a small town, it’s something that I love doing. It genuinely makes me happy to get out and look for what lives around me, and then taking part in something important like documenting what exists here. I don’t want this to take my faith away in this service because I do not want any of my pictures being fed into Google’s AI. It won’t help with accuracy. It will just make it harder for people like me to catalogue what they see. Will make it where people are reluctant to come here because they won’t know if they’ll actually be getting real information.

I can’t say I’m the most eloquent person here. I’m just a random artist who likes to take pictures of bugs. But my thoughts mirror a lot of the concerns that others here have. It would especially be nice to know what is being sent to Google in the first place. How do we know they won’t use this to gather images to feed into their image generator? How do we know the comments and research people post won’t be put into Gemini to spit out generated text? This just isn’t a trustworthy scenario and it doesn’t seem like it’s worth it in such a community driven environment in the first place.

Extremely disappointing, terrible news!

I don’t want it either

Did iNat choose the grant or did Google give it to them bcoz it seems like Google gave it to them.

Generative AI is actively worsening climate change, which is in direct opposition to the conservation work many of us do here. Your choice to use it is an insult to the intelligence and very existence of iNat users.

Google destroys everything it touches.

Generative AI is unreliable and will result in untrustworthy and inaccurate conclusions. Didn’t Wikimedia Foundation do something just like this and it was a total flop wasting millions of dollars and losing users’ trust?

This is hugely disappointing. Like the vast majority of users above, I do not want my observations or data to be used for this purpose.

There is already a lot of good feedback I agree with here. There are pros and cons to this project, but I believe there are many more cons.

User blazeclaw rightfully said
"I have seen long and vitriolic threads on bugguide where experts outright hate iNat because it feels too untrustworthy in the accuracy of its IDs from the CV and less knowledgeable users who are overconfident in their abillity to ID."

I work with insect taxonomists, and the best of them hate iNat for the reason stated. They prefer BugGuide to be more reliable for checking images of a species, since the iNat information is not as curated and reliable for research use (as pointed out by user tworm above).

Team iNat, your strength is your community of volunteers as many have said here already. Respectfully, this decision should be reconsidered. The same mission could be accomplished with a curated wiki of experts helping explain the difference in taxon. Perhaps if you perfected that, then you could implement a system to have the app/website refer users to the specific page and also include it in search results.

Make a human-editable wiki to collect and display identification guides like this. Generative AI flies in the face of the "Gestalt of human expertise" that naturalist used to allow

We urgently need the complete text of all agreements entered into between iNat and Google.

We also need a full explanation of exactly which data are being exchanged and for which purposes.

I did not consent to and will never consent to my data or images being used without compensation by Google to train gen AI models, or for Google staff to use my data to learn how to train gen AI models.

iNaturalist needs to hit an immediate pause on interactions with Google and do a MUCH better job of explaining things AND getting the iNat community's buy-in before proceeding.

Oh, so sad.. I've happily been an iNat. user for almost ten years, and while sometimes irritated by the AI ID gross mistakes when suggesting an ID, I DO NOT WANT Google involved, unless it is a Grant to use in ANY WAY (Such as promoting Curators doing more to explain IDs, etc. ..People-based, Community-based.) I echo many of the concerns voiced in the MANY comments preceding mine, one in particular:

"I'm not sure how a collaboration with Google (an organisation whose carbon footprint rose 48% in 5 years as a result of its expansion of data centres to service AI) aligns with the iNaturalist vision of "a world where everyone can understand and sustain biodiversity". Not to mention Google's removal of their pledge not to use AI for weapons, and the huge monopoly they have over search and advertising. I've removed Seek from my phone and replaced it with ObsIdentify, an EU based non-profit which isn't working with an organisation that has done more to damage the online experience than almost any other."
Posted by rgbartlett about 3 hours ago
It is COMMUNITY which makes iNaturalist attractive and USEFUL! Please, DO NOT do Google's bidding!

One of the things I used to love about iNaturalist is that it still seemed to be a proper social network (other than a citizen science platform), centered on its actual human users and still relatively safe from enshittification and environmentally catastrophic fads. Of course I should have understood before, when I saw there was no way to make the app work properly without the Google Play Services localization on, and no version of the app on F-Droid. I guess I will stop using it and recommending it to everyone.

Copying my message from the iNat forum:

In August, my iNaturalist account will be 10 years old, and I've been around for many of the site's changes through the years. Even though I can see that many commenters here share my views on the matter, I still want to leave my two cents for what it's worth.

Personally, I am against the use of generative AI such as described above on iNaturalist. In short, despite many benefits it could have for the platform (the speed organisms could be identified would be useful for people not particularly invested in wildlife, but that begs the question why they'd be on iNat in the first place), it exacerbates and creates new problems for identifiers for no real reason other than "AI" being the latest new and catchy buzzword for investors such as Google to profit from in the short-term.

Many others above have commented on the environmental and ethical considerations that should be addressed before something like this is added to iNat, so I won't restate what they have said. Instead, I want to further comment on a point brought up by a few others (e.g. @blazeclaw), namely that there is an intrinsic difference between how the AI used for computer vision is utilized compared to the proposed model. CV is a tool that gives suggestions which can be overridden by either the observer (if you know already know the species you can enter its name when submitting observations) or by identifiers once they pass the 2/3rds benchmark. This system already has issues, namely in the often large amounts of time and effort it takes for multiple identifiers to review observations of particularly challenging groups. While some discussion about what this means for arthropod, plant, and fungi identifiers has occurred, this would also negatively impact identifiers for groups such as mammals and birds. For example, many of the smaller bats (e.g. Pipistrellus) can not be identified from pictures without audio verification of their vocalizations, which has not stopped CV from listing possible species from images alone. I primarily identify birds of prey, and I often have to correct CV identifications from (usually new) users even in well-iNatted regions such as in the eastern U.S. (the CV loves to identify Red-shouldered hawk as Cooper's hawk and vice versa, not even mentioning Sharp-shinned vs Cooper's hawk IDs). Gen AI descriptions would add more confidence to incorrect identifications that newer users would accept at face value. This would increase the amount of misidentifications on the site (and where iNat data is shared including GBIF, academic papers, and now Google) and the amount of time and effort needed to correct these misidentification (accounting for individual observations and maintaining the AI) for seemingly no real reason other than just to capitalize on the AI trend.

