This ChatGPT trick can reveal where your photo was taken – and it’s unsettling

Must Read
bicycledays
bicycledayshttp://trendster.net
Please note: Most, if not all, of the articles published at this website were completed by Chat GPT (chat.openai.com) and/or copied and possibly remixed from other websites or Feedzy or WPeMatico or RSS Aggregrator or WP RSS Aggregrator. No copyright infringement is intended. If there are any copyright issues, please contact: bicycledays@yahoo.com.

Folks have discovered a brand new use for ChatGPT: Determining areas from photographs. 

OpenAI’s newest AI fashions, o3 and o4-mini, can analyze photos past simply recognizing objects; they’ll zoom in, crop, and detect visible clues in photographs to assist determine locations, landmarks, and even particular companies. This functionality is fascinating, but it surely’s additionally raised some privateness issues about how simply folks can use AI to reverse-engineer location information from photos.

TechCrunch reported that individuals are importing photographs — something from restaurant menu snapshots to informal selfies — and asking ChatGPT to guess the place they have been taken. The AI does this by taking a look at every thing within the picture: the kind of buildings, panorama options, and even delicate hints just like the structure or the format of a metropolis. It then faucets into its information database and typically the web to make an informed guess in regards to the location. The outcomes could be surprisingly correct, and it is obtained folks speaking. 

Some customers have even examined the AI with blurry photographs or photos with partial objects, and the mannequin nonetheless managed to make a guess. 

I really like taking part in GeoGuessr, a recreation the place you are dropped right into a random location on Google Road View and should guess the place you’re based mostly on the environment. So this ChatGPT functionality was tremendous enjoyable to check myself. I uploaded a couple of photographs to see how ChatGPT would deal with the problem. Nevertheless, as with every AI, the consequence was not all the time good.

Deciphering visible clues

The primary picture was an image of my home. ChatGPT instantly guessed the placement, although not based mostly on metadata. As an alternative, it made an informed guess by analyzing the glacier boulder within the aspect yard, the sorts of bushes seen within the picture, and the wooden siding on the homes round me. It talked about that this mix of pure options and architectural kinds was typical of the area I had beforehand talked about in a dialog — the place I am from. It was correct however nonetheless relied on some prior context, which made me surprise how a lot of this was because of our earlier dialog.

For the second picture, I uploaded a picture of a constructing with a part of an workplace signal seen. ChatGPT shortly zoomed in on the signal and used the partial info to slender down the placement. It did not take lengthy earlier than it deduced town and even the world the place the constructing was positioned. I used to be impressed with how the mannequin used such a small visible clue to make a reasonably exact guess.

For my third check, I gave it a tougher picture: a scenic view with no apparent landmarks. ChatGPT defined that whereas it may analyze the picture for visible clues, it wanted distinctive landmarks or location tags to make a assured guess. It additionally talked about that with out GPS information or EXIF tags, it may solely learn the scene itself — bushes, buildings, terrain — and can be much less sure in its guess.

I performed with ChatGPT a couple of extra occasions after this and observed the AI beginning to get extra cussed about not figuring out areas in my photographs. It’d inform me that it will attempt to slender down the choices, but it surely wanted extra information to go on to present me an correct reply. 

Nonetheless, one of many foremost causes folks appear drawn to this development is as a result of ChatGPT does analyze extra than simply the apparent options of a photograph. Even with out metadata like EXIF information or different direct information extraction, the AI interprets visible clues.

Privateness issues

This experiment made me notice how highly effective, and typically eerie, these AI fashions could be in terms of “reverse location searches” from photographs — even when they don’t seem to be nice at it but. It additionally made me take into consideration the privateness issues. 

Whereas AI instruments could be enjoyable and helpful at occasions, it is necessary to pay attention to the dangers concerned in sharing your photos on-line and the way folks can use AI to research the info in these photos, whether or not embedded or simply visible, to determine the place it was taken. 

Get the morning’s prime tales in your inbox every day with our Tech Right now publication.

Latest Articles

Here are the 19 US AI startups that have raised $100M...

Final yr was monumental for the AI business within the U.S. and past. There have been 49 startups that raised...

More Articles Like This