University of Helsinki uses artificial intelligence (AI) to generate attractive faces based on brain data.
What does our brain find beautiful and attractive? Can artificial intelligence (AI) machine learning learn from patterns of human brain activity the secrets to personal attraction? In a new cross-disciplinary study published last month in IEEE Transactions on Affective Computing, University of Helsinki researchers use AI and brain-computer interface (BCI) technologies to generate attractive faces.
“The study of aesthetics, or the perception of beauty and experience of attractiveness, has a long tradition within psychology and related disciplines,” the University of Helsinki researchers wrote. “Despite the common idea that taste is intensely individual, psychological research consistently shows a strong consensus on the visual features that are considered attractive. Symmetry in faces is known to be seen as attractive, perhaps because symmetry in general is an important evolutionary signifier.”
The researchers modeled personal attraction using four phases. First, a generative adversarial network (GAN) was trained using data from the CelebA-HQ that has 30,000 pictures of celebrity faces. Then study’s 30 participants assessed the images randomly sampled from the GAN as their EEG brain activity was recorded by a brain-computer interface. An AI classifier associated the EEG data with the participants’ assessment of the images. Next, new images were shown to participants as their EEG data was recorded, which were then classified by the images that the participants found attractive. In the final stage, new synthetic images were generated.
“While we instantaneously recognize a face as attractive, it is much harder to explain what exactly defines personal attraction,” the researchers wrote. “This suggests that attraction depends on implicit processing of complex, culturally and individually defined features. Generative adversarial neural networks (GANs), which learn to mimic complex data distributions, can potentially model subjective preferences unconstrained by pre-defined model parameterization.”
The generative brain-computer interface had a high degree of accuracy in a double-blind evaluation of the synthetic images compared against matched controls. “The results show that GBCI produces highly attractive, personalized images with high (83.33%) accuracy,” wrote the researchers.
“The generative brain-computer interface (GBCI) is able to generate a priori non-existing images of faces that are seen as personally attractive,” wrote the researchers. “Uniting BCI methods with a GAN allowed us to generate photorealistic images based on brain activity. Importantly, the generated images did not rely on external assumptions of the underlying data (such as what attributes make a face beautiful). Thus, the GBCI is able to generate attractive images in a data-driven way unaffected by current theories and opinions of beauty.”
Using a brain-computer interface, scientists enabled AI machine learning to create synthetic images that are personally attractive to humans with a high degree of accuracy. Ironically, although machine learning is able to produce synthetic attractive images based on what it learned from human brain data, we are not closer to understanding exactly what makes a face beautiful. What is beautiful remains in the brain of the beholder and the opaque black box of artificial intelligence machine learning.
Copyright © 2021 Cami Rosso. All rights reserved.
A version of this article first appeared on Psychology Today.