If iNat is now contractually obligated to follow through with adding this, I would strongly suggest having it as either an opt-in feature as mentioned above, or incorporating it into a separate app such as Seek. This would at least reduce the amount of incidental damage caused by the tool while still having it as an option for those who would benefit from it. Additionally, not only would I support the creation of a community wiki, but I would also volunteer time to contribute to ensure its information is up-to-date and correct. I hope that the iNat staff has considered this before the announcement, but I do agree that we should still wait for more information about how this system will work before deciding where to go next from here.

This doesn´t seem like a nice path for I Naturalist. I feel that every app is in a rush to somehow get AI involved, but end up forgetting the core of the app itself.

Also, AI is not always exact, it confuses things, so what´s the point in an app like this.

And lastly, I think it takes away a bit of community around here.

There, al least, should be an OPT-OUT option.

Why use a GenAI tool which hallucinates and is mostly inaccurate when you have an ARMY OF PEOPLE on the site who are much more accurate, don't cost a single penny, and do the EXACT SAME THING.

Increased environmental harm, human exploitation, and data "theft" with decreased community interaction, scientific gravitas, and reliable results... no thanks.

Give us a way to OPT OUT

iNaturalist as a platform has meant the world to me. Its been the primary way I can document my connection to nature, and its the number one thing I point people to to learn more about the wonderful flora and fauna around them. But it wasnt the CV model or the massive database that kept me here, its the community. Ive met so many wonderful naturalists on here, and together we've made a difference. Now all of that is getting thrown away. This feature COULD be revolutionary, if its crowd-sourced. Gen AI is not only known for making things up, its destroying the very nature we are trying to document. Worst of all, stealing the work of our community to make slop tags goes against the spirit of citizen science. You are scientists, you KNOW (our should) that generative AI is destroying our world, and suffocating the human spirit. I hope that you change your mind and use that grant to support this wonderful community we've created instead of bowing down to AI

I do not want this feature, and would rather delete my observations from the platform than to let them be used to train gen AI. This is a bad move. Do not implement this.

I don’t want this and I don’t think you’ll find many of us that do, I would rather continue to connect to nature through people, not AI slop which is not accurate enough to rely on for anything, let alone something as important as ecological surveying. This will alienate a lot of your user base and I will be looking for alternatives if this happens.

I have donated money to this platform because I believe it’s important, but this has shaken my trust. I do not consent to this.

We are not on this platform to train your LLM. Do not implement this.

This is a mistake.
Having a human community as the backbone of iNaturalist was a feature, not a bug.
The unintended consequences for all the data collected by individuals and shared as part of a community is problematic.
AI isn't the only approach to this community need, as you have well established up to this point
The question isn't over applying computer power to identification, the question is where do we draw the lines for roles, and where best do humans fit into that "proofing" process.
I wonder if there should be two options:
iNaturalist
&
aiNaturalist
and let folks choose with their feet?

bugguide is rife with lack of data autonomy and throwing away of usable data because a curator thinks the photo is 'frass' (insect poop) and it is just deleted. I recognzie there are ways bugguide does certain things better than iNat, but turning iNat to bugguide would be as big of an issue for me personally as turning iNat into an AI dystopia.

I have concerns about the accuracy and hallucinatory aspects of gen AI.
If it offered an explanation as to how to tell two different species apart I would have to check in other places (guide books etc) to corroborate it and reach an ID, just like I'd do without the AI, unless the difference was very obvious, in which case I would have spotted it without the AI... untrustworthy suggestions are sometimes worse than no suggestions.
I think a wiki of tips about how to ID species made by the experts already on iNat would be far better. Also, I know that involvement in gen AI will make many people quit the platform, and could alienate some of the best IDers who have invested a lot of their time and experience in iNat.

I agree with most of the comments above and this feature worries me. I cannot conceive why iNaturalist would get AI-generated identification tips when it is still impossible to filter for observations with comments in order to find those that include identification tips from experienced identifiers. A user-edited wiki wherein contributions can be verified by curators or the whole community would make much more sense. Considering the small sets of RG observations of most species, AI-generated identification tips would probably be unreliable anyway! I mean, this AI didn't even notice that the frog on the left has a uniformly white underside whereas the one on the right has an additional brown band above its belly!

Please don't fall for the money iNat 🫤

Another vote for wiki, this is one of the last good places on the internet. Let's keep it in the hands of the community.

If you go ahead, please allow users the chance to opt out.

This is incredibly disappointing to see. AI viewing something and suggesting an idea is generally ok. Many times we see this fail, and we understand that it simply does not know, in the same way we don't. Allowing AI to create new information to pose as fact is irresponsible at best. You might as well believe everything anyone says online. This is more likely than not, going to be a detriment to this community.
A user-edited wiki would be great, though. That would bring something useful to the community, and help make real scientific knowledge more accessible to the public.

Don't do this. Google has done nothing to earn the trust of this (or any) community, will erode the community and the integrity of the app. It has no stake in this project and has no interest in seeing it thrive. Google's goal is simply to push its products everywhere and collect data on users. If they suddenly decide to monetize their platform built on inat (as they have done with other apps in the past) then the community loses out on everything we have given them for free.

I believe iNaturalist should not take this direction.

A lot has already been said on the environmental impacts of generative AI, unethical data mining by large corporations, the potential inaccuracies of these kinds of system, and the possible human-focused alternatives. There's no need for me to expound upon this, I just want to add my voice in condemnation of this handover of identification expertise to a generative AI system.

In addition to being an iNaturalist fanatic (and monthly contributor for the past 5 years), I am a biodiversity researcher who uses iNaturalist and other machine-learning tools for work, including custom-built systems for making biodiversity monitoring more scalable and accessible to non-experts. I am not against AI. The iNaturalist computer vision and human expert validation system is one that I constantly use as an example of how AI should be used. However, I believe that we should not aim for AI to replace experts, but to empower them to share their expertise with non-experts. In this context, what I mean by 'experts' is well-informed identifiers, not necessarily just scientists, and acknowledging that not everyone adding identifications really knows what they're doing.

Generative AI is fundamentally different. I see the computer vision model as a tool. The addition of a generative AI will not be adding a new tool, so much as adding a different kind of 'expert'. I predict that non-experts on iNaturalist who are just starting out identifying things will naturally gravitate towards learning from the generative AI model, rather than from real human experts. This is a dangerous feedback loop to create, as AI-trained 'experts' will then feed the model, rather than human experts. Let's keep the iNaturalist community a community of humans.

As a mega user of iNat from the under-resourced part of the world, I fully welcome this initiative, and look forward to seeing how it evolves - from Tony Rebelo, who is in South Africa with me.

For Africa, its observers and identifiers I will wait and see. It could be helpful, and we mostly don't have access to 'taxonomy and original descriptions behind a paywall in academia'.
Deleted accounts piling up a fresh behemoth of Needs ID - that is daunting.

