o3 and o4-mini are wildly good at location guessing from the most mundane of photos

What to make of this? #
First, this is really fun. Watching the model’s thought process as it churns through the photo, pans and zooms and discusses different theories about where it could be is wildly entertaining. It’s like living in an episode of CSI.

It’s also deeply dystopian. Technology can identify locations from photographs now. It’s vitally important that people understand how easy this is—if you have any reason at all to be concerned about your safety, you need to know that any photo you share—even a photo as bland as my example above—could be used to identify your location.

As is frequently the case with modern AI, the fact that this technology is openly available to almost anyone has negative and positive implications. As with image generation, it’s important that people can see what this stuff can do first hand. Seeing this in action is a visceral lesson in what’s now possible.

Watching o3 guess a photo’s location is surreal, dystopian and wildly entertaining

Simon Willison details their exploration of the "guess the location" workflow with o3 and o4-mini. Currently this feature is still behind a $20/month paywall for ChatGPT+. What's neat about o3 and o4-mini is its ability to run python code as part of its thinking. You should really check out the transcript to see how wild it really is.

That said, this feels like a workflow that it's learned in consuming the internet. As someone that works in fraud, I've seen people do this kind of understanding and reasoning first hand.

There's a very real case here that posting photos online (of anything) without stripping EXIF data is a dangerous endeavor. I hope that iOS, Android, Chrome OS and mac OS focus on making a default toggle to setup to remove all EXIF data from photos being shared.