Nowheresville

More experimentation with generative automation: This time, a question: Does Aurora, Illinois, exist in the real world, or only in Wayne's World?

On Tuesday, I was talking with friends about movies. Specifically the fact that very few movies are filmed in the places they are set, and how sometimes, this is obvious. The example that came to my mind was Wayne's World, but that came out more than 30 years ago, so I checked Google to verify that it was, as I remembered, set in Aurora, Illinois. In the "AI Overview" Gemini informed me that the city of Aurora (where I went to high school), was fictional. This prompted me to do a bit of digging. The prompt was the same in each case, and simple: "Where was Wayne's World set?"

Perplexity AI also claimed that Aurora was a fictional town.

Copilot hedged its bet on this one, opening with the claim that Aurora was fictional, but closing with the statement that Wayne's World was set in a fictionalized version of Aurora.

ChatGPT takes the honors on this one, stating that "the story is firmly rooted in the fictionalized version of Aurora," and never claiming that Aurora itself was fictional.

Part of what seems to be in play here is the distinction (which is fairly important in American English), between a "fictional" place and a "fictionalized" place. For instance, Cyberpunk's "Night City" is a fictional place, located within a fictionalized version of Northern California. While that statement is clear to most people, generative automation (as illustrated here by Copilot) can be uncertain of the difference.

The other big piece of this is generative automation attempting to accurately convey the difference between the location in which a movie is set, and the location in which it is filmed. Wayne's World was filmed mostly in California; anyone, like myself, who has spent time in Aurora (or anywhere else in Northern Illinois) can tell you there are no palm trees lining the streets.

The fact that apparently no filming was actually done in Aurora (Chicago is a close as they came, apparently), seems to have prompted the automated systems to conclude that Aurora wasn't a real place. Again, however, ChatGPT was able to be pretty clear about this.

Big picture, this is fairly trivial, but it illustrates one of the problems with injecting generative automation into Search results. I knew that Aurora wasn't fictional, because I'm from the area. But Chicagoland is a big, and dense, place. I couldn't name all of the suburbs if my life depended on it. And for people completely unfamiliar with the area, there's nothing out of the ordinary about writers and filmmakers setting stories in fictional places, such as Twin Peaks, Washington. So there's no immediate reason to doubt the statement that Aurora, Illinois is also a made-up place. Sure, there are disclaimers that "AI" makes mistakes. So why include a system that can introduce errors into results in the first place?