Commenting to echo the concerns of others. Generative AI is unreliable and spreads inaccurate information. A user-contributed wiki seems like a much better investment.

I do not need to reiterate any of the criticisms already stated by others. My personal take on this—the CV is already terrible for a great many taxa—just look at earthworms, where there are tens of thousands of misidentified photos as “Lumbricus terrestris” which makes iNaturalist useless for most oligochaete records. This feature can only make that situation even worse. Add on all the experts who won’t be using inat because of AI, and suddenly we have a website that churns out misinformation by the bucketload.

Why not just have user-written comments that pop up the exact same way as these potentially pointless AI comments? I would welcome that feature, since it would be an extremely useful way to give disclaimers such as “The CV cannot identify species of [Taxon]” or “There are undescribed species in this genus; see these observations” or “Species A and Species B are only identifiable from photos of the underside”. Carefully worded statements written by experts (who can, unlike an AI, then be contacted for specific questions!) could bring much good to iNaturalist.

The post says it, actually—iNaturalist’s great strength definitely is its network of volunteer identifiers and experts, and while there will never be enough of those, they cannot be replaced by generative AI.

I assume this response wasn't anticipated by iNaturalist staff -- why not? In the end, this is a relatively small grant with, in my opinion, the likeliest outcome (if iNat continues with it) being a tool that is tried out but ultimately rolled back because it doesn't work very well. But my biggest concern is that there are steps leading up to this where the community's sentiment on this topic wasn't really considered, or was at least misjudged. As another user said, this is a serious and controversial topic, which should be treated with similar sensitivity as something like introducing ads. I think iNaturalist has recently done a good job surveying its community, just not on this yet, so I'm hopeful about the resolution. When organizations start to consistently disregard community sentiment, it is a death-knell for the community. So I think this is a great opportunity to course-correct. It's not worth burning good will and trust of the community (not to mention donations, observations, and IDs) over a relatively small amount of money that has no guarantee of leading to a useful outcome anyway.

In the spirit of trying to be constructive, here are a few things that might help:

• A user survey on this topic, both so people are able to give feedback, and to help the iNat team make decisions (although there is plenty enough feedback here already)
• Updating the terms of use, as suggested in Michael Pirrello's comment above
• Not moving forward with this grant, and instead creating new functionality like a wiki or upvote system. If there is no funding for that, try setting up a goal-specific funding drive -- I am sure it would reach its goal. The community response has been overwhelming, and this grant is not even that much money compared to operating costs. Some people have already deleted accounts, and many newcomers won't even try something with the perceived contamination of genAI (which is reasonable after all, we've seen it happen to so many sites; we want something we love to be safe from it).
• Maybe there are also ways to make this better (like releasing the agreement with Google to the community in its entirety, the grant proposal in its entirety, making any tools opt-in, etc.), but realistically I worry that there will be a lot of damage if this goes forward in any form.
• Side note on the messaging here: It's not quite right to consider genAI and computer vision the same thing. iNaturalist calling the CV algorithm "AI," has always seemed a bit generous, but especially now that LLMs, a very different beast, are around. Also, this announcement is still unclear about whether identification comments are being included (instead of, maybe, keywords generated from the photos themselves). But it's important to note, the response here is NOT just due to a communication issue. Sure, more details are important, but the community has real concerns beyond just how this was communicated (e.g., in the LinkedIn post, tweet, announcement, etc).

Most of my concerns have been expressed by others, but chiming in to add that I have serious reservations about this. Even suspending the potential ethics/privacy concerns for a moment, this could severely damage the credibility of iNat as a scientific resource. The people are what makes iNat special. I have been a monthly supporter since 2020, but this has me reconsidering that support.

How do we opt out?

I am heartbroken to hear that iNaturalist is partnering with Google and their AI. I agree with all of the ethical and environmental concerns mentioned in all the comments above. Machine learning is great for science - but Generative AI is a whole different ball of garbage.
I work with teachers and promote iNat and Seek to them as a way to engage students in thinking about observing and quantifying the world around them. I likely won't be doing that going forward, because I don't want to reinforce the idea that Generative AI is legit.

Don't mess with perfection - iNat is great as is.

Want to "improve species suggestions"? Keep the status quo. Or better yet, use the grant money to add the user wiki people have been asking for.

Companies have been adding Generative AI to their interfaces prematurely and haphazardly for the past few years and I've yet to see any mainstream platform where this has been well-received. This is anecdotal, but almost any instance I've seen of people actively speaking out in support of Generative AI integration has been at least one of the following:

(1) People with limited grasp of technology, who essentially view GenAI as magic (presumably not the main userbase for an online platform like iNat)
(2) People who are highly business-oriented (isn't iNat a nonprofit?)
(3) People with socially conservative political leanings (iNat is predominantly American, and the political right in the US currently tends towards anti-intellectualism and conspiracy theories, so unlikely to make up a large portion of the userbase on a citizen science website)

So why on earth would this be a good idea? Why would it benefit people on here?

There's so many ways to make iNat more technologically advanced without making rash decisions to piggyback on a halfbaked fad like Generative AI (I echo some other users here in comparing it to NFTs or cryptocurrency, though I would add that at least cryptocurrency often does what it's supposed to even if it's objectionable for other reasons). Maybe you could aggregate comment threads on existing observations of certain species – aggregate only, not synthesize. As in, going to a taxon page would show not just About, Taxonomy, Map, Similar Species, but also a new tab with a paginated list of comment threads (with usernames and timestamps included) discussing identification of that species. Maybe even having this be built into the Similar Species tab. There could probably be an AI use case here still, in the form of filtering out threads where the only comments say things like "Great photo!" with no identification discussion involved.

I have enough interest in tech to not harbour some kind of blanket hatred for all AI. But I'm already somewhat bearish on Computer Vision in its current state (Dargida procinctus. Don't make me tap the sign.), and outright hawkish on Generative AI considering its ineffectiveness at reliably doing anything other than writing code or corporate emails. I do believe that GenAI will have potential someday, but it's really just not there yet, and it's disappointing how much it's being rushed out everywhere without any regard for users. There's no reason to jump headfirst into GenAI right now the way that every big tech corporation has already been doing. I really don't think having a system that provides poor-quality or outright misleading identification information is worth alienating many of your most active users.

In the absence of actually listening to community input, hopefully it won't have to come to this, but with iNat being open-source...I don't have the technical skills myself, but from what I've seen from this community so far, I do have faith that someone will set up a competing website minus the GenAI if need be. (Worst case scenario).

