Reading Time: 4 minutes
I was adamant that the artificial intelligence was wrong. My family had given me a bird feeder that came with a camera. As the birds appeared, the camera would take pictures and make an assessment of the bird’s species. It gathered a location for the camera, so it could geo-locate its selections, which made sense. It was also why I found it so irritating that it suggested that the bird I was looking at—that I had referred to as a purple finch for my WHOLE LIFE—was a house finch.
I have been interested in birds for as long as I can remember. I do not seek them out but, when I’m out on a bummel, I look for them. Everywhere I have lived, there are common birds. You see the same ones over and over until you start to learn their names. I have never been good with remembering bird song, although I have been known to whistle back at a bird (pretty sure it was a black-capped chickadee although it wasn’t the “chick-a-dee-dee” song they are famous for) of a morning.
It’s like the bon mot attributed to President Grant, among others: “he knew only two tunes; one was ‘Yankee Doodle,’ and the other wasn’t.” I am actually pretty good at hearing and recalling music but bird song largely eludes me. Also, yes, I’m aware of the Merlin app, the Shazam for birding. I am not sufficiently interested in birding to bring any additional technology to it other than my camera.
This is why I was so certain in my identification. The purple finch is a red-colored finch that has been around the Midwest as long as I have. I’ve specifically looked up in the past why it was called a purple finch when it wasn’t purple. I was very confident in my identification.
I’m also a librarian, though, and I know I can be fallible. So I looked it up in my bird book. This is a Golden Book field guide I’ve had my whole life. Surely it wouldn’t lead me wrong? The fact that this upstart camera with its software was suggesting that this bird was a house finch couldn’t be more wrong. The book said about the house finch:
Abundant in bottomlands, canyons, suburbs, and ranches in the West; uncommon but increasing and spreading in East, especially at feeders.
A Guide to Field Identification: Birds of North America, p. 296 entry on House Finch
Uncommon in the East! Ha! I mean, you can just look at the map next to the entry in the book linked boave. There was no coloration east of the Rocky Mountains.
But the artificial intelligence was insistent. In fact, it was your typical AI. Certainty without accuracy. It is not like we live in an aviary. We see about 10 species: sparrow (house, etc.), finches (PURPLE, American gold), grosbeaks, cardinals, cowbirds, red-winged blackbirds, and so on. We hear a blue jay but the pipsqueak stays in trees, warning the world about its impending doom. Very occasionally the garden clears out and you can often see the hawk circling above or, on one memorable occasion, being harried by the littler local denizens.
House finch. House finch. House finch.
The thing is, it was also clearly making errors. The AI claimed one set of images was a house finch, a northern cardinal, a purple finch, and a house sparrow. It was one bird in the pictures.

Wut? Same, finch, same.
Let’s leave aside the lack of a serial comma, Mr. Art Intel. This is a 30 second clip. There is almost no way that 4 different birds would visit a feeder amicably in that period of time. That’s like those universities that schedule two classes on either side of campus, giving a student 5 minutes to walk two miles.
More often than not, it makes an unrecognized species alert because the bird wasn’t courteous enough to stand in front of the camera. I have started to refer to the feeder as the Bird Butt-y, because I see more photos of bird butts than anything else.

I can’t say this for a fact but this appears to be an AI that should be drawing from a limited amount of information that is largely fixed. The National Audubon Society offers images of “over 800” birds for North America. Cornell’s All About Birds says there are 760 species. The point being, this is a pretty well-known body of information. It is also geo-located, so it should only be making choices about birds in my area. I would have expected the AI to get it right every single time.
That’s not really fair, though. Also, it’s bird identification. No one is going to die if the software identifies a bird as one species when, in fact, it is another.
And there is also the possibility that the AI is correct in cases that I disagree with it. I am big enough to admit that I am as fallible as some AI. I don’t want to be that guy. I am not a birder and I am using a bird identification book that was published over 50 years ago. I am clearly not an expert. In that span of decades, it looks like the house finch has, like so many creatures, skipped Flyover Country and is now prevalent on both coasts.

My world is, as you can imagine, shaken. The very foundations of my knowledge are on shifting sands. Or, rather, I probably should recognize that books that are 50 years old may need to be updated. (Also, my ego is a bit bruised from the amount of ridicule I sustained when I disclosed this fact to my family.) A quick look at All About Birds confirmed my worst fears:
The House Finch is a recent introduction from western into eastern North America (and Hawaii), but it has received a warmer reception than other arrivals like the European Starling and House Sparrow. That’s partly due to the cheerful red head and breast of males, and to the bird’s long, twittering song, which can now be heard in most of the neighborhoods of the continent. If you haven’t seen one recently, chances are you can find one at the next bird feeder you come across.
House Finch, All About Birds
Nice: “the next bird feeder you come across.” Rub my beak in it.
The bird feeder has been a huge amount of fun. You can run a live stream, which I don’t for both battery and surveillance reasons. In fact, I was very touched when my family gave it to me. They know that I am extremely resistant to any internet-capable surveillance technology like an Alexa home assistant, Samsung fridge, Google Chromecast, a Ring device or a Tesla car. While none of those devices will ever land in our home, this little feeder with its camera and it’s AI identifications has been a real hit.
I even have come to appreciate the artificial intelligence’s attempts at guessing what a bird is from it’s butt feathers. But I am learning to be gracious, and, in the case of birds, anyway, giving it the same latitude for error that I require. The big difference is, whether birds or more serious areas, I will always be able to validate an AI output but I need to be sure to use the right tools myself.