Software developer Christopher Schmidt said he was impressed that the photos on the website This Person Does Not Exist weren't images of real people.

Built by software engineer Philip Wang, the website loads a new, random image of a person's face every time visitors refresh the homepage.

Though the images appear convincing, they're not actually photographs of real people.

Instead, the website uses an AI algorithm called StyleGAN, which stands for "style generative adversarial networks." Designed by U.S. tech company Nvidia, the algorithm randomly selects traits from a database of millions of photos of real people to synthesize images of people who don't exist.

"I saw This Person Does Not Exist, and I was like, 'Wow, these look really real. That's impressive. I didn't know computers could do that,'" explained Schmidt.

Inspired by Wang and This Person Does Not Exist, Schmidt decided to build his own website, This Rental Does Not Exist.

He was able to build his website using a publicly available database of bedroom photos compiled by researchers at Nvidia.

Schmidt also used a rudimentary text generator AI made by Google to create random descriptive text to accompany each listing.

Concerns about the public domain

Much like the GAN algorithms that power deepfake videos of celebrity faces spliced onto porn actor bodies, or of politicians appearing to say something they never said, StyleGAN has raised concerns about the development of deceptive or misleading content.

Carl Bergstrom and Jevin West teach a course called Calling Bullsh-t at the University of Washington, and created the Which Face Is Real? website. They're also worried about Nvidia's decision to make its StyleGAN research publicly available.

"In some ways, it's scary that anyone in their basement or any large company can create these kinds of images with this technology," West said in an interview with Spark host Nora Young.

Still, West was quick to point out that developers could also use StyleGAN for positive purposes.

"It makes me a little more hopeful when you do have this technology accessible to all those that want to, let's say, do good for society."

New conversations for a post-AI world

While experts like Bergstrom and West are concerned about the potentially harmful uses for AI that can mimic humans, others are keeping an open mind about where the technology may lead.

OpenAI co-founder Greg Brockman believes it's important for members of society to be aware of technologies like StyleGAN, in order to begin having conversations about artificial intelligence sooner rather than later.

"I think that knowing what to do... is actually quite hard, and there's a really good case for the best way of getting to a good outcome is for everyone to know about this stuff, everyone to be thinking about all the different ways that this will be used and abused down the road many years from now," said Brockman.

Brockman also said that he doesn't want AI research to turn into an arms race between different countries trying to build more powerful algorithms.

"You can imagine if you're in that race, the first thing you're going to throw out is safety," Brockman said.

"So I think that the first thing we hope for is for what we refer to as coordinated competition," he said.

"It's very clear that countries are going to be competing on applications and trying to push forth this technology, but when it comes to safety, that's something we can all coordinate on."

As for Christopher Schmidt, he believes that websites like This Rental Does Not Exist and This Person Does Not Exist are valuable because they bring attention to developments in AI research, while forcing humans to question the content they consume online.

"Don't assume just because there's a picture out there, it's real," said Schmidt.

"Because anything that [Philip Wang] or I could hack up in a day or a weekend or whatever is going to be far below the capabilities of a dedicated bad actor."

This article originally appeared in CBC Radio