I think it may be hard to even do a demo with the backlash.
Honestly I wouldn’t pin this on scientists necessarily, more so tech industry executives. I doubt the backlash against GenAI (across the internet, not just here) would be so massive if big tech companies weren’t so hellbent on adding GenAI to every app and website without a second thought. For example, I have (had) the free version of Adobe Acrobat. I also have a decade-old computer. The addition of GenAI features to that app (features that I couldn’t use even if I wanted to, since it’s the free version) made the whole thing too slow to be usable on my device anymore – but the only thing I was ever using it for in the first place was to read preexisting PDFs.
And this ties back to why I don’t think accurate reasoning is necessarily important here, in terms of opposition to iNat adding GenAI features. Comments like those by @natev are refreshing to me, as far as making sure my own perception of GenAI is accurate, so I will say that I do appreciate those (though personally I do continue to oppose the feature announced here). But at this point, for a lot of people, GenAI is just synonymous with enshittification, for lack of a more eloquent term. Big tech companies have dug themselves into a PR hole by deploying GenAI irresponsibly, so to quote @spidercat , I think it’s important for smaller organizations like iNat to “read the room” when it comes to things like this.
Wowsers this has certainly stirred up quite the hornet’s nest. There are a number of points that need making.
- Those who have already deleted their accounts have been ridiculously foolish. They have jumped on a bandwagon without giving any time to consider things properly and to take a calm decision. In doing so they have caused permanent damage to iNaturalist.
- Those who are fulminating about not wanting Google to “benefit from their work” in any way at all are profoundly ignorant of reality. Images on iNaturalist are indexed by Google. That means images on iNaturalist already benefit Google to some degree. So all the getting hot under the collar about that particular aspect of things is pointless heat instead of useful light.
- The above said proceeding with this is a VERY bad idea for many reasons.
- Google enjoy a bad reputation which they have thoroughly earned. This combined with the ideological bias which seems prevalent on iNaturalist has produced a great deal of the backlash already seen.
- The promotion of this has been hamfisted to say the least. Vague genralisations combined with opaque decision making processes create a vacuum and that vacuum has been filled with a huge amount of negative material.
- It’s distinctly possible that the reputational damage this has caused iNaturalist already de facto outweighs the monetary value of the grant.
- Those who raise the issue of AI hallucinations are correct to do so. AI is confidently wrong about a great many things.
I would strongly advise iNaturalist to back off with this. Postpone any implementation until clarifications have been issued over legitimate concerns. Also consider pulling the plug on the whole thing. The damage done could already easily exceed any “benefit” of the project.
This might be one of the things that most riled me emotionally upon seeing the news. There’s so much that could be improved about the platform, there are so many feature requests in limbo and so many fundamental design challenges that need to be sorted out, and all of that remains on the backburner while this is being worked on?
I understand the huge financial appeal of the Google grant, but it still makes me worry that iNaturalist’s leadership has lost touch with the community, the experience of using iNaturalist for its own sake, or any kind of community- or science-oriented vision for the platform.
Having read this entire thread, I figured I would chime in with some thoughts that folks have echoed but have gone overlooked for the most part in this discussion.
I downloaded iNaturalist a few weeks ago because I needed something to keep me grounded in light of the chaos that is the current state of the world, and my usual activities on my phone have become so bogged down with generative AI. Being a visual artist whose career and livelihood have been directly harmed by use of generative AI, I do my best to avoid it at all costs. (Please note, I’m not fundamentally opposed to all AI, just generative.)
The thing I have come to love about iNaturalist since downloading it is the community. It is very difficult to find meaningful, positive interactions with real people on the internet nowadays, and for me iNaturalist has been an oasis in a desert of AI slop. These past few weeks I’ve been going outside to look at bugs and birds regularly for the first time since I was very young, and I have been learning from other human beings who are as passionate as I am. To incorporate generative AI would hinder the aspect of human connection that makes this platform so charming.
I would hate to see that happen. I would always, always, always rather learn from another person with a beating heart who genuinely cares about what they are doing, than a machine that is repeating information it has gathered. Even if the machine could teach me quicker than a human could. A machine is optional, a person is going to keep doing what they love no matter what.
