markscherz.bsky.social
Curator of Herpetology & Associate Professor of Vertebrate Zoology at Natural History Museum of Denmark • ERC StG: GEMINI • Co-host of SquaMates Podcast and AnatomyInsights on Youtube • He/Him
510 posts
3,743 followers
1,268 following
Regular Contributor
Active Commenter
comment in response to
post
I occasionally get to be this guy in meatspace too.
comment in response to
post
Exactly how I felt about achieving tenure.
comment in response to
post
You use a special castable resin for printing these pieces, right? I just acquired a 3D printer and am beginning to understand the full scope of your incredible models.
comment in response to
post
"Unfortunately it turns out that iNat staff, like many of the rest of us, tend to be nature nerds who are happier hanging out in a bog with frogs than interacting with people, and thus the communication skills may be a little … lacking … at times."
very relatable
comment in response to
post
Unbrielievable.
comment in response to
post
Is all the segmentation done by hand, or do you have scripts that help? Do you segment in Amira/Avizo, or something else?
comment in response to
post
When I was about his age, I threw a pocket knife into one of my toes. A few years earlier I closed one on my finger and never told my parents because I wasn’t supposed to be playing with it.
comment in response to
post
Shit, hope he’s doing okay!
comment in response to
post
Further thoughts: bsky.app/profile/mark...
comment in response to
post
I also am an emphatic believer that iNat is systematically inflating identification confidence, because the current system is hard to implement fuzziness and unidentifiability in. And I am worried that genAI will make this worse, not better.
comment in response to
post
This could be done with the help of genAI, but it will need expert validation, and ultimately will likely struggle to capture whatever it is that the computer vision model is actually using.
comment in response to
post
If they want to be a field guide and have this information, there needs to be an infrastructure change that allows for keys, feature lists, trait-based ID, and taxon comparison. None of this really exists right now.
comment in response to
post
The computer vision model on iNat is incredibly good. I mean, really amazing. And I can see the desire to have in words what the computer vision model is finding so that users can learn to ID too. But I think @inaturalist.bsky.social needs to figure out what it wants to be.
comment in response to
post
I appreciate this. There’s a real need for nuance, and it is not common. I think the reaction we are seeing is a result of months of disappointment by the scientific community in the rampant adoption of genAI without a clear need, transparency, etc. from students, services, publishers, etc.
comment in response to
post
If you are a user of iNat, this project may well improve your user experience. If you are a taxonomist, it may decrease your incentive to participate.
comment in response to
post
Even if it is good for the casual uploader, if it causes experts like me to be frustrated, disappointed, and ultimately leave the platform, the entire project will collapse. iNat NEEDS experts. Human experts. A lot more than it needs TaxoBot4000, the regurgitating robot.
comment in response to
post
I think the team at @inaturalist.bsky.social need to sit down and read the comments and feedback that they are getting on their blogpost, and across social media, and seriously consider if this endeavour is going to be able to deliver what they want it to deliver, without estranging their user base.
comment in response to
post
What I really don’t understand is why community-based organisations like iNaturalist are hopping into the genAI train without asking their communities whether they could see a use case for genAI. The response has been almost universally negative here, and for others before.
comment in response to
post
Also, my own abilities (and the taxonomy) have changed over a decade using iNat. Accommodating these subtleties will be extremely hard for an LLMonster.
comment in response to
post
In this case, an LLM is very close to 1:1 replacement. Maybe useful if I die or leave the platform, but otherwise a bit of a slap in the face.
comment in response to
post
In my case, the community is small. In fact, for many taxa, it seems I’m the only active iNat user who actually knows how to identify them properly. After all, I described quite a few of them myself.
comment in response to
post
(Though I have serious doubts that even this synthesis will be half as good as that which the already existing experts on iNat would be able to put together through a wiki or equivalent community-led annotation/guide)
comment in response to
post
I can see the attraction of LLMs here; when the quality and volume of input is high, the quality of output can be high, too. In groups where there is a lot of taxon expertise, there’s a large community to draw on, and it can synthesised.
comment in response to
post
careless best hit assignment, contamination, or data confusion. 😬 Not good.
comment in response to
post
That is to say, just because they are paying so much, does not mean they are necessarily more engaged with the teaching material. Though I grant you that I cared more as a foreign fees student at Edinburgh University, than some of my fee-less classmates. More on the line.
comment in response to
post
Many answers you have are from people in countries that have tuition fees. Denmark has fees only for foreign students (most Danish students are paid to study, at least partly). I am not sure if you see students here less motivated than in other countries, even though it is not ‘costing them dearly’
comment in response to
post
TL;DR requirements differ, pedagogical education of lecturers is attempted, but sometimes the tried and true rote learning of facts is the only way to convey what the students on a given topic need to know.
comment in response to
post
What we might then do in a university setting is add practicals or other activities that break up the droning lecture, and reinforce the learning. That can be key. But we can’t do that for everything—there is simply too much to learn.
comment in response to
post
In many cases the didactic techniques involved with teaching those facts—because there are so many of them, often much more detailed than K-12 required—are pretty limited. Is classroom teaching/lecturing the best way to acquire that knowledge? I don’t know. But I think it helps.
comment in response to
post
But on the other hand it dramatically reduces the amount of information we can convey. It’s much slower. In fields that are a lot about learning facts and figures, like medicine and zoology, these methods struggle. We *need* rote learning because the students *need* facts.
comment in response to
post
On the one hand, the students do prefer it; it keeps them activated and attentive. It is often easier for them to remember things from the class, and ideally it should help especially with grasping concepts and reasonings behind them…
comment in response to
post
At @ucph.bsky.social, faculty who teach are required to do at least some pedagogical training every year, and many also required to take basic or more advanced pedagogical training. That training focuses on bringing a lot of the K-12 techniques to the university classroom. I have mixed feelings…
comment in response to
post
Does titchy share that etymology?
comment in response to
post
Pretty sure these are actually the historical plans for Settlers of Catan
comment in response to
post
We are starting to think it’s shotgun shot, tailored for small animals. This specimen is from the Noona Dan circumnavigation in the 1960s.