Well i don’t see this happening. Much of the work that feeds AI on iNat is provided by unpaid volunteers.
8 Likes
Preface: I don’t believe taxonomy is antiquated or that AI will replace taxonomists. Certainly AI has and will continue to change the field of taxonomy.
Main point: It’s hard for me to imagine a folly as great as intentionally stifling scientific progress in the name of propping up an antiquated way of doing something. Ask the coal industry how that’s working out for them.
6 Likes
Taxonomy is what you do after you’ve done the systematics of a group. Are there full time systematists/taxonomists? Maybe if they’re retired. Most are academics with teaching and/or administrative responsibilities or grad students who move on to other things and might not publish their findings. Which is why it might take forever to get a new species described or a group revised.
5 Likes
I’m not a professional taxonomist, but based on what I know from other local scientists, there’s always 10x more work than anyone will ever get funding to complete. As long as it’s just used to get more done, and isn’t used to fully replace humans, I don’t think AI getting better on iNat is going to put anyone out of work any time soon. Should just go towards getting backlog resolved quicker if anyone sane is in charge.
5 Likes
Only 2 years
https://www.inaturalist.org/observations/113720433
Will it be described in my lifetime ?
We have 9 obs
https://www.inaturalist.org/observations?place_id=123155&preferred_place_id=113055&subview=table&taxon_id=419559
5 Likes
computer vision taxon classification is one application for such a dataset, but there are potentially so many other applications for this data beyond that. i think it’s fair to debate whether iNat should (dis)continue its computer vision efforts (with the understandng that others would still continue down that path), but i wouldn’t want to conflate that with a debate over whether folks should observe and / or identify observations in general, since that would seem to be sort of a tail-wagging-the-dog situation.
personally, i think this particular case of a technology (CV) taking over a task that was previously done by humans is just another variation of the same thing that has been happening for as long as humans have been around. it’s true that a particular technology can take jobs away from some people, but usually a lot more people benefit from the mass availability of the product of the tech. and then some people whose jobs were replaced by the tech – instead of moving on to something totally different or being forced to work for much less pay – embrace the tech and find ways to expand or build upon what they had been able to do (often for much more pay / free time and impact).
8 Likes
One opinion I have seen expressed: “AI is not going to take your job, but people who use AI might.” That makes a lot of sense to me.
What if AI could be trained to write species descriptions? Maybe something would finally be done about the “waiting to be described” queue that several replies in this thread have complained about.
4 Likes
If the herbarium had a project that allowed the most prolific collectors to receive a preliminary ID from specific taxonomists before specimens were ever collected and sent in, would that result in better specimens in the herbarium? Just a thought I had. I’d probably lurk on a project like that if there was one. Maybe the plant ID that you’re doing is of old specimens? If it’s recent stuff, it seems like a collaborative community ID in addition to the ID given by the taxonomist that’s part of the project before the specimen was sent in would be interesting. Then if it’s confirmed by physical examination in the Herbarium, the obs could go to quintuple bonus RG.
Sorry if I’ve made all sorts of assumptions about how your job (or Herbariums in general) works.
1 Like
What is your usual clientele at the herbarium that needs plants ID’d?
It varies but mainly:
-
Consultant botanists working on environmental impact assessments
-
Government departments
-
Staff and students from educational institutions
-
Members of the public
Thanks to everyone that’s commented so far, it’s great to see that opinions are many and varied.
6 Likes
I did similar work in addition to curatorial tasks when I worked in a vertebrate museum collection. But I wouldn’t call myself a taxonomist because I was not naming or revising taxa. I suppose some of those IDs could be done in the future by putting a physical specimen in a scanner of some sort and having it spit out an ID for you. Same as CV but using the actual specimen.
2 Likes
For those interested there is a book Managing the Modern Herbarium https://www.universityproducts.com/managing-the-modern-herbarium.html
Perhaps @mftasp could comment.
1 Like
The latest Computer Vision model can identify almost 90,000 taxa, which is impressive, but bearing in mind that:
- it’s already learned the easiest and most accessible IDs and the next 90,000 will be harder
- it still generates a non-neglible number of errors and is incapable of understanding when it is wrong
- more than 2 million living species have been described of which the Computer Vision can recognise fewer than 5%
- that there are many more, perhaps millions more, species remaining to be described
we are a very long way from not needing taxonomists any more. Whether the powers-that-be recognise this and allocate adequate funding for taxonomists is another question, but AI will definitely be nowhere near replacing human expertise for a long time to come.
18 Likes
I thought that the jobs for taxonomist disappeard in Holland already years a go. If it is about taxonmy it is about DNA not photos…
Further…
Resitence is futile…in Holland 99% is done by volunteers i think
Zijn hoogleraarschap ging gepaard met een bestuursfunctie als wetenschappelijk directeur van het Rijksherbarium. Ook in deze hoedanigheid bewees hij de Nederlandse botanie een belangrijke dienst. Drie jaar na zijn aantreden begon een hevig snoeibeleid in het onderwijs. Herbaria en andere biologische collecties vielen niet langer onder de kerntaken van de universiteit en zouden voor de bijl gaan.
1 Like
Describing new taxa is only a rare event because few people have the time and money needed for it. Plenty of work waiting there if the AI takes away the rest of your work.
8 Likes
Indeed I suspect the vast majority of major ID contributors are working as volunteers. Even those of us such as myself who are lucky enough to have a paid biology job, rarely see “iNat IDs” falling within their work description, so we apply our skills after hours as volunteers. Same with taxonomy. The majority of (entomology) taxonomy papers that have crossed my desk in the past couple of years were authored or co-authored by volunteers (a group in which I include retirees). It seems to me that after a “golden age” of govt-supported taxonomy in agriculture, forestry, and academia, we’re returning to the Victorian model of it being practiced largely for free by those with wealth - either in spare money or spare time, to practice it as an avocation. It would be great if society valued taxonomists more, but the reality is that society writ large does not. But taxonomy will survive - it’s in good hands with the volunteers who do it for the sheer love of it rather than for a paycheque.
9 Likes
One wonders whether this will lead to a shift away from genomics-based back toward morphology-based taxonomy.
3 Likes
Conversely, even without knowledgeable AI, most hope is lost for professional taxonomists (speaking as one).
Also, regarding training AI to work – about a third of the species I’ve described are known from a single specimen, and another third from less than five. There are many plants here that are now down to less than 50 individuals, and have few photos of them. These are the most important ones to be able to identify, and the ones that an AI is least likely to get right.
10 Likes
Possibly. But on the other hand, how much can AI be used to increase total observation of species, and thus increase the number of people who care about potential plights? I know that before iNaturalist, I had very little interest in photographing and identifying species I see regularly, now I do 20-50/month, and sponsor a local penguin rehabilitation. Full disclosure: I do work for an AI company, and understand that my work there will likely one day remove the need for my job.
4 Likes
In my opinion, I think that using technology to pre-sort material will save a lot of time. It will change the nature of the work, though.
Right now it seems that we want to gather sufficient evidence that you do know the plant is X. Instead, imagine that you’d be trained to search for reasons to disprove the AI’s thesis that the plant is X.
This shift would, I think, still be helpful. Humans would still be doing what they’re really good at – recognizing patterns and anomalies within patterns. Critical and abstract thinking. Why not let a computer take all the initial measurements eg the length and curvature of the thorns or whatnot?
It would reduce the issue of confirmation bias because you’re explicitly trying to DISprove the thesis. and if you don’t find anything wrong or unusual or notable, then the thesis is accepted.
just a theory though. I wish I could… express my thoughts more clearly.
6 Likes