People have found a new use for chat: detecting places from photos.
The latest AI models of Openai, O3 and O4-Min, can only analyze images beyond recognizing objects; They can explore visual clues in photographs to help you identify locations, sites and even specific businesses, harvest. This ability is attractive, but also some privacy concerns have been raised how people can easily use AI for reverse-engineer location data from images.
Too: Chat has made it easy to find and edit all the AI images that you have ever produced
Tekkachchan Told that people are uploading photos – anything restaurant menu Snapshot for Casual selfie – And ask the chatgipt to guess where they were taken. The AI does this by looking at everything in the image: types of buildings, landscape features, and even subtle signals such as architecture or layout of a city. It then taps to make an educated estimate of its knowledge database and sometimes the location in the Internet. The results can be surprisingly accurate, and it is talking to people.
Some users have also tested with AI Staining photos or picturesAnd the model still managed to make an estimate.
Too: Is the Chatgpt Plus worth your $ 20? Here is described how it compares free and Pro plans
I like to play LandA game where you have been dropped in a random place on the Google Street View and will guess where you are on the atmosphere. So this chat capacity was super fun to test itself. I uploaded some pictures, how to handle the challenge to see. However, with any AI, the result was not always correct.
Explain the visual clue
The first photo was a picture of my house. Chatgpt immediately estimated the location, although not based on metadata. Instead, it made an educated estimate by analyzing the glacier boulder in the side yard, the type of trees seen in the photo and the wood siding on my surrounding houses. It mentioned that this combination of natural characteristics and architectural styles was the specific of the region I mentioned in a conversation – from where I am. It was accurate, but still depended on some pre -reference, making me wonder how much it was due to our earlier conversation.
For the second picture, I uploaded an image of a building, which appeared with part of an office sign. Chatgpt quickly zoomed to the sign and used partial information to reduce the location. It did not take a lot of time before the city and even the area was reduced, where the building was located. I was impressed by how the model used such a small visual clue to be estimated quite accurately.
Too: How to use chat: a starting guide for the most popular AI Chatboat
For my third test, I gave it a more challenging picture: a beautiful view with a clear scene. Chatgpt reported that when this scene could analyze the image for clues, specific sites or location tags were required to estimate a confidence. It has also been mentioned that without GPS data or exf tag, it can only read the scene – trees, buildings, terrain – and its estimates will be less sure.
I then played with Chatgpt a few more times and saw that AI began to be more stubborn about not identifying places in his photos. This will tell me that it would try to reduce the options, but I needed more data to go to give an accurate answer.
Nevertheless, one of the main reasons is attracted to this trend, as Chatgpt analyzes more than the clear characteristics of a photo. The AI explains the visual clue, even without metadata such as metadata such as EXIF data or other direct data extraction.
Privacy concerns
This experiment made me realize how powerful it is, and sometimes terrible, these AI models can occur when it comes to “reverse location search” from photos – even if they are not great yet. It also thought of me about the concerns of privacy.
Also: Openai’s most impressive step is nothing to do with AI
While AI devices can be fun and useful at times, it is important to know about the risks involved in sharing their images online and people can use AI to analyze data in those images, whether embedded or just visual, whether it was taken to find out where it was taken.
Get top stories of morning with us in your inbox every day Tech Today Newsletter.
