First, let me tell you that I stared at this photo for 15 minutes before I wrote the page you viewed before arriving here.
The explanation is that this is not a person all, but a human likeness created by artificial intelligence (briefly explained below). This girl or woman you believe you are seeing is nobody. She never existed.
The longer I stare at the photo, the more frustrated I become. I want her to be real. How can she never have existed? The thought struck me that she could have been my child had I married a different woman. Whatever my imagination could dream up, the fact is, she is nobody, created in two dimensions by a computer.
Still, I want her to be real. I want to know her. She looks so nice. I want her to be happy, maybe married with a wonderful husband and children. I want her to come to my home for dinner, give me a big hug when we greet, and then we’ll sit on the patio and talk, laugh, and watch the kids play.
I feel very sad she never lived. But that’s dumb! How can seeing her photo, knowing she is no more real than a cartoon character have such a profound effect on me? My eyes tell me that she must be real. She must be!
For me, trying to understand this photo is sort of like trying to understand the concept of endless space. No matter how long I think about it, it can’t conceive of it.
Here are two other people who never existed.
And, now, here’s how these “people” were created. It’s through a technique called GAN or Generative Adversarial Networks. It’s a relatively new concept in machine learning, introduced in 2014. The goal is to synthesize artificial samples, such as images, that are indistinguishable from authentic ones. A common example of a GAN application is to generate artificial face images like you see here by learning from a dataset of celebrity faces. One of the main challenges is controlling their output, i.e. changing specific features such as pose, face shape and hair style in an image of a face.
It’s hard to believe, isn’t it? What do you think? Please leave a comment.