Fig-leaved Goosefoot (Chenopodium ficifolium) Photo  By Dalgial – Own work, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=15995459

Dear Readers, naturalists (including myself) are often nervous about using new technology to identify plants or insects or, indeed, anything else. “What’s wrong with a field guide!” we cry, “and get a hand lens! Or some binoculars!” But gradually, many of us have realised the potential of phone apps like Merlin or Plant Net, or some of the extremely useful expertise-based websites, such as iNaturalist or Facebook groups. I, for one, have been astonished at the wealth of expertise on some of the groups dedicated to wild plants, spiders and bees and wasps on Facebook.

However, AI is a whole new ballgame, so in my Botanical Society of Britain and Ireland (BSBI) magazine this month I was very interested to read an article by botanist David Steere. He decided to try out Google Gemini AI to see how it would perform, and how it compared to his own observations.

First thing was, don’t upload your photos already named with the species, or the AI will ‘cheat’.

Secondly, Steere found that you needed to upload as much information as possible with the photo – size, habitat, plant height etc. He then told AI what he thought it was, and asked it to verify. Its  great strength is looking in detail at the photo, going down to pixel level to look at hairs on the stem etc (though it occurs to me that the quality of the photo will have an impact on this).

If you don’t include additional information, AI does get things wrong – Steere describes how it  decided that the Fig-leaved Goosefoot in the photo above was a young Tree Mallow (Malva arborea). When he gave it the size of the plant  and the location, AI got it right. The key thing here is, of  course, that Steere knew what he was looking at, and could correct the AI. A lot of us won’t have that depth of knowledge.

Steere found that the sources that the AI was using for identification were sound – including Clive Stace’s recently updated Flora of the British Isles, and the BSBI’s own website. It was also very good at comparing closely related plants, and compiling a useful table of the differences between them. But he came to the conclusion that, at present, Gemini AI at least is only really useful for botanists with a bit of experience, who have a rough idea of what they’re looking at. I would add that it makes sense always to go back to the field guide and see if what you’ve been ‘told’ makes sense.

I suspect that AI will have a role to play in identification, but it’s far from the last word. Personally, I’d rather trust a human being with experience and knowledge, who can explain why what I’m looking at is what it is. Plus, it’s the enthusiasm of humans for their subject which is so contagious and brings such joy. I’ve not yet met an AI engine  that does the same.

I am very curious to hear about how you are all getting on with the AI revolution (if indeed you are :-)). Pop something in the comments if you’re using it! Like all technologies, it’s neither good nor bad, it’s just how humans choose to use it.