Chapter 1: An Intriguing Discovery

I stumbled upon a remarkably unsettling website known as ThisPersonDoesNotExist.com. When you visit, a human face appears, accompanied by the striking declaration, “This person does not exist!!!” The images you see are entirely generated by Artificial Intelligence (AI)! Each time you refresh the page, a different, unique face is displayed, as the algorithm draws from a vast image database to create portraits of non-existent individuals. It employs Generative Adversarial Networks (GAN), which consist of two neural networks: one that produces an image and another that evaluates its realism. The GAN training process continues until the generator successfully tricks the discriminator. The faces created are indeed imaGANary... (apologies for the corny pun).

This innovation was developed by Philip Wang, a software engineer at Uber, utilizing technology from Nvidia back in 2019. For more insights, you can check out additional information on sites like Verge, along with a great video here.

The results are both impressive and unnerving. While most images are nearly flawless, with only occasional odd backgrounds, the facial features often appear convincing. However, the generator can sometimes produce bizarre errors that the evaluative neural net fails to catch. Here are a few examples:

A woman with an unusual hole in her head.

A lady with glasses that only appear on one side.

...and the occasional mix-ups with earrings scattered throughout the generated images.

Or sunglasses — “Should I wear them or not? I can’t decide...”

It seems the “judge GAN” might have taken a break while these two were generated...

The results are simply astonishing for these two images. While the technology is impressive when it functions correctly, its failures can lead to more fundamental mistakes.

I find this technology intriguing, yet I struggle to identify practical applications for it. I'm curious about your thoughts: What potential uses do you envision for this technology? What benefits could it offer? Please share your insights in the comments section!

Also, I previously wrote about comparing various text-to-image algorithms and experimented with some unconventional words, yielding intriguing results.

Chapter 2: The Uniqueness of AI

Not all AI is created equal. Each AI system is distinct, just like individuals. Explore the limitations of text-to-image generators and compare their capabilities.

The first video titled "AI does not exist but it will ruin everything anyway" delves into the implications of AI and its potential consequences.

The second video, "Explained: The conspiracy to make AI seem harder than it is!" by Gustav Söderström, sheds light on misconceptions surrounding AI technology.

If you're interested in exploring more articles like this, consider signing up for Medium, where you can find similar content not just from me but from other writers as well!