Tuesday, February 22, 2011

The Watson Effect, or, The Future of Bird iDentification

In light of the supercomputer Watson's recent dominance on Jeopardy, I was wondering if birding could undergo a similar technological revolution- one that would allow computers to identify birds in the field. It's an interesting concept, and, if it ever is actualized, a controversial one. How would this change birding? Would it be good or bad? Would birders become redundant? Is it even possible?

To answer the last point, yes, I believe it is, and it is probably feasible using current technology- no need to invent anything new. Two products lead me to this conclusion. One is Shazam, the popular app that allows you to figure out what is playing on the radio. It "listens" to part of the song, searches its database, and quickly comes up with an answer (and, of course, gives you the link to purchase the song on iTunes). Could this be adapted to birdsong? Probably. I'm not an expert on the technology, but it seems that it would not be too difficult to program the database to account for the variation in birdsong. It wouldn't be able to id everything, but it could probably do a lot. The second product is Google's new program "Google Goggles." Known as an augmented reality program, it allows you to point your phone camera at an object (say, a storefront) and gives you information about it, overlayed on the image. This is a bit more far-fetched, but it doesn't really seem like crazy statement to suggest that within 5 years a similar program could match a name to an image of a bird that you take, if it is a distinctive bird and a good photo. Within 15 or 20 years, the majority of birds could probably be identified with a camera and computer program (or app, probably). Would this be difficult to create? Yes. Is it economically worth it for anyone to try? Maybe, maybe not. Is it possible? Yes, it seems so.

On the surface, the prospect seems both enticing and a bit scary. Would it take all of the fun out of birding to be able to point your phone at a bird, and have the id immediately? It seems sort of like cheating- like you wouldn't be actually identifying the bird, just letting other people id it for you. And in some ways, this would be true. It probably would take some of the skill out of birding- particularly, it would narrow the gap in knowledge between beginning birders and someone like me- who can identify most birds given a good look, but is not particularly adept or experienced at tough ids. Maybe that's a good thing- if more people can feel accomplished at birding without putting too much effort in, then maybe more people will become birders. Additionally, technology like that could help a lot with bird research- instead of spending days analyzing data (say nocturnal flight call recordings), one could use a computer to make the identifications and give scientists important data.

On the other hand, the technology probably could not exceed human capabilities anytime soon- I don't think a computer could handle empids or gulls, considering all the variables involved (lighting, individual variation, hybrids, feather wear, etc). For expert birders, it might be completely useless- if the computer can id it, they can too, and they (rightly) wouldn't trust it with any ids they wouldn't be confident about themselves. In a discussion of a possible Thayer's Gull in the Northeast for example, one could not simply pull out their iPhone and point it at the bird for a definitive id- the result would just be another point for the human observers to weigh. In the same vein, the accuracy of the reports could be doubted- if a beginner iDentified (yes, that's what I've decided to call it) all of their birds, then there could be many erroneous reports in ebird and elsewhere of confusing but rare species. Obviously, the program would have to use probability as a factor, but this might not be enough, or could lead to the opposite extreme- rare birds are overlooked because the program does not even consider them.

All in all though, it seems that a computerized bird identification system could revolutionize birding, and is not necessarily that far off. And though I can already picture the experts grumbling about how "birders these days don't even look at the birds, they just point their phone at them," and do agree with them to a point, it seems that such a thing would be a net positive for birding. As a birder, I, for one, welcome our new computer overlords.

Do you? Leave your thoughts on whether a bird-identifying computer or app would be possible to create, and whether it would be a good thing or a bad thing for birding in the comments. 

Wednesday, February 16, 2011

Bird Photo Quiz 2

Well, given the response to my last photo quiz, which, if you haven't seen my comment, was a first year male Common Eider, I'm going to try to do a weekly photo quiz, if I can take enough photos to keep up.

This week's bird should be a bit easier than last week's, perhaps too easy (though I thought that about the eider too, so I don't know...). Photo taken in December in the Northeast somewhere. As always, leave your answer in the comments. 
Oh, and to add a comment if you don't have a google account, just select Name/URL. You don't need to put in a URL, though you can. 

Saturday, February 5, 2011

A Quick Quiz

Click to enlarge. Leave your guess or answer in the comments. I am actually not 100% certain myself as to what it is, but I do have a guess. Photo taken in January in Massachusetts, if that helps.