Artificial What?

I started this blog intending not to write anything controversial, but I've really got to make some comment about AI. Forgetting the dangers and huge inconveniences outside of birding, I find it very much a double-edged sword, but at least we are able to choose which bits we use. After all, modern cameras and processing programmes use a lot of AI, but there are still manual settings available in many cases, and lots of options. They are far better at most things, but if you leave it to "auto" all the time you won't get the best photos. An interesting "effect" which only happened on one frame was the scattering of the light from a laser pointer of the Sri Lanka Frogmouth which is shown below. I've no idea what caused it, but thankfully it only affected one frame (the laser was still on for subsequent two frames but they were fine).

It's up to the individual, but I think most birders take a similar view to me when it comes to apps like Merlin. I always much prefer to identify something for myself, or if I can't, then take a photo for further "research". Some think differently, and will send records to ebird as they walk along. I know it's the modern approach, but I can't bring myself to rely on an app for identification. The big drawback is the location, if you happen across a real rarity I don't think there's anything in the programming that allows for that. I don't see Merlin picking up the 2006 Long-billed Murrelet at Dawlish. I'll give it a go, now!........

......As I suspected, Merlin hadn't a clue, at least it didn't suggest something ridiculous, or worse, Ancient or Marbled Murrelet. Removing time and location gave a list of best matches- in order of probability- Red-throated Loon, Long-billed Murrelet, Pigeon Guillemot, Marbled Murrelet, Black Guillemot. A reasonably competent birder with some knowledge of auks worldwide would surely have narrowed the possibilities to Murrelets, I can only think the relative abundance of Red-throated Diver caused the AI to suggest it. When I put the location back in, the App reverted to "don't know". Why not Black Guillemot or Red-throated Loon, eminently possible at that location?

I do use Obsidentify for moths, and in many cases it's better than me. I always check with my books and the Norfolk Moths website to confirm but if it gives a probability of over 95% it's almost always correct. I know many moths can't be positively identified without dissection but this is reflected in results, and it is brilliant at micros. The only slight drawback is that English names aren't always the same (I believe the app is Dutch) so Scarce Bordered Straw becomes Cotton Bollworm, which until I looked it up I didn't believe occurred in the UK. As it's only a migrant that's understandable, but any result of less than 90% needs to be viewed with suspicion. Some suggestions are downright hilarious, I've had Red Phalarope, European Honey Buzzard, Common Toad, and Wild Boar suggested for my back garden (I wish!), the first two for micros! I'm quite happy to use an App for mothing, since you don't have to  consider such things as call, jizz etc., but then I don't take it too seriously. Birding in the field, AI is restricted to my camera. I went out to Carlton Marshes yesterday and photographed the Long-billed Dowitcher. Without AI that shot wouldn't have been possible, but on the way back I found a Cetti's Warbler and got a good view, albeit brief, without the bird singing. That was far more satisfying to a birder.