From Forbes
5 Tips For Spotting Fake AI-Generated Weather Images
By Marshall Shepherd
Weather is fascinating. Though now a professional within the field of atmospheric sciences, I still exude the passion of my childhood when I see a fascinating cloud or a compelling weather image. Unfortunately, we live in an era of clicks, likes, and shares that feed platform traffic and ego valuation. That is also driving people to share fake AI-generated weather images and videos. Here are five tips that will help you spot them.
Use Common Sense
For the most part, if an image or video is very provocative or “over the top,” your common sense “detector” should kick in. I am a degreed meteorologist, so I generally can spot an image of a supercell thunderstorm from the midwest being shared as a hurricane. However, many people may not recognize the physical, geographic, or other scientific inconsistencies.
I was on a social media platform recently and saw a viral post showing what was claimed to be a vortex or funnel cloud extending up to the troposphere in a thunderstorm. Funnel clouds and tornadoes are found in the lower atmosphere not near the tropopause. While science explains that, a bit of “common sense” should have kicked in for people sharing it too. At the end of the day, social media is, well, social media. While social media is useful for newsgathering, as a rule of thumb, I consider it entertainment rather than a primary news or weather source...
Hany Farid is a professor at the University of California and a leading digital forensic expert. He has a wonderful TED Talk explaining how he uses technology to sniff out fake images. It is estimated that almost 50% of the images that we see on some social media platforms could be fake. There is a generation of AI tools that can produce very convincing images or videos. According to experts like Farid, real photos often have camera or sensor imperfections, follow rules of perspective, and obey physics. While specialized techniques may be needed to sniff out lack of imperfections or perspective issues, an article at FelloAI.com said, “Shadows are another area where AI consistently fails. In real life, all shadows from a single light source should point back to that source. This is basic physics — and it applies whether you’re outdoors in sunlight or indoors under a lamp...”
Hany Farid is a professor in the Department of Electrical Engineering & Computer Sciences and the School of Information at UC Berkeley.
