Summary

  • ChatGPT’s new o3 and o4-mini models are causing a stir on social media
  • The models can use ChatGPT’s tools and can also perform deep reasoning on visual input, allowing them to guess locations based on photos and images
  • To use the function, the user needs to select the preferred model, upload a photo, and ask the question “where was this photo taken?”, with the most reliable results from the prompt “Scan this photo and tell me exactly where it was taken”
  • The AI doesn’t rely on image metadata but looks for clues inside the photo, such as landmarks, names, and signs
  • It uses cropping, zooming, and inverting colours, as well as searching online for cropped sections to match clues
  • While not always 100% accurate, the AI is good at pinpointing locations from unique objects in the photo;

By Karrar Haider

Original Article