Inat is better off staying away from megacorporations like Google. This $1.5M grant towards genAI is only going to do harm. It’s basically “Google trusts AI more than the identifiers who work their butts off to help everyone.” I’m sorry, but I don’t want my power as a prolific identifier taken away by a trap. Palantir may take over this genAI down the line and continue to ruin iNaturalist’s uniqueness.
I personally think genAI is one of the most dangerous inventions created, if not the most dangerous. Let’s keep Inat free from this and make curators, identifiers, and observers stay on Inaturalist.
Also, double standards are wrong.
Sorry for coming across really harsh, I just am anti-genAI and love Inat the way it is now.
-Robby
I think you misunderstood me: I am not pinning anything on scientists; I am explaining that scientists who utilize AI algorithms/predictive models in their methods are at risk of having these negative connotations erroneously pinned on them
Hey folks, let’s please keep the gender and AI discussion out of this topic and focus on iNat and AI. You’re welcome to continue it in personal messages if you want, I’ve hidden the recent posts about it.
Just a heads-up that we added some more clarifications and explanations to the original blog post. They don’t address every concern people have, but I hope they provide some clarity as to what the grant does and does not entail and the general outlines of what the approach will be.
LOVE to hear that they’re planning a live virtual Q and A. I feel like they might actually be listening! Thank you for letting us know about the update!
The edit is reassuring in some aspects. But i still hold concerns. Especially over the community itself. Many people were left in fear, anger, disbelief, and so on.
I dont know if this single edit is enough to quell emotions. Its a challange to regain trust and reassure many who still have valid concerns. Im also unsure how best to handle this tough situation.
The Q&A though is excellent and should be advertised in a blog post to reach as many as possible.
Communication with the community is key. Instead of linking a form to fill out after the blog was closed, a link to the forum post here should have been left to encourage public discussion. Filling out a form is not public discussion.
The blog post update makes pretty clear that staff intend to go forward with at least an early stage of this despite the fact that this has generated far more opposition than any other step iNaturalist has ever taken. The update leaves no room for, “maybe we will listen to the community and not do this at all.” The Q&A is a good idea, but announcing beforehand that you plan to shove forward despite the substantial risks is announcing that it is image management rather than consultation.
a lot of folks have yelled about the environmental costs and the data ownership issues so i don’t want to add to that yelling in particular, but i have to say i don’t see having hallucinated descriptions as particularly helpful with the kinds of bad faith identifying folks have been commenting on in the forums this year. or why i would want to spend any amount of time with responses arguing in favor of the hallucinations? i’m not even a taxonomic specialist (which a lot of people want here) and it just seems very much like it will add to the load in an antagonistic way. particularly when we start thinking about the photos that aren’t so clean as those frogs used in the blog example. because that is so unrealistic in practice.
it seems bad as an observer and an identifier. what a bummer.
For a good example of AI being confidently wrong with regard to nature specifically, here’s an exercise for anyone who wants to see for themselves: go to a LLM and ask it for identifying characteristics of a relatively obscure, poorly known species with which you are familiar. It will likely give you a confident, detailed, and completely incorrect description. I tried with two separate AI models (GPT 4o mini and Mistral), asked them to describe the crawfish species Procambarus elegans (which is relatively rare and poorly known, with a very restricted range). GPT gave me a description of the red swamp crawfish (P. clarkii), even calling it by that common name, but claiming the description was for P. elegans (which looks very different). Mistral just made up random descriptions for various crawfish features (such as coloration, rostrum shape, claw shape, etc), many of which were completely wrong. Both gave zero uncertainty in their answers, even though it was obvious they did not have any data for the actual descriptions of this species. And their answers were very detailed. What would prevent something like this from happening if inat implements AI species descriptions?
Yeah, i think this is a major mistake. The community has spoken, and on a scale they never have before.
On a community driven site, that should mean something.
Closing the comments to the main blog post after posting the update is certainly an interesting decision; I don’t think silencing the most visible avenue of communication is the right step to deescalate tensions right now. But I digress, perhaps the most telling part of the update are the concerns which are not addressed: most glaringly, no mention at all regarding the many (MANY) requests for being able to opt in or out of such a wildly unwanted feature. All said, it reads very much as “we see you, we hear you, you are valid (here’s why we are right and you are wrong), we are doing this whether you like it or not”. As a member of the site for nine years who has recommended iNaturalist to hundreds of people, I feel like iNaturalist really does not have a pulse on its community anymore.
