Chinese scientists have introduced Difface, a neural network capable of generating three-dimensional portraits of people using only their genetic material. The technology opens new horizons in forensics, anthropology, and historical research.
G. Ostrov
A scientific team from China has made a significant breakthrough in applying artificial intelligence to genetic material analysis. The innovative neural network model called Difface demonstrates the ability to transform a person's genetic data into detailed three-dimensional facial images.
To train the system, researchers used an extensive database including three-dimensional facial scans and corresponding genome fragments from approximately 8,000 representatives of the Han ethnicity — the world's largest ethnic group, comprising about 91% of the People's Republic of China population.
Current Capabilities and Limitations
At this stage of development, the technology demonstrates promising but still limited capabilities. The average error in facial point positioning is about 3.5 millimeters. With the addition of supplementary biometric parameters of the subject — such as gender, age, height, and body mass — accuracy improves, and the margin of error decreases to 3 millimeters.
During testing on a control group of 2,000 participants, the Difface system demonstrated the following results:
- In 3.3% of cases, the model was able to unerringly identify a person by their DNA;
- In 42% of cases, the correct match was included in the list of the 20 most likely candidates.
Development Prospects
The developers view the current version of Difface merely as a first step toward creating a more sophisticated system. Their plans include significantly improving reconstruction accuracy and adapting the algorithm to work with genetic material from various ethnic groups. To achieve these goals, they intend to substantially expand the training database.
Successful development of this technology could have revolutionary significance for several fields:
- Historical anthropology — the ability to visually reconstruct the appearance of historical figures based on preserved genetic materials;
- Forensics — creating accurate portraits of suspects from biological traces found at crime scenes;
- Genetic medicine — improving understanding of the connections between genotype and phenotypic manifestations.
Despite the promising prospects, the technology also raises serious ethical questions related to the privacy of genetic data and the possibilities for its use.
As this technology continues to advance, we may witness fundamental changes in our understanding of the relationship between our genes and physical characteristics.