As a user of nearly ten years, a monthly contributor for three, and as an artist whose career has been taken out back and shot by generative AI, the decision to tie iNat in with Google's systems in this way (generating identification cues) has honestly broken my heart. I'm glad that so many are making their opinions known.

I feel that the discussion has become emotionally charged.

Our goal is to build a working demo by the end of the year, and we’ll share more updates as the project evolve

Let the grant be spent on looking at some tools. Should the tools not work, produce unacceptable results, or turn out to be substandard, we can drop it.
But give the technology a chance. 30 years ago we did not have internet, smart camaras and iNaturalist. Who knows: AI ++ may just deliver tools that specialist taxonomists (and I am one) will find indispensable. Imagine if sentiments had killed iNaturalist in 2010: "how on earth can laymen contribute to science and conservation: what an absurd notion! It will be useless: inadequate photos, wrong identifications, a confusing mess interfering with science and taxonomy."

Half the pro-genAI statements here seem to be going "um... did you guys know that iNat already uses an AI for its photo identification feature? 😏" I think most of us very well understand that they're incredibly different types of AI. We do not want this incoming generative AI. I'm adding my voice to the list of scientists and educators here as an incoming entomology PhD student to say I strongly oppose using genAI to summarize user comments to produce identifying characters for taxa.

Sure, this might work for something as distinct and common as an American robin. Many species, especially local favorites, are charismatic enough to have genAI probably write an accurate description for them. But the vast majority of species cannot be identified with the naked eye, let alone with even a few photographs, and such organisms are misidentified often on iNaturalist. A lot of people have been demonstrating that LLMs are reaching near human levels of accuracy, but much of iNaturalist's community is understandably not that accurate or completely lacks the ability to even guess when it comes to very small and/or indistinct species like invertebrates, plants, and fungi. An LLM trained on iNaturalist comments would not fare much better on these species, if there's even a bank of comments for those species to pull from. There doesn't seem to be a point of adding genAI if it's going to keep making the same mistakes or have the same knowledge gaps as the data it's trained on.

Furthermore, professional species descriptions that exist for many indistinct organisms are 1. barely available on the wider internet and 2. written in incredibly arcane language that even some scientists have a hard time parsing. Some terms even mean different things for different organisms. Especially for point 2, I doubt a LLM would be able to regurgitate such information accurately, let alone properly summarize it into common terms. Oftentimes, such species descriptions have no choice but to use extremely specialized jargon. It's the nature of taxonomy. We are dividing species by literal hairs sometimes.

Instead of focusing efforts to futilely ID such organisms to species, we should save writing these explanations for broader, more easily-identifiable taxa. Even obscure species would get some label, and researchers who know those groups may even be able to use those observations from there. I completely understand that writing a concise, accurate, and accessible description for every species is impossible. It would be near impossible even with automation. Both genAI and humans would need fact-checking when writing these. So why not just stick with humans, who can definitely feasibly write descriptions for the much fewer broader taxa out there? The species that genAI could easily write for could also be easily and quickly written by people! There's over 1 million species of insects, but probably only a few thousand or less insect families. I implore iNaturalist to keep encouraging community participation by hiring experts or organizing trusted and knowledgeable volunteers to write the descriptions for these broader taxa.

@peaceblaster
Keep at it, the more that people who do support this decision continue to insult and invalidate the concerns and feelings of those against it, the worse your side looks. You're seriously telling everybody here that we were "never serious about this project, science, or nature"? I'm certain most who have threatened to leave are only doing so because they wholeheartedly believe that this change runs completely counter to the mission of this project. Passionate iNaturalist users from around the world have been trying for years to implement something like this but made through the work of human hands and collaboration, and to see all of that dedication and effort being dismissed as a "knee-jerk reaction" is frankly insulting. The hostility being shown to the people who feel rightfully betrayed and ignored by this announcement isn't going to save iNaturalist, I can tell you that much.

@pythakorean It is only being applied to species that are "in" the CV, and have justifiable use for the added "remarks". They are also manually vetting them for accuracy before they are released. I have been reading paragraphs of people explaining why this will fail on obscure species or those that you can't identify from photos. It clearly will. But as far as I see this isn't going to be a problem because of the prior requirements. Little to no efforts are being used to futilely ID such organisms (aside from those that shouldn't be in CV, but that's a separate issue entirely).

It also isn't going to scan the internet for species descriptions. The project, and when it was proposed a year ago, outlines that it is using existing CV training to produce "text explanations" of how it selects certain options. It is not going to make anyone redundant, because the whole CV as it stands is a continual collaboration among everyone. If anything, this may be an improvement because it will be bringing ID commentary across observations to a more accessible location.

Personally I see many occasions where identifiers din't leave comments or explain IDs, and so I then comb through every observation or spend time online trying to understand. Sometimes I can't figure it out. I think this situation could help solve some of these issues. Knowledge to identify species should be available for all. This will not be a perfect solution, but it will be a start.

The biggest flaws of this project are, I still argue, the same as CV at present, which are people who blindly agree with suggestions given to them. That's going to happen, as it always has, and while it might theoretically increase the instances of this, at the end of the day we should be encouraging users to do their due diligence either way.

PS. and yes, while they are "different" kinds of AI, the overall system is still part of CV and so that's why I group it like that.

Please don't partner with Google AI!!! Aside from how incorrect it can be, the horrendous environmental impact of AI data centers is completely antithetical to iNaturalist's purpose. NO AI!!!

Yeah, no. The failure modes for this are bad (can you imagine people repeating bad mushroom ID rules once they read them?), and I'm not loving either "genAI slop now being part of iNat" or "iNat providing training data for genAI".

I used to trust inaturalist as a project pretty implicitly (knowing the computer vision engine itself isn't perfect--still a great tool). This feels like stepping off a cliff.

Can we opt out? Or are we stuck deciding whether to delete our accounts / thousands of uploads if we don't want your LLM garbage?

This is a terrible decision, your focus is clearly on short-term profits and not conservation and I don't know if I can continue using this app in good conscience if you follow through with this decision. At the very least give us the option to opt-out of gen AI data collection and gen AI-related features. You've alienated your user base with this decision, we're all so tired of AI being forced down our throats like a bunch of ducks being force fed to produce foie gras .

It's been a very long time since I've seen such agreement about a topic from such a wide variety of people. Mistaken or not, for better or worse, I think iNat should side with what appears to be the overwhelming majority of users and rethink this idea. Any benefit to the ID process provided by genAI would be vastly outweighed by the harm that it would cause to the community as mentioned by so many above. The voice of the people says that they would very much rather have ID tips on the species page.

