3 ms·
Many people have problem telling identical twins apart, so under that metric it's true that human faces aren't globally unique. There's even a project to find "
by dalke 11y ago
Many people have problem telling identical twins apart, so under that metric it's true that human faces aren't globally unique. There's even a project to find "twin strangers", which has had some success (see the images at the bottom of https://twinstrangers.net/ https://twinstrangers.net/ , though they also had the same makeup applied, which makes them look even closer).
> how can fingerprints not be unique?
How can the hash of two different strings be the same? It's all part of the Birthday Paradox, combined with that it means two identify two physical objects as the same thing.
Now, at some point no two human-scale objects are the same. Even with two mass-produced objects, there are slight variations in production because of non-zero tolerances. For example, two sequential $1 bills differ of course by the serial number, but also contain minute differences in the paper.
So as you say, we need to distinguish between how fingerprint analysis is done, vs. the most detailed possible analysis. And we know of several cases where fingerprint identification has been incorrect. One is Brandon Mayfield, a US attorney who the FBI believe, on the basis of a fingerprint match, to have been one of the Madrid bombers.
As a result of that case, and similar ones, people have started to ask the police and others for data which shows that "all fingerprints are unique" is really true. This myth started over 100 years ago, but was based on a small sample size. There's never been the analysis needed to figure out the actual error rate, especially if there are, say, 100 million entries in the system.
See also https://en.wikipedia.org/wiki/Fingerprint#Validity https://en.wikipedia.org/wiki/Fingerprint#Validity . And notice how that Wikipedia page also includes "No two fingerprints have ever been found identical in many billions of human and automated computer comparisons.[citation needed]". Where is that citation?