Summary of Key Points
Recently, AI-generated “standard faces” have sparked widespread physiological disgust among netizens due to their striking similarity and artificial appearance. The topic has even made it onto hot search lists, with the People's Daily criticizing the trend, stating, “We’re tired of AI-generated faces; it’s time for a return to the ‘sense of authenticity’.” These AI-created faces all share common characteristics: large round eyes, high noses, and pointed features, along with a superficial smile that lacks genuine emotion. This phenomenon is attributed to a combination of three factors: technological limitations (algorithms naturally tend to produce average-looking faces), cost considerations (pre-made templates are inexpensive and less prone to copyright issues), and commercial pressures (the demand for content is high, but there’s a desire for speed and efficiency). As a result, aesthetic homogenization has occurred, potentially damaging the internet’s aesthetic ecosystem.
Detailed Analysis
1. Why do AI-generated faces cause physiological disgust? – The “Uncanny Valley” Effect
The discomfort we feel when looking at AI faces is not just a matter of being fussy; it’s an instinct evolved over millions of years. Japanese scientists have identified the “uncanny valley effect”: when something looks 95% human-like but has minor discrepancies, our liking for it plummets, even turning into fear. AI-generated faces fall precisely into this category. While they have all the necessary facial features, the details are often flawed—pupil reflections that aren’t aligned, unnatural muscle movements, and skin textures that appear plastic. Our brains’ “fusiform face area,” responsible for recognizing faces, can detect these issues in just milliseconds, and our fear center (the amygdala) immediately signals, “This isn’t a normal human!” We may not be able to articulate exactly what’s wrong, but our subconscious mind rejects them, leading to an indescribable sense of disgust. The more we’re exposed to homogeneous AI faces, the more this discomfort intensifies, eventually turning into aesthetic fatigue.
2. Why do all AI-generated faces look so similar? – A compromise among technology, cost, and law
It’s not that AI can’t create diverse faces; rather, creators deliberately use the same templates.
- Technological limitations: AI models learn from a large collection of publicly available images on the internet, which tend to represent mainstream aesthetic standards (young, symmetrical faces with standard features). Algorithms assume that what appears frequently must be correct, resulting in an average-looking outcome. To avoid errors (e.g., misaligned eyes), they narrow down the range of possibilities, limiting creativity.
- Cost and law: Developing unique AI faces can lead to copyright issues or require significant investment in debugging. Using pre-made templates is both cheaper and risk-free, so everyone ends up using the same generic images.
3. Why has this approach backfired despite low costs and high demand? – The pursuit of speed and efficiency overlooks user experience
Although the cost of AI face generation has decreased significantly (creating a single face may only cost a few dollars), the demand for content in short videos and micro-series has skyrocketed. Producers focus on mass production without considering whether users will accept such content. Perfect but soulless faces may seem fresh at first, but after repeated exposure, they become unbearable.
4. The long-term harm of AI-generated faces: A “great aesthetic depression”
The real concern is the potential for a cycle of aesthetic degradation. Once these AI faces dominate the internet, they will be used as training data for future models, leading to even more homogenized faces. This could result in an “aesthetic depression” where everyone’s taste becomes uniform, and truly unique, authentic content is overshadowed.
5. A way forward: Moving from “perfect faces” to “authenticity”
There are already some attempts to address this issue:
- Using non-human characters (e.g., fruits or pets) in short series, which are less likely to trigger the uncanny valley effect.
- Developing more detailed AI-generated faces with realistic skin textures and expressions.
The People’s Daily’s call for a return to “authenticity” emphasizes the need for content that reflects reality. Even if not perfect, content with warmth and personality is what users truly desire.
Conclusion
The issue with AI-generated faces is essentially about finding a balance between technological advancement and user experience. While technology can make content more accessible, we must not forget the human element. After all, we use the internet to explore a “real world,” not a collection of artificially created faces. What are your thoughts on AI-generated faces? Feel free to share in the comments section!