Nothing to worry about:
The Chinese Academy of Sciences can now create 3D facial images from just a person's DNA, with stunning accuracy. The technology, named Difface:
The Chinese Academy of Sciences has developed an advanced AI technology called Difface that can create highly accurate 3D facial images solely from a personβs DNA. This breakthrough uses genetic data, specifically single nucleotide polymorphisms (SNPs), to reconstruct detailed 3D models of human faces, capturing subtle features such as the shape of the nose, cheekbones, and jawline125.
Difface operates by first mapping high-dimensional SNP data and 3D facial images into a shared low-dimensional feature space using a transformer and spiral convolutional neural network. It then employs a diffusion model, an AI technique that generates lifelike images from noise, to reconstruct the 3D facial structure from the genetic features15. The model was trained on a large dataset of 9,674 Han Chinese individuals, each with paired genome sequencing and high-resolution 3D face scans, enabling it to learn the complex relationships between genetics and facial morphology245.
The technology not only reconstructs current facial features but can also project how a person's appearance might change with age. Including additional information such as age, sex, and body mass index further improves the accuracy of the 3D reconstructions, reducing the average error to under 3 millimeters456.
This innovation holds significant potential for forensic science, allowing investigators to generate facial images of unknown individuals from biological samples at crime scenes. However, the method is still being refined and validated across diverse populations to ensure broad applicability beyond the Han Chinese dataset it was initially trained on46.