At the University of Helsinki, they used an AI to find out which faces are attractive just by looking at the neural activity of the participants.
Defining what physical beauty is has been one of the debated philosophical questions for centuries. And although beauty also depends a lot on the person who looks at it, now an Artificial Intelligence (AI) can know which faces we find beautiful just by reading our brain waves. But how is this possible?
New research on this topic has been published in the scientific journal IEEE Transactions on Affective Computing. To conduct the study, they invited about 30 participants to a special Tinder session. In order to read the mental waves of these people, they used “electroencephalography (EEG) measurements to identify what kind of facial features people found attractive and then sent the results to an artificial intelligence (AI) program,” as explained in Science Alert.
They also used a machine learning system known as an adversary generative neural network (GAN). At first, this AI was made familiar with the subjects who were considered desirable by showing it pictures of celebrities. Then he was asked to make other people with physical beauty from those 200,000 images that he had been given to learn. And the result was perfect new faces, unattainable for the observer.
Caps with electrodes to measure brain activity
At the special Tinder session, the participants were introduced to these new images; But they also wore caps with electrodes to measure their brain activity while they looked at the pictures of their perfect dates. Experiment participants didn't have to swipe right if they liked the person they saw, they weren't really there to flirt - and those people weren't real too. However, the hats gave the researchers all the information they needed: "They didn't have to do anything other than look at the images," explains cognitive neuroscientist Michiel Spapé; one of the study's principal investigators. "We measure your immediate brain response to images."
In this first session, the GAN interpreted the brain waves in terms of how attractive each artificial face was to observers. He then used that information to generate new faces, based on the neuronal response of each observer.
During the next session, the observers were asked to rate the images. Among them were the new faces generated according to the previous activity and others totally random. “Ultimately, the results validated the researchers' test, and the participants rated images designed to be attractive as attractive in about 80 percent of the cases, while the other faces were selected only 20 percent of the time. ”, They explain in Science Alert.
Small AI study on attractiveness
Only 30 people participated in this experiment conducted by a team of psychologists and computer scientists from the University of Helsinki in Finland. However, it is a good example of the extent to which AI is being refined for things as intimate as an attraction to other people.
Beauty can be very subjective, but once AI knows how you like faces it can help you choose a new partner. Although, unfortunately for these participants, they were not real people.
“If this is possible in something that is as personal and subjective as attractiveness, we can also analyze other cognitive functions such as perception and decision-making. Potentially, we could guide the device towards the identification of stereotypes or implicit biases and better understand individual differences ”, explains Spapé.
Human behavior is very interesting for psychologists, hence the importance of experiments like this artificial intelligence to learn more about how we appeal and beauty in others. In addition, we are thus aware of the extent to which AI can know us and learn from us.
0 Comments