This AI GAN Can Tell Your Preference of Attractive Faces Through Machine Learning [STUDY]

We all know that the way we perceive beauty among people is different. Like the common saying, "Beauty is in the eye of the beholder, "you might think that the perception of beauty is only dependent on the person who's seeing it.

However, when technology interferes with this discussion, the concept of beauty will be changed, as artificial intelligence has different discoveries in this finding.

AI GANs Can Actually Tell You What Faces You Find Attractive

In a report by Science Alert, researchers said that they could know if the features are enough to be categorized as 'attractive' to people. Through the use of electroencephalography (EEG) measurements, that will be possible. The AI program will then evaluate the results.

The generative adversarial neural network (also known as GAN) is an innovation of machine learning. In the study, GAN was able to establish familiarization in the faces that people found beautiful. The system can now make new sets of faces with synthetic beauty, close to the real ones.

In the experiment, experts from the University of Helsinki ran a similar Tinder set-up test, with 30 participants who participated.

The subjects were tasked to view the sets of faces from the computer, and none of them were real persons. Instead, the Finland-based team used AI portraits that only look realistic. They were produced from around 200,000 images of random celebrities.

Furthermore, the participants were instructed to wear elastic caps packed with electrodes to monitor brain activity while looking at the faces. On the said dating site, you can swipe right if you like someone, but that is not the case here.

Through GAN, the neural activity is evaluated, and the brain interprets the responses of the person's perception of attractiveness to different faces.

The machine-learning system was also able to create new faces which the person perceived through the attraction identifier set by EEG.

Michiel Spape, a cognitive neuroscientist, explained that the participants are only instructed to look at the images. By a simple command, his team was able to conduct measurements regarding the brain response to the faces in the images.

The Second Experiment

The next experiment involves people who were asked to rate the faces with corresponding 'attractiveness' value. The group of faces flashed in random.

This led to a conclusion that people were attracted to around 80% of the faces, while the rest accounted for 20%, which other participants have selected.

Although this is only a small study, seeing AI GANs in the experiment just extended the machine-learning capabilities beyond human comprehension.

"If this is possible in something that is as personal and subjective as attractiveness, we may also be able to look into other cognitive functions such as perception and decision-making," Spape said.

Spape ended that the study could form a device that aims to set apart implicit. Through this, people can have a better grasp of perspective differences to others.

To assess the full study entitled "Brain-computer interface for generating personally attractive images," visit IEEE.

This article is owned by Tech Times.

Written by Joen Coronel