A child who looks like the mom, dad, or Brad Pitt: AI tailors the physical appearance according to the customers’ wishes
Should the child look like the dad, the mom, or perhaps like Brad Pitt?
In Spain, the company Fenomatch has developed an algorithm that, using up to 12,000 facial measurements, helps a patient undergoing fertility treatment find an anonymous donor who matches a certain physical appearance.
“The clinic uploads a photo of the person the baby is supposed to look like, and then an AI algorithm finds the best facial match among the photos in the donor database,” says Sergio Gonzalez, product owner at Fenomatch.
This service is typically used by, for example, single mothers who want the child to look like them. On the other hand, a couple usually wants the donor to look like the parent who has not been able to provide sperm or eggs themselves.
Fenomatch has been around since 2018 and, according to Sergio Gonzalez, was founded by data analyst Luis Arenaz Villalba, who himself had gone through a fertility treatment and was frustrated that the couple is not able to see their donor.
"In Spain, you can only use anonymous donors, but he wanted a more transparent process, so he founded Fenomatch,” Sergio Gonzalez says and emphasizes that you can of course also match your image against open donors.
Approaching a more accurate appearance prediction
The company has so far helped more than 8,000 families at more than 100 different fertility clinics find the right facial features in a donor, and according to Sergio Gonzalez, most of their customers are from South America and European countries that mainly or exclusively work with anonymous donors.
Although it is in most countries possible to obtain information on phenotypic traits such as hair colour, eye colour, height, and education, Fenomatch claims that their technology can approach a more accurate appearance prediction—without actually seeing the donor.
In some of the markets that Fenomatch is currently exploring for new potential customers, the wishes may be different.
“Especially in the USA, where it is the patients themselves and not the clinics who choose a donor, we believe that many single mothers would like to be able to show a photo of a celebrity and find a donor who looks similar,” Sergio Gonzalez says.
Fenomatch is not actually able to see the many photographs and other data used by the algorithm. But they have written the more than two million lines of code which make up the encrypted and ISO-certified online platform the fertility clinics can subscribe to and enter their donor data into.
Sergio Gonzalez emphasizes that they cannot promise that the child will look as much as the desired outcome as some patients believe it will. But he says that after running the algorithm 8,000 times, he has great confidence that the algorithm can copy the most important features almost as well as if you had seen them in a photograph.
The platform is based on three types of matching. In part, you are informed of the phenotypic traits of the donor, which the clinic has entered based on a fixed form, which can include hair colour, height, and leisure interests.
The next step provides the opportunity to upload data from gene sequencing, which enables screening for similar mutations in genes that can cause illnesses in both donor and patient.
The third option is this face-matching with the many thousands of data points that measure whether the people in the photographs have the same distance from nose to mouth and so on.
Research shows that if people look alike, they probably also share genes. How do you avoid problems with that if single mothers want the child to look like themselves?
“Of course, nobody wants inbreeding. We therefore have a blocking mechanism which disqualifies a donor if the match is greater than 80 percent because it means they could be a relative,” Sergio Gonzalez says.
As for diseases, Sergio Gonzalez emphasizes, many fertility clinics today offer genetic testing of sperm and egg donors, and that data can also be uploaded to the platform, so that these results are also taken into account before choosing a “face”.
In the user interface, each match within the three categories will be highlighted with a specific score, and then the clinic or the patient can weigh up the benefits vs. risks.
In the future, Fenomatch wants to connect patients, clinics, banks, and donors all over the world.
“That way, both clinics and patients get a huge selection of donors to choose from,” Sergio Gonzalez says.
Fortsæt din læsning
Techtopia #277: GPT-4 – spøgelset i maskinenPlusIT27. marts