"who knows?"
--> and yet you + others continue to unhelpfully speculate. I'm unsure you understand what 'pivot' means if that's how you interpret a proposed demo which may never progress further after testing

also fyi using iNat data in research ≠ 'working with them'

have a great day

"pivoting to generative AI" is a strange and very misleading characterisation of the planned demo that does not accurately described the currently intended tool whatsoever

did you use the Shiny app, or the R package? The latter allows you to use infinitylists for anywhere in the world, the former only for select regions as a demo of the functionality

humans make mistakes all the time when IDing stuff. Could be any number of reasons: it's a woodpecker, and it's olive-coloured, so that's what they typed (this is why there are so many black house ant misIDs). Or, there's a local species with the common name olive woodpecker as well

that first ID wasn't made by AI, and that species isn't one of the suggestions at all for that record when you query the AI. In fact, when you check it the top suggestion by the AI is the correct species ID

nice one Francesco! I've added both species to iNat for you ;)

We extensively discuss why accuracy was so high, but key is that many experts have already spent years sharing their knowledge on iNat. It's clear: consistent expert engagement with citizen science, and the learning by citizen scientists that results, drives high quality biodiversity datasets.

What did we find? Almost 11,000 records were reviewed by at least one expert during the event. Of the 7,000+ records from our dataset that were identified to species or finer, 92% of them were correctly identified. And if only considering the ~3,500 Research Grade records, these were 97% accurate.

We recruited over 50 taxonomists, botanists and other experts with knowledge of the Western Australian flora, and gave them three weeks to assess the ID accuracy of thousands of iNat records from three regions in WA as part of an 'expert ID blitz'.

But discovery hasn't finished! Since we ran the numbers earlier this year, at least 50 new species (mostly introduced) have already been found for Royal, with a smaller handful for Yosemite. There's still lots to discover, so head out there with your camera and plant press!

Second, the most complete picture of the plant biodiversity from each park relied on combining all three data streams: each had unique species that the other two didn't. Excitingly, @inaturalist.bsky.social records contributed 63 new species for Royal, and 20 for Yosemite

Key take homes?

First, compiling species lists for a reserve requires lots of manual curation! We found many errors, including misIDs, wrong coordinates, taxonomic issues, and databasing errors. These result in differences of 100s of species across curated vs non-curated lists

And what about Yosemite? It reached 1,632 species, including 1,409 natives, in an area of 300,000 ha.

How many vascular plant species do you think are in Royal NP? We compiled a total of 1,414 species! This includes 1,146 natives, more than the entire British Isles in an area of just 15,000 ha.

A few years ago on the midnorth coast, I walked through a tall, closed wet forest with tens of thousands of razor grinders calling simultaneously, it was close to unbearable. The overall sound was so loud I felt like I was almost hallucinating, and the air seemed to be almost pulsing with sound.

@siobhanleachman.bsky.social a mysterious, slightly dodgy online store that seems to have now disappeared off the face of the earth...