AI asserts that our fingerprints could not be unique.

Columbia University’s research challenges the idea that each fingerprint is unique.

A US university team taught an AI technology to identify the same person from 60,000 fingerprints.

The researchers say the technique can recognize various fingerprints from one individual with 75-90% accuracy.

They’re unsure how it works.

“We don’t know how the AI does it,” said Columbia University roboticist Prof. Hod Lipson, who oversaw the project.

Forensics

The researchers assume the AI tool analyzed fingerprints differently than previous approaches, focusing on the direction of the ridges in the center rather than details like ridge ends and forks.

It’s not employing forensic signs from decades ago, said Prof Lipson. It appears to be leveraging the middle swirls’ curvature and angle.”

The result startled Prof. Lipson and undergraduate Gabe Guo.

“We were very suspicious… we had to check and double-check,” said he.

Some in the field may already know that.

Graham Williams, Hull University forensic science professor, said unique fingerprints were never set in stone.

“We don’t know that fingerprints are unique,” said he. To our knowledge, no two persons have the same fingerprints.

Crime scene

Columbia University’s findings may affect biometrics and forensic science.

Unidentified thumbprints at crime scene A and index fingerprints at crime scene B may not be forensically related, but the AI technology may be able to identify them.

The Columbia University team, which had no forensic experience, acknowledged the need for additional research.

AI technologies are taught on massive volumes of data, therefore more fingerprints are needed to advance this technology.

The model was developed using only complete, high-quality fingerprints, but partial or bad prints are more common in real life.

Mr. Guo said our program is helpful for forensics leads but not court evidence.

However, Staffordshire University associate professor of forensic science Dr. Sarah Fieldhouse did not believe the findings would have a “significant impact” on criminal casework at this time.

She wondered if the markings the AI tool was focused on stayed the same depending on how the skin twisted when it touched the print surface and if they lasted a lifetime, like traditional markers.

As with many AI-driven products, researchers are unsure what the AI is doing, making this difficult to answer.

The peer-reviewed Columbia University study will appear in Science Advances on Friday.

However, the Cheshire twins may be ahead of the pack. Grandmother Carol told the BBC her two grandkids can open each other’s iPhones with their fingers.

She claimed Christmas is when they showed me. We were told they were identical at birth, but I can see they’re different now.

She said her grandchildren can evade the phones’ face recognition.

Prenatal fingerprints develop. Last year’s research showed that they may be genetically comparable to zebras and leopards’ markings, an idea initially presented by codebreaker Alan Turing in the 1950s.