Briefly, news has carried out an experiment in which we have tested whether people can distinguish images generated by artificial intelligence (AI) from real photographs. You can take the same test further down in the article. Ten participants were presented with four pictures of different motifs. Only one of the pictures was a real photograph. In the experiment, only 17.5 percent of the answers were correct. Media researcher Arild Fetveit at the University of Kristiania says that the computer-generated image is now so much like a real photograph that it represents a watershed. Some of the participants, who study technology, are positive about the technology and saw no problem with many missing the right answer. The summary is made by an AI service from OpenAi. The content is quality assured by news’s journalists before publication. The corridors at the University of Porsgrunn are teeming with students. Many are cramming for the last exams before Christmas. Today, the students will also get to brush up on another task. An experiment that news has brought to campus. news shows four pictures to the students. One of the pictures is a real photograph. The other three are generated with artificial intelligence. For example, are the students able to see which one of these pictures is a real dog? Swipe right to see the four pictures of dogs. Is this real? Or maybe this one? You can find the conclusion at the bottom of the case! news tests ten different people who try to guess four different motives. They are presented with images of dogs, beaches, monuments and women. You will find the same test further down in the article. If you are impatient, you can click here. Around a table in the canteen sits Angelica Salomone. She only manages to guess one correctly when she tries herself on news’s test. – It is very scary, says Salomone. – why is that? – Because you don’t know what to believe anymore. In news’s experiment, only 17.5 percent of the answers are correct. This is how news has worked with the case In this case, news has generated fake images with artificial intelligence. I have used Adobe’s software Firefly. In the program, write a simple text that describes what you want the program to generate an image of. The photograph I have used in the test, which is real, has been taken by news from the image bank Colourbox. They state that all material on their website goes through a quality check. This is how they say that there are no AI-generated photographs in their bank. The company states that the check is a manual process, and that errors can occur. But only for a limited period until the AI-generated content is discovered and removed. – A watershed The fact that people guess wrong is met with great understanding by media researcher Arild Fetveit at the University of Kristiania. – Computer-generated images can now be so similar to photographs that they represent a watershed, says Fetveit. Fetveit says the picture that news presents to people is almost impossible to distinguish. He points out that the criteria one could previously use to distinguish between AI and photography have gradually become outdated. In the past, you also needed a high level of expertise to create fake images of good quality. In, for example, Adobe Photoshop. With artificial intelligence, you only need to write what you want to picture. – Today’s technology is more accessible, faster and easier to use, says Fetveit. Positive towards technology news shows the photo of a group of boys at the University of Southeast Norway studying technology. They also only manage to hit one of the alternatives correctly. Jardar Fagerlund, on the other hand, is not so frightened by the experiment. – I don’t see any problem with this. As long as it doesn’t affect people, it’s fine, says the technology student. – But it’s easy to deceive people now, isn’t it? – Yes, but one has already been able to deceive people for many years in various ways. Artificial intelligence is primarily a tool, says Fagerlund. Jardar Fagerlund was one of the students who took part in the experiment for news. Photo: Sindre Thoresen Lønnes / news Arild Fetveit at the University of Kristiania says AI-generated images mean new challenges for most people when they have to navigate the media landscape. The researcher believes that computer-generated images, which cannot be separated from photographs, can weaken trust in the photograph. – What I experience is that I can trust the technology itself to a lesser extent. Thus, I have to rely to a greater extent on those who use it. So trust will be more firmly rooted in the institutions that use the material and present it, says the researcher. The situation will also place greater demands on media houses and content creators. In this article, Faktisk.no comes up with tips on how you can reveal fake photos. Here is a short summary: How to reveal a fake picture? Check if the details are consistent If there are several pictures of the same event, you can see if you find differences that don’t make sense. Look for classic AI drawings Although the technology for generating images has come a long way, it is still not perfect. The contrast in the images can look strange, and they can have a cinematic or computer game-like feel. Midjourney 5 manages to make hands better than previous versions, but they can still look weird. Nevertheless, great progress has been made in generative AI in a short time. Therefore, it is not certain that you will be able to distinguish between an AI-generated image and a normal image with the naked eye in the future. Use reverse image search Faktisk.no has written several articles about how you can use reverse image search to find more information about an image. You do this by searching the image itself in a search engine. Check news websites For spectacular events, such as a dramatic arrest of Donald Trump, it may be enough to check traditional media. If you find no trace of the arrest either in Norwegian or major international media, this is a clear indication that the incident is not real. If the incident was real, it would not have taken long before it was picked up and received massive publicity. Try special tools for AI images There are separate programs that try to analyze whether an image is generated by artificial intelligence or not. Such tools cannot be used uncritically. Source: Faktisk.no Take the test! Are you thinking that people have been stupid enough to find the real picture among the fake ones? See if you’re any better then! In the videos below, you will be presented with four images. Only one of them is a real photograph. Fasit Then there were the dogs. Which one of them is true? The correct answer is the dog marked with a green ring. Pat on the back if you saw it! You are obviously interested in this case since you read it all the way down. Please advise me of other angles I should look at in relation to this case. Please also send me an e-mail if you have any tips for things I should look into.
ttn-69
Artificial intelligence and photography – can you tell the real from the fake? – news Vestfold and Telemark – Local news, TV and radio
