OpenAI's latest ChatGPT models, o3 and o4-mini, have enhanced capabilities to identify the locations of photos with high accuracy, leading to a viral social media trend. Users have been uploading images and challenging the AI to guess where they were taken, often pinpointing specific addresses or landmarks based on subtle visual clues such as architecture and landscape features. This functionality has been likened to the online game GeoGuessr. The trend has gained popularity on platforms like X and YouTube, where users showcase ChatGPT's ability to analyze elements within photos, including partial signs, to make educated guesses about locations. This has raised concerns about privacy and the potential for doxxing, as the AI's reverse-engineering of location data from images could be exploited for malicious purposes. OpenAI has acknowledged these privacy concerns and is working to train its models to refuse requests for private or sensitive information. The company also monitors for and takes action against policy abuses to mitigate the risks associated with this new feature.
How To Prompt The New ChatGPT, According To OpenAI https://t.co/g4A5MzFR4l https://t.co/4qdmmG1N5D
🚨 EXCLUSIVE: OPENAI’S FIRST BACKER—PERSUASIVE AI ISN’T COMING, IT’S ALREADY HERE AI already knows how to hijack your attention. Soon, it could hack your decisions, beliefs, even your values. @vkhosla: “What I fear most isn’t AI on the battlefield—it’s persuasive AI used on https://t.co/h3n1MxTBiM https://t.co/RnuW2yyafu
🚨 EXCLUSIVE: OPENAI’S FIRST BACKER—AI IS ALREADY HACKING YOUR MIND From ad targeting to influence operations, the line between manipulation and control is vanishing fast. @vkhosla: “That doesn’t even take sophisticated AI. Facebook’s algorithms can do it. TikTok? I believe it https://t.co/PoULJqvB2R https://t.co/RnuW2yyI52