4 Ways to Identify AI-Generated Photos: Spotting Artificially Generated Images
Just last month, an image of Pope Francis in a stylish white Balenciaga puffer jacket it quickly went viral, convincing thousands that the Pope was ready for a night on the town. Days ago, another series of photos surfaced on social media, showing the arrest of former President Donald Trump at the hands of riot gear-clad New York City police officers.
There’s just one catch: the images were completely fake.
AI generated photos
These viral sensations are produced by artificial intelligence systems that process the user’s text prompts to create images. While such AI systems — like Halfway through the trip, DALL-E and Stable diffusion — have been available for several years, only recently have the images they spew become convincing enough to fool less savvy observers. And some experts believe we’re moving fast toward a future where it’s virtually impossible to tell the difference between real images and AI-generated fakes.
Read more: AI can now make music from text descriptions
“The systems are already very good,” says James O’Brien, a professor of computer science at the University of California, Berkeley. “Even experts, without using specialized image analysis tools, can’t necessarily find some glaring, obvious signal that proves it’s fake. And they will be fine.”
O’Brien cautions that it’s important not to rely solely on visual clues to determine whether an image is legitimate. For now, though, there are some signs eagle-eyed viewers can watch out for — among other ways to spot potential visual misinformation without getting fooled.
How can you spot fake images created by AI?
Fake photos aren’t exactly a new phenomenon. Photos have been fabricated and manipulated almost as long as photography has existed. A famous portrait of US President Abraham Lincoln, for example, it is actually a composite created by stitching an image of Lincoln’s head with (ironically) an engraving of pro-slavery politician John S. Calhoun. However, AI imaging systems don’t require nearly as much work, capable of quickly creating realistic-looking photos with a simple text prompt.
Today’s AI tools differ in their ability to generate compelling images. AI systems have struggled specifically for digitize human hands, causing crooked limbs that sometimes have too many (or too few) fingers. Spotting these unusual fingers in images may be one way to weed out fakes, says Claire Wardle, co-director of the Information Futures Lab at Brown University’s School of Public Health.
Read more: Your Robotaxi is almost here
“It’s kind of a fun exercise, like when you were a kid and you looked at these image problem puzzles,” she says. “It won’t be long before these images get so good that these things like fingers won’t save us. But for now, there are still some things we need to be careful about.”
1. Watch out for unstable fingers and teeth
Because the data sets that train AI systems tend to only capture pieces of hands, the technology often fails to create realistic human hands. This can result in images with bulging hands, stretched wrists, thin fingers or too many digits – telltale signs that an AI-generated image is fake. IN viral image of Pope Francis, his right hand (and the coffee mug he’s holding) appears crushed and distorted. Teeth can also cause problems.
“Your hands have structure,” says O’Brien. “You have five fingers, not four or six or whatever. The [AI] models have trouble with this type of structure, although newer ones are getting better at it.
Indeed, as technology advances, some AI tools like Midjourney V5 have started cracking the codeat least in some examples.
2. Be careful with overly smooth textures
Some AI image generators produce textures that are too smooth or plastic-looking skin with a shiny sheen. This means that a jacket may appear – say, the Pope’s flapping coat too nice, says O’Brien.
“Instead of looking like a material that has some wrinkles in it, it can come out a little too perfect,” he adds.
3. Look for details that don’t fit
Perhaps the biggest things to watch out for are inconsistencies in the logic of the image itself. Both O’Brien and Wardle refer to a recent series of AI generated images of a great white shark washed up on a beach which also went viral on social media.
“The easiest way to tell that this image is fake is to look closely at the shark,” says O’Brien. “You’ll notice in each photo that the pattern around the eye is different.”
Other inconsistencies, O’Brien continues, include details such as clothing fabric that blends with different objects or background patterns that repeat perfectly. But we may be more likely to miss these details if we want to believe the reality that the false image presents. IN A 2021 study co-authored by O’Brienparticipants were more likely to accept an image’s credibility if it was consistent with that person’s prior beliefs.
4. Do your research
When in doubt, don’t be afraid to compare what you see with other credible sources.
“The things we’re seeing now, we just have to immediately Google to find other information about them,” says Wardle. “Yes, there are little clues you can see in the image, but just think and do your research.”
The future of AI-generated images
O’Brien doesn’t think the moment when we can no longer tell the difference between real photos and AI-generated images is in some distant future. He suspects we could cross that threshold in the next few years.
“That moment may not be as far away as some people would like to think,” says O’Brien. “Because the rate of improvement of these systems is getting faster. […] I think as a society we need to get better at accepting that the things we see – the images – may not reflect reality.”
Regardless, the future of AI image generation will be bold, if nothing else. Scientists around the world have already begun using AI systems to recreate images that people have seen based on their brain scans, according to preprint published on bioRxiv the end of last year.
“There are amazing things coming out of this new technology,” adds Wardle. “But it’s a bit like if a faster car comes along; we just have to drive a little more carefully.
Read more about AI-generated photos : Can AI language models like ChatGPT unlock mysterious ancient texts?
AI-generated photos Source link