Granted this thread discussion is about how people feel about their contributions and platform including “AI” components (companies),
and under what terms and conditions it is ethical to repackage and re-market (as “AI”) the contributions from users, professionals and copyrighted field guides, which is what the ‘training data’ is [this is what many copybot “AI” algos actually do]
But serious side-topic:
The iNaturalist web platform is impressively and generously implemented, complete with a fully featured and very useful API (that wisely throttles responses)
But the overall web services charges seem high?
Would be interesting to see cost breakdown in terms of storage, CDN, compute, etc.
Is there not some philanthropic, tax-advantaged mechanism by which Large Web Hosting Organization(s) reduce the pricing of web-related services for nonprofit/community organizations and partners?
Made an forum account just to comment on this.
- iNat has had AI for years, since 2017 they’ve had a computer vision, CV, model, trained on user data to provide ID suggestions. Its AI. Its trained on user data. Its not always right, but its pretty good for common species. Of course for more rare organisms, its less accurate, but then again so are people. I don’t understand how people see GenAI as some huge unexpected jump.
- As the iNat team has been saying for OVER A YEAR, they wanted to have good explanations for the CV ID’s rather than just spitting out the top species. I can’t even say how many times I’ve seen the top 2 or 3 suggestions for a species and had no idea how to distinguish them. If I’m lucky I can google a diagram, but that works less for less common species. Adding an explanation, or at least verbalizing the CV’s internal processing would be helpful.
- Some people suggest a user-contributed wiki: I think this would actually be a great idea, and the current CV model could just link to that. Just as a note though, how many people will contribute to this? How many less-common species will get covered? Tons of my observations get no additional ID for years, and the ones that do get ID’d by people rarely have explanations.
- As for taking money from Google- come on guys. iNat is a non-profit, its not like they’re pocketing this money just for the fun of it. If Google made a private donation of the same amount to iNat, should they reject it? How many non-profits reject donations or grants because the money comes from the elite who made their money exploiting people? That is literally how many non-profits stay afloat, so unless everyone complaining has personally donated to iNat, I recommend getting off the high horse.
- They didn’t say they are going to implement this technology. They got a grant to EXPLORE the possibility. They are going to see what the environmental impact and user response is before rolling anything out. Honestly a 1-day turn-around with a response to user criticism is about the best possible scenario.
Overall I’m shocked at how negative the response is. iNat has been using AI for years, but now that people know the buzzword “Gen AI” its as if they are committing a crime. I mean, deleting accounts because they are trying a new technology out? Are you kidding me? Every previous blog post on iNat trying out new AI models I’ve seen hasn’t been unanimous agreement, but the general sentiment was “Oh cool idea, it doesn’t seem like its 100% there yet, but I hope it gets better!”. Why is this different? I really hope people actually read the update instead of blindly deleting their account. I’m trusting iNat on this- and they’ve rarely let me down before.
To be fair to the folks moderating the comments on the blog post (are they even paid to do so?), it’s probably a nightmare to deal with right now. Maybe I’m being too nice, but I am trying to be as charitable as I can while this pans out, since there’s still a lot of unknowns. The end result could be very different from what it seems now, for better or for worse.
On that note, I hope to see all of y’all at the Q&A, whenever that gets a date and time. You BET I will be there. It’s an opportunity for them to undeniably hear what we all have to say - the more of us who show up, the better! If you haven’t filled out their feedback form for the Q&A, you owe it to yourselves to do exactly that! https://inaturalist.typeform.com/to/W44CV9GC
I think a larger issue is the comments section on journal posts isnt designed to have that much traffic in such short time and to host such a large discussion. Discussion of this size is best on the forums where many quality of life features exist both for easier responding and actually reading/navigating.
Although if one does not give them benifit of the doubt, its easy to view this as silencing. They absolutely should have had a link to the journal post here.
Also curators are not paid, we are volunteers.