A concern I have is how exactly groups with many similar species (that can pretty accurately by identified by the CV model) are going to be approached. Sometimes there are dozens of possibilities, and without already knowing how to separate them, only a wall of complicated text can allow the average person to accurately diagnose their specimen. Personally, I do not trust in AI to accurately generate information on obscure, but common species, that only a select few humans know how to identify. I have spent a lot of time identifying obscure beetles on here, and would gladly write about their differences on species pages if I could, I believe a user curated wiki is the best move. Adding AI generated ID remarks to CV suggestions (or similar) seems near pointless in my opinion. It is such an easy way to spread false ID tips, and if remarks are going to be reviewed or rated by curators (or normal users) already, why not just have these curators write their own remarks? Also, I like the CV because not only is it convenient, but it provides no explanation, which in my opinion creates a level of distrust in its suggestions, which I believe to be good, as (hopefully) the average user understands that it may be mistaken. If this AI feature will be implemented, I would at least want it to not provide information when it is uncertain, similarly to the CV.

I saw above the concern that people are lumping other iNat use into the 'carbon cost' calculation. I guess i am concerned in the other direction, that focus on the carbon cost of this one use of AI in particular (rather than the obviously harmful overall cost of this form of AI) just leads to people not wanting to drive to do ecological surveys, not want to use their phones and computers to record biodiversity data, and instead just turning off their lights and sitting in their rooms waiting to die. It is important that we don't waste resources but ultimately if we have a situation where the people who care most about nature stop doing anything while those who don't care keep intentionally being wasteful, we end up with more harm than good. Avoiding waste, and efficiency, are important, but we also need to recognize that the answer to climate change is never going to be "some people minimize their impact while others do nothing". It will be societal change to other energy sources and better conservation and restoration of habitats. (and, of course, much of the discussion is a massive distraction and why oil companies invented the "carbon footprint" concept in the first place) Unless one is being excessive, like idling an H2 next to a wetland for an hour while you observe orchids or watever, i don't think angsting over the carbon impact of collecting biodiversity data, which has to account for like 0.00000000001% of carbon emissions and promotes conservation of carbon-sequestering habitats, is actually productive.

That's just to address that one issue. I still think that any further iNat collaboration with "Gulf of Capitulation" "Wait, no, be evil now" google is a bad bad idea, that ignoring the wishes of the user base is a bad idea, and that ignoring so many other user requests while doing this thing that no one wanted is really frustrating.

"They are also manually vetting them for accuracy before they are released"

Who will do the vetting? Obviously experts on the specific taxa, otherwise there's no point. But these experts are eager to write their own identification notes and have been saying it for years. They just need iNat to create infrastructure for that (all the talk about user-edited wiki). Writing is fun and creative work — even more if done communally and collaboratively. It allows you to think through the topic again and sharpen your insight. Vetting somebody else's writing is a chore. Vetting a computer-generated, impersonal output would be soul-crushing and I can't imagine many people being super motivated to do that.

"the overall system is still part of CV"

Why do you think so? They say "For years we have grappled with how to surface the most useful identification tips shared by these members of the community, and we think that generative AI could provide a scalable way to synthesize and share useful information about how to identify the 100,000 different species included in our current modeling process." That sounds much more like a separate system working in parallel with CV. This generative AI would be synthesizing user's comments. It wouldn't interact with CV. In fact, I can't really imagine these two systems working together. CV doesn't even "know" why it suggests an identification, so it can't communicate that information to the LLM for it to create a human language summary... assuming such connection, bridging two very different types of SW, would be even possible.

I feel absolutely sure the iNat staff did this with only the best motivations and expectations. No question. There's no way they expected this kind of response, and I know they're working through this as best they can. I'd be happy to burn Google, Meta, and the rest right to the ground (figuratively), but none of what has happened changes my belief that the iNat staff is acting in good faith.

To the iNat staff - I sure don't envy the position you're in right now, but just know that this isn't a reflection on you. You're collateral damage in a firestorm of exasperation with this technology. Whatever decision you make, I know you'll do only what you believe is in the best interests of iNat.

@coreyhusic I feel like if there were so much more to it, you would have said that in response instead of just saying I don't understand... and yes, I do fully believe these are personal hangups about the words "AI" and "Google". We can call it ideology if we want to dignify it to that degree, but this is looks like a typical Twitter storm to me. Let me go through what I'm seeing in this thread, and feel free to point out what I'm missing:

• Complaints about Republicans, crypto, profits, etc.: fully irrelevant, sounds like they just want someone to talk to

• Complaints about carbon impact: this is a pretty low impact model, certainly lower impact than the one that actually does the IDing. Fortnite on your phone is probably using more power than this model.

• Complaints about Google: Sorry you don't like Google? This is basically a grant- no one's giving them data, no one's giving them money, and at most, this is just Google buying some PR (e.g. "look at us! we support this cool non-profit!").

• Something something scraping data for gen AI: iNaturalist is an open dataset anyone can use for anything, and has probably already been scraped by every major LLM.

• Something something AI IDs suck: yeah, probably, but that's why we have people confirm them. AI is not replacing anyone here. Also, don't pretend everyone here has a PhD- literally anyone can download this app, and the AI we have now is probably better at IDing than most users.

I'm just saying that either not reading, or deliberately misunderstanding the post, and then responding by pulling your contributions is short-sighted at best and outright selfish at worst. Your data has already been scraped for AI- all you're doing is taking it away from science and what you see as legitimate use (I think AI is a legitimate use, but that's another discussion).

Very well said @neontetraploid

This news is disappointing for a number of reasons. AI is becoming more and more difficult to avoid, and as this blog post states, more and more users are joining iNaturalist, meaning that this will increasingly contribute to the harmful impact that AI is having on the environment. No one ever asked for these services and I fail to see how they align with this community's care for nature. Even if these new descriptions appear to be helpful, they eliminate the need for critical thinking, which is an essential part of making observations and being a naturalist.

What I've always appreciated about iNaturalist was the combination of the machine learning algorithm and human expertise. I am concerned that this devalues human expertise. Additionally, generative AI has major problems with hallucinations, as well as extreme environmental impacts.

Just yesterday I was lamenting generative AI's parasitic appropriation of the intellectual labour of human beings. (I trust the iNat community will pardon any slight my metaphor might implicitly express against actual parasitic organisms, which have their legitimate place in the natural order.)

My perspective is informed by the principle of subsidiarity popularized in the late 19th century by the previous Pope Leo, number XIII, and later paraphrased by the Catechism of the Catholic Church: "neither the state nor any larger society should substitute itself for the initiative and responsibility of individuals and intermediary bodies."

