Beauty is in the AI of the beholder

We are living in a time unlike any other.

Two nights ago, I laid down next to my 20 year old son, home for college, and shared some AI-generated images with him. It is funny how different our reactions are. Me, having a long background in tech stretching back to low-resolution monochrome screens, cannot help but marvel at what I can now create with a simple text prompt. And my son, a sharp-eyed and sharp-minded prodigy, who instantly sees past the marvel and scrutinizes every image.

MidJourney: 20-year old young man and his 48-year old father lie next to each other in a queen size bed, together staring peering at a phone. The lighting is dim, most light is ambient with the main illumination on their faces being from the phone screen — s 500 — v 6.0

Where I see wonder, he immediately picks out flaws. At first, I saw this as negativity and pessimism, but he quickly explained that he is ever training his eyes to be able to identify what is real from what is not.

Wow. Have I been looking at these images all wrong?

One thing I keep noticing, though, is how if you tell most models to create a person, there is a strong tendency towards striking beauty. These models, it seems, are trained with, well, models.

MidJourney: happy young couple walk through a park holding hands

Seldom do you get an odd-looking person. A large birthmark (without prompting for it). And any asymmetry is seen as a flaw in the AI model.

So, this sparked a thought. If my son is training himself (and presumably other younger humans are as well) to recognize AI-generated imagery, will they naturally develop a sense that anything too perfect or too symmetric isn’t “real”?

Extending that thought, will his generation develop a slight aversion for classic definitions of beauty from this association?

How will his generation, and generations to come, define beauty differently?