Asger Risom
AI Learns to Read Your Face — And Your Future
In an age where artificial intelligence is reshaping everything from medicine to media, one of the most debated frontiers is how deeply machines should peer into our lives — and bodies.
New technologies are not just interpreting data, but also making inferences about health, aging, and even longevity. As AI systems gain access to images and personal information, questions of privacy and power follow close behind.
An AI that sees more than meets the eye
Researchers at Mass General Brigham have developed an algorithm called FaceAge that can estimate a person’s biological age from a photograph — with potential implications for both medicine and privacy.
According to tech.wp, the system could help doctors decide on optimal treatments by gauging how “aged” a person appears compared to their chronological age.
The research, published in The Lancet Digital Health, is based on deep learning models trained on nearly 60,000 facial images. The AI estimates biological age by detecting subtle facial cues, such as skin tension or muscle tone, rather than more obvious indicators like hair color.
Better outcomes through better guesswork?
Testing the algorithm on over 6,000 cancer patients, the team found that those who looked significantly older than their actual age fared worse under treatment — a signal that appearance may reflect deeper biological wear.
In clinical simulations, FaceAge helped improve doctors’ accuracy in predicting short-term survival compared to their unaided assessments.
This could be particularly useful in tailoring treatments like aggressive chemotherapy or surgery to individual resilience, rather than age alone.
Powerful tool, dangerous temptation
But researchers also warn of risks. FaceAge, in the hands of insurance companies or employers, could be used to discriminate based on predicted longevity. As one lead researcher noted, “That’s something we absolutely have to guard against.”
The team is now developing a second-generation version of the model, trained on a larger data set of 20,000 patient images. They’re also testing how variables like lighting, makeup, or cosmetic procedures could trick the system — an essential step toward responsible implementation.
The technology may be revolutionary — but like all revolutions, it comes with questions that demand answers.