If appropriate attitudes are held, the Compendium of of the Social Doctrine of the Church explains, "intermediate social entities [here read: identifiers] can properly perform the functions that fall to them without being required to hand them over unjustly to other social entities of a higher level, by which they would end up being absorbed and substituted, in the end seeing themselves denied their dignity and essential place" (186, emphases added). The torrent of disappointed comments here is a predictable consequence of "an absent or insufficient recognition of private initiative … and the failure to recognize its public function" (Compendium, 187).

I make heavy use of pattern recognition by iNat's computer vision feature, always understood to be subject to human verification. Generative AI is a different matter altogether. Setting aside the problems of hallucinations and environmental externalities, I see gen AI as ultimately corrosive of the community that has made iNat so appealing to me.

I see the grant as a trojan horse, and I struggle to imagine the conditions under which I would be prepared to continue participating if the funded project proceeds.

Dang. . . this is going to confuse beginners even more when using CV for identification as misinformation is inevitable with a LLM. If you're going to do something AI related on INaturalist, maybe improve the scope and accuracy of the CV without changing it too much? It works fine.

If nothing else, let this be the ultimate question. What does the site have to gain from following this course, and what would it lose? Is it worth it?

Disregarding environmental and ethical concerns that are being voiced very well by other people here: The ability for community curation of species differentiation notes is crucial. I've seen so many observations with perpetuated reasons for identification that are incorrect or within the morphological range of multiple species - just because sources are few and some are incorrect.
I'd be very worried about the perpetuation of incorrect or unhelpful identification notes through this and the possibility for it to fundamentally undermine the quality of 'research grade'

I'm with @charlie on this one. Lots of good uses of AI with iNat that could have been explored (really hoping for the sound CV)--but this isn't one.

I'm seeing lots of ill-informed reasons why people don't like the proposed use of genAI (well summarized by @peaceblaster). For example, one can run genAI on a laptop--not all genAI needs a coal-powered plant to run it. But there are still reasons why it's not the best course of action (if only because of the distrust). I think it's pretty obvious that there's nothing that the iNat staff can say that will give users faith in this approach.

This undermines the scientific, educational and environmental goals of iNaturalist, as LLMs are known for creating false informations and being very energy-consuming.

If this is implemented I'll delete my observations and account.

@pfau_tarleton agree that this isn't a necessary feature, especially given the pushback it seems to be getting. I still think that pushback is teeming with misinformation, blind rage, and an appalling lack of nuance, but I'm still worried how many people are willing to delete their contributions regardless.

I've been a major supporter of iNaturalist, but this gives me pause. Going to reflect, but I don't think this is something I want to be a part of and am strongly considering deleting all of my contributions as well.

This is very, very weird and in my honest opinion, just gross. Many of the users here have already expressed opinions that are very much in line with how I myself am feeling but if I’m to put in my two cents I just have to ask. Why? What is the point of an LLM that scrapes from user discussions to provide ID tips when you could create a basically identical system that simply relies on those actual human identifiers to write out the knowledge? I myself have seen the discussions on this site and know for a fact that absolutely not every single one would be trustworthy to source these “tips” from, which I get is why the feedback system is mentioned, but that just reinforces the “why”. You would not NEED a feedback system if you already had a group of trustworthy figures curating these ID tips together. I fail to see how a human curated system of the same nature would be less efficient than a Google sponsored AI. Which I guess leads me to add another concern. There is absolutely no way that iNaturalist staff have not seen the poor performance of Google’s AI systems or LLM AI systems in general. In just the last several months Google’s search engine AI has lied to me about there being a breeding population of white sturgeon in Mexico, as well as lied to me about the names used to refer to certain sturgeon hybrids, all while confidently linking websites that completely DISAGREED with everything the AI was sourcing them for. Once again, with how we ALL know these LLM programs falter, it is just impossible to properly fathom any non profit-based incentive at this stage that would genuinely lead to iNaturalist seeing an AI-curated system like this being beneficial to the platform. Thousands of users, myself included, have left countless comments on forum posts and the like discussing how willing we would be to contribute to a human curated system just like this, not to mention to guides and potentially a database far more in depth. It feels, to me, very disgusting, knowing that iNaturalist staff have most certainly seen these prevalent discussions and have now decided that instead of inviting the community into more collaborative work, the platform will now have an AI to scrape their contributions and regurgitate them into something most of them don’t even want to be a part of. In my eyes, that is iNaturalist telling it’s users they are resources to be sourced from more than they are the contributors that built this amazing platform to what it is now over the last 2 decades.

GenAI would be filled with errors and it would only worsen wrong confident identifications. A user created wiki would make problems less rarely and could fix them easily. Truly this is a spit to the face of every identifier on this site.

This is an exciting opportunity. iNaturalist's success has always come from combining its incredible community with a willingness to embrace new technologies while maintaining high standards. To stay at the forefront of biological recording, it's essential to explore the cutting edge.

The pace of progress in this field is staggering, and I don't see this as an "either/or" choice between AI and a user-driven wiki. The two approaches could be complementary, even symbiotic. I'm optimistic about this experiment and trust the iNat team to navigate it thoughtfully.

I have over 800 videos and 2000 photos, which will equal somewhere in the ballpark of 500+ observations, sitting on my computer waiting to be processed and uploaded to iNat. I'm not going to bother going through them until you change your stance and clarify what is going on. It's your move, iNat. You're making me withhold valuable data because I no longer feel comfortable sharing my observations with y'all.

@tonyrebelo
"What is this really about? What is proposed is that the current CV (which we already have) will now have an explanation as to how it made its ID. The CV is OK, but it giving a reason is not."
The CV doesn't make its IDs with logic like the kind you and I use. it makes its IDs by analysing many images and going "the ones with these coloured pixels in this arrangement are often identified as these species". it doesn't know what a wing or a nose or whatever are, so whatever text you get isn't actually a reflection of how it identified the species, it's an amalgamation of other people's comments pasted onto the CV's guess, most likely with very little acknowledgement of what features are visible in the actual observation.

The kind of AI that generates text is, for all intents and purposes, basically overly fancy predictive text. It learns what words tend to come after each other when different words are present and then puts words together in a convincing sounding order. It can't think or know things the way humans can. It can just make sentences that sound like a sentence a person would put together

So I'll add a few thoughts. I think the intent here is positive. Many people who are novices have said something along the lines of, "I wish iNat would tell me why this is a Plantago wrightiana rather than a Plantago helleri. This is meant to be able to explain that, probably in more of a layperson's terms, which makes nature more accessible than plugging away for hours with a dichotomous key.

