December 3rd, 2013 at 7:54 pm by Dr. Drang
I’ve been using Touch ID since I got an iPhone 5s in mid-October. Generally speaking, I like it, and I find it faster than the old swipe-and-passcode method, but I’ve felt compelled to reteach it my fingerprints twice already. I know this sounds impossible, but its recognition of my prints seems to decay with time.
I rescanned my fingers this weekend, and Touch ID has been amazingly fast and accurate since then. Just as it was when I first got the 5s, and just as it was a few weeks later when I rescanned my fingers for the first time. Just before each rescan, though, I was so frustrated with Touch ID I felt like throwing the phone across the room.
Sometime in early November I turned Touch ID off and went back to my old settings: four-digit passcode, but no need to enter the code if the phone’s been off for less than an hour. I tried this for nearly a week before giving up. Forget about the times I had to enter my code—just having to swipe to unlock was more than I could stand. Even bad Touch ID was better than that.
So am I fooling myself, or is it possible that Touch ID’s recognition gets worse with time? I don’t know enough about how it works to hazard a guess, but if the software is continually updating its internal “picture” of my prints with each scan, decay is at least possible.1 We tend to think that gathering more data always increases our understanding, but it doesn’t always work that way.
Google, the Oliver Twist of data, has had to adjust its Page Rank system because it had been gamed by link farms to give high rankings to shitty websites with weak, derivative content (but lots of ads). Even better, though, is the story of the Google Translate API, which was shut down a couple of years ago, because the indiscriminate use of it had reduced the quality of the corpus Google used to learn from. In a nutshell:
- Crummy Google translations were being posted to the web.
- Google would then scoop up those translations and add them to its knowledge base.
- Google’s translation engine got stupider.
EMpTy Pages, in the aforelinked post, called this “polluting its own drinking water.”
If Touch ID is polluting its own drinking water, I won’t be the one that finds out. There’s no way in hell I’m going to count its successes and failures and collect that data over a long enough period of time to draw any conclusions. But I’m going to be mighty suspicious if I feel like throwing my phone again by Christmas.
I’m assuming here that fingerprints themselves don’t change appreciably over a few weeks. If they do, the whole concept of Touch ID is flawed. ↩