This portrait can still be distinguished as AI-generated but that won't be the case soon, Source: Pixabay

Why do AI-generated faces look more realistic than real ones?

Why do AI-generated faces look more realistic than real ones?

And how this will be a chronic issue in the digital space

Faces of white individuals created by artificial intelligence are consistently perceived as more realistic than faces of people of colour. White faces generated by AI now appear more "realistic" to us than actual human faces because they are, on average, created to be more proportionate and physiologically aligned with accepted beauty standards. However, this does not hold true for other skin tones. The reason behind this? Algorithms disproportionately train on white individuals, as revealed in a new study by experts from the Australian National University (ANU).

If faces of white individuals created by artificial intelligence are consistently perceived as more realistic, it could have serious implications for people of colour and reinforce racial prejudices online," warn researchers in a paper to be published in Psychological Science, a peer-reviewed journal.

This problem is already evident in current AI technologies used to generate professional-looking headshots. When used for non-white individuals, artificial intelligence alters their skin and eye colour, adjusting them to resemble white individuals as much as possible, researchers have found. This results in a kind of 'AI Michael Jackson' effect, mimicking his years-long skin bleaching and surgeries to adopt more Caucasian facial features.

The Hyperrealism of AI

Researchers have discovered that one of the issues with the "hyperrealism" of artificial intelligence is that people often don't even realize they've been deceived and are convinced that faces crafted by artificial intelligence are actually photographs of real individuals.

Once again the reason for this is that faces generated by AI algorithms tend to be more proportional and aligned with statistically more attractive representations than what can be observed in the general population, leading observers to mistakenly interpret them as images of real people.

However, Australian researchers caution that we won't be able to rely on these physical cues for much longer. AI technology is advancing so rapidly that the differences between synthetically generated and human faces will likely soon disappear entirely.

Recent research indicates that it's becoming increasingly difficult to discern synthetic faces from natural ones. Simply visiting the website is enough to understand what this entails.

Their synthetic faces are created by an AI algorithm known as a "generative adversarial network" (GAN), which consists of two neural networks, essentially computer models inspired by the way neurons are connected in the brain.

The hyperrealistic artificial depictions of human faces, which become difficult to distinguish from real photographs raises numerous questions and challenges, including potential implications for privacy, identity, and social interactions. As technology advances, we need to remain constantly prepared to understand and adapt to the changes it brings, so that we can address the multitude of challenges this brings along and potentially harness the benefits it offers.

This article is part of Read Twice – an EU-funded project, coordinated by Euro Advance Association that targets young people and aims to counter disinformation and fake news by enhancing their skills to assess critically information, identify vicious and harmful media content and distinguish between facts and opinions, thus improving their media literacy competences.

The contents of this publication are the sole responsibility of its author and do not necessarily reflect the opinion of the European Union nor of TheMayor.EU.



Growing City


Smart City


Green City


Social City


New European Bauhaus




ECP 2021 Winner TheMayorEU