This said, I do understand concerns people have, particularly if they haven't explicitly agreed to have their notes and comments used this way. I wonder how this gets into details with Terms of Service. If this data gets scraped, and somebody deletes their account, is the data still available in perpetuity or does it get removed from CV model / LLM suggestions?

Why not leave things like this to the huge and knowledgable userbase we already have on this site, instead of outsourcing human teaching interactions to an AI that uses unconscionable amounts of energy to achieve usually less accurate results? I would be really disappointed to see the site management follow through with this. For me, iNaturalist is a place to learn from other people and share my own knowledge. I'm sure I'm not alone in enjoying the genuine human interactions I experience here, especially in a world that seems more and more focused on replacing those with computer models.

This isn't even going into the possible safety issues that would arise should people trust an AI 100% to tell them what mushrooms are safe to eat, etc.

Please take some time to consider whether incorporating the technology-of-the-moment is worth robbing human users of the joy of sharing our knowledge with others. As someone who was hoping to monetarily support the site in future, I can safely say I won't be doing that if this comes to fruition.

Yeah… no. Absolutely not.

Going to Google and asking for generative AI, when your userbase is dedicated to conservation, education, and appreciation of our world's environment and biodiversity? You will lose donors, you will lose users, you will lose observations, you will lose scientists, you will lose public opinion, and you will gain a lot of bad blood if you incorporate Google's genAI into this website.

Frankly, no thank you.
Generative AI and the whole host of issues that come with it are antithetical to the community here and their mission. As a relatively new user, I’m super disappointed to see yet another cool company succumb to Gen AI. It’s everywhere and we don’t want it. Please get rid of it, your community here doesn’t want it.

Really disappointed about this to be honest. I really do not not want gen ai in my inaturalist. This is no Reese’s peanut butter cups.

Please don’t do this. I’m a huge fan of iNaturalist and recommend it to people often, but I wouldn’t feel comfortable doing that or even using the site if it suggested information on identification made by AI. This will likely turn iNaturalist into a platform that spits out misinformation, and even at its best it will discourage people from doing their own research. Experts understand nuance, Generative AI is a just space heater that lies to you.

Oh absolutely not. There are a whole host of issues raised by the use of generative AI that are annoying at BEST, and downright lethal at worst. There's plenty of examples of people using identification books written with the use of genAI being told that something is entirely safe, when in reality it ends up very deadly (button mushroom vs destroying angel is the first that comes to mind).

And that's glossing over the fact that to even use gen AI, there's a ton of resources that are being used up. Clean water that would be better used for drinking or watering plants is getting sucked up by greedy companies to keep the machines cool, there's a ton of unclean energy being used to even power these things, etc etc. Personally I'd rather NOT have my data get used for something as soul-suckingly earth-destroying as this. ESPECIALLY by a group that seems to care about the planet a lot.

None of us want this genAI slop in our identification app. Especially not when it's hand-in-hand with the corporate greed that google has been shown to have. I am incredibly disappointed and I hope that this gets removed promptly.

iNaturalist's strength is its community. There's a risk that emphasizing generative AI could inadvertently reduce human interaction on this platform, which is where a lot of learning and mentorship happens. How will you maintain that balance and ensure identifiers for certain taxa don't leave the website, effectively leaving iNaturalist with just the generative AI (which can then go completely unmoderated.) I'm assuming only experts will have the knowledge to actually be able to validate the AI's claims on many species, so will you be hiring taxon experts to do this? It's one thing to have experts such as botanists, entomologists, mycologists, etc volunteering to help identify organisms on this platform, but expecting them to moderate all of the AI's comments as well and ensure it doesn't spread misinformation is going to be considerably more difficult. I wonder if a financial benefit will be required as I worry this is looking very labour intensive and in many regards requiring highly qualified expertise. Without constant moderation this AI could spread a lot of misinformation and all it takes is a difficult taxa to only have one or two experts who either leave the platform or stop watching the AI and the AI could effectively spout as much nonsense comments as it likes for as long as a taxon expert is unavailable on the site to correct its behaviour. Normal everyday people are not going to know when the generative AI is lying or making things up. The only people who can correct this false information generated by the generative AI would be the taxon experts who specialize in said taxon. This is putting a much bigger strain on them... Many difficult taxa only have one or two identifiers... I'm assuming you'll be paying these taxon experts if their work will be absolutely necessary for the accuracy and success of this generative AI's implementation? Surely a failure to get these highly educated identifiers onboard will mean the AI is spouting information and you'd have no real idea on it's validity? Generative AI is known to often produce factually incorrect yet seemingly confident responses, I imagine it'll especially try to do this with difficult to ID taxa as well. Who is actually going to moderate this when only a very select few can actually confirm or contest what the AI is saying anyway?

An example I'd like to give: An expert identifier could state in previous comments that the incredibly rare Greater Spotted Haggis Fly has a furrier abdomen compared to the Lesser Spotted Haggis Fly. If there were only one taxon expert identifying these Haggis Flies on iNaturalist and then the expert was to disappear... Who will correct the AI if it started saying that the Lesser Spotted Haggis Fly is completely hairless? Who would even know the AI is lying if it's making things up or warping the facts about an obscure species that nobody has the expertise to correct? In this kind of instance the only way to ensure the AI doesn't do this is to pay people to moderate it, yet in regards to the Haggis Flies there is only one person on the website who is knowledgeable to identify them in the first place. Would iNaturalist pay this identifier to stick around and ensure the AI doesn't lie? Many normal everyday people will not be able to differentiate different genera of flies, never mind species. It seems likely the only people capable of validating or refuting the AI's claims are those who are experts in the taxa already.

Terrible decision. What happened here? Whose families are being held hostage for a nonprofit environmentally-focused organization to decide on the random implementation of GenAI?

I mean seriously, if we think about how iNat's image recognition AI (which is COMPLETELY different from GenAI) can change its mind on a post, and how its opinions can differ even between platforms, then that means that we will 100% be facing an amount of generations OVER one per post, which is just an absurd amount of energy and water at the rate that iNat's database is growing.

I am very disappointed by this announcement. If this goes forward I will withdraw my support and my contributions to iNaturalist

Hell no. I’ll delete alll my observations and withdraw my donations if you do this. INaturalist is about people and planet and the beings we love and want to know more of. It brings us closer to nature. Google and LLM use enormous amounts of fossil fuels to run their servers that if we apply this to everything our observations will be nothing more than a museum of dead creatures.

I posted something similar to this on the forums, but I think it's important to get it both places for posterity and to make sure it's seen. iNat is extremely important to me as an epicenter of citizen science and a source of truth for identifying and documenting the wildlife around me.

It’s extremely worrying for me to see that iNat is accepting this grant, because there are a ton of problems I can see from reading both the public grant terms, and from my personal experience with genAI in the tech industry.

GenAI algorithms have a persistent and incurable problem with hallucinating statistically likely but baseless information when they’re invoked, and it’s an artifact of the way they’re made. This is not a correctable problem. This is not fixable by some kind of multi-model oversight, by some kind of “reasoning” framework, or by anything else. It’s embedded into how genAI frameworks are created and trained.

GenAI algorithms are also, at this point in their development, experiencing significant issues with diminishing returns on capability based on the saturation of their training data with genAI-created content. This will be a problem for any genAI system that is intended to produce output in natural language, or that is descended from any major model in use today. Like the hallucination issue, this is not a correctable problem.

Additionally, after reviewing the public grant terms, I can immediately see a poison pill embedded into them:

[each grant recipient] will receive a share of $30 million along with six months of technical training, pro bono support from Google’s AI experts and access to Google Cloud credits.

This is a trap, pure and simple. You can’t run a genAI model without access to compute power, and Google will definitely be ensuring that you’re running on Google Cloud. Cloud credits are ruinously expensive - the only way that Anthropic, OpenAI, and all the other getAI providers can exist is because they’re getting free cloud credits from Google and Microsoft.

I know iNat is always concerned with server costs and infrastructure costs – a nonprofit science website doesn’t exactly rake in the cash. Think about what happens if you integrate genAI, and then Google pulls your Cloud Credits. Suddenly, you’re not just paying for your own servers and infrastructure, you’re paying whatever Google wants to charge you for a critical portion of your infrastructure. Your site has become a hostage to Google – and you can’t remove the genAI because Google's experts helped you integrate it in a lynchpin position. This outcome would OBLITERATE iNat’s funding, and the only assurance you’d have of it not happening would be Google saying “we would never!”

This is not an exhaustive list of my concerns, but it’s the top few. We cannot allow the seductive lure of money from an untrustworthy benefactor to compromise the functionality, integrity, and usability of iNaturalist, under any circumstances.

After carefully reading your description, I remain unconvinced and urge iNaturalist to drop this plan to add AI functionality in partnership with Google. The environmental footprint of generative AI is unacceptable. I suspect that Google stands to gain far more value in access to training data than iNaturalist will receive. The output of genAI models is fundamentally flawed and unreliable. I don't trust Alphabet/Google and don't want my observation data used by that company. If you proceed, it will undermine one of the last great places on the internet, and I am likely to cease contributing.

Generative AI is seductive, very seductive. Lots of smart people know that. The sentiments being expressed here are overwhelmingly of caution and concern. In my own experience of doing > 97k identifications of plants, many observations will never be identifiable to species, no matter how well AI is trained, because the observers don't show the features (most often the leaves) that are essential for correct identification. AI won't be able to fix that. Educational tools for serious observers could improve the quality of some observations. Taxon-specific Wiki pages with expert-contributed tips for identification AND observation sounds like a great idea.

This is a terrible idea. Google is not a trustworthy company. Generative AI/LLMs makes things worse, not better. The data centers that are built for AI are a negative impact on the environment. Others are more well versed with their dissent against AI than I am, but I agree with them. Google just wants more of our data to feed its machine that can't do anything useful even with all of the data of the internet.

If iNaturalist goes through with this deal, then I will remove my data and delete my account

I say this as an enthusiastic user of iNaturalist with over 11,000 observations to date who deeply values this platform and the community of naturalists it has fostered. This idea would be a mistake to implement.

As many have said, this is a risky idea to begin with and there are better solutions that preserve the integrity of iNaturalist while also allowing continued improvements to the platform. Implementing GenAI (particularly in association with a company like Google) is a huge mistake. The environmental impact of GenAI combined with the major privacy concerns are enough to discourage any use of GenAI in association with something like iNaturalist. Considering the risk of jeopardizing the relationship iNaturalist has with its users and the wider scientific community, the picture just gets worse. iNaturalist already has a community of experts ready to contribute this kind of information, we don't need Gen AI lying to us while Google steals our data and destroys the environment.

@naelin i think the only possible pro is convenience. but there are many more possible cons imo: misinformation will likely be spread, our observation data (or as you suggest, more of it) could be used as training data without our consent (this could become illegal in a number of years, rendering it pointless for poor Daddy Google too) , the community would generally be less tight with less people explaining their IDs, server space and energy would be wasted on it, and of course, it would further exploit people and the earth in a world of pre-ai laws. but technology is always abused until the law catches up i guess so its not like im surprised iNat's jumping on the bandwagon. anyway, these are all pretty minor effects but considering the limited benefits it just seems like a bad decision. at best, pointless. at worst, harmful.

In principle, I think providing explanations of how an ID suggestion is made is a good idea, but is that actually what is happening if you use generative AI to explain how the CV (non-generative) AI made an ID? Isn't that just two AI tools running in parallel? I can imagine cases where the standard CV ID is sound, but the image uploaded don't actually show some feature mentioned in the explanation the (generative) AI provides. Then you may just confuse users. "Where is the white stripe? I don't see any stripe in my picture. I must be misunderstanding...".

This makes me feel completely unsafe and uncomfortable adding any more observations that I make with the Cornell Cooperative Extension, from a place of academic integrity and respect for the First Nations which Google is participating in the oppression of.

Not reading the entire thread.

But please, please, don't. This is one of the only social media platforms left on the internet that hasn't been ruined by AI. Please don't take this step. Please don't ruin the amazing thing that you have created here.

@lothlin sooooo true

Please let the amazing naturalists on this site write descriptions for species and help Inaturalist stay unique. Too many websites are using AI as their main source of information. Code should never be the reason humans lose their expertise. I don't want false information being advertised because an AI thinks so. I made https://www.inaturalist.org/journal/yayemaster/111412-a-guide-to-identifying-difficult-plant-genera-on-long-island without any AI help and has helped many Inatters get better photos and IDs.
AI will only bring Inaturalist further away from its goals of connecting users to nature and exploring the biodiversity of this amazing planet. This will inevitably lead to many users(including me) to leave Inaturalist.
In my opinion, the world was better off without AI, and so Inat should refrain from this grant.

iNat has functioned without this for nearly 2 decades. Pretty please, no.

GOOGLE?!?!? 😨