The danger of misinformation: how to identify hyper-realistic images created with AI

The development of the artificial intelligence (AI) allowed the ultra-realistic image creation from scratch, from easy-to-use tools that are available to everyone. This favored the propagators of false news (“fake news”), a dangerous and widespread practice in the information age, as well as Internet users who use it to make memes.

In cyberspace you can see a version of the Pope Francisco with a white jacket or a black suit dancing in a bowling alley. TO donald trump detained by police officers or serving time in prison with the typical orange suit of American prisons (according to Hollywood). also to a old man with bloody face after being repressed in a protest against the pension reform promoted by the president of France, Emmanuel Macron.

Google will add an artificial intelligence chat to its search engine

Despite how impressive the veracity of the images is with the naked eye, they were created from scratch with programs that can create an infinite number of images, even of non-existent human beings, based on a database made up of user orders and the combinations that are generated accordingly.

The image of Pope Francis with a white jacket went viral on social networks where a debate will be held regarding its veracity. It was finally conceded that it was created with AI.

The level of realism that the images reflect can sometimes be confusing for some, especially when they are related to sensitive issues for public opinion or issues of partisan politics. This opens a dangerous door to the spread of misinformation.that can only be solved by learning to identify these images.

How to detect images created with AI

There are programs like Halfway, DALL-E either stable diffusion that because of how simple they became popular among Internet users. It is enough to write the characteristics of the desired image in a search engine so that the algorithm generates a new image from millions of images in its databases, pixel by pixel.

Will a robot replace me? These are the 20 jobs most threatened by Artificial Intelligence

For example, the Twitter user elliot higgins I asked the software for an image of DDonald Trump pretending that he is detained by the policesomething that never happened. she did it “while waiting” for the former US president to appear before a judge in New York, where he was technically detained and criminally charged for falsifying business records.

Although the result is realistic, there are some details that indicate that it is not a real image. For example, there may be watermarks that identify the software that created the image, as in the case of DALL-E a multicolored rectangle.

However, these details can be removed by users who want to pretend that the image is real. To verify the image, a user can search for it again in the search bar to retrieve its previous occurrences back to the original source.

Artificial Intelligence 07042023
Source: Twitter @eliothiggins.

The importance of details

Another way to spot fake images is by paying attention to details. AI images for example do not usually adequately reflect reflections or shadows. In general, the funds are generic and tend to look blurry, the grain point of the image is different from reality (it looks more like a movie scene). Also, if there is text in the image, it usually doesn’t make sense.

In the case of Donald Trump, it is seen that what is written on the police caps does not make sense. Another detail is that the tycoon has a baton attached to his pants, as if he were just another policeman.

Artificial Intelligence enters the labor market: What will be the jobs of the future

“You have to find the inconsistencies in the details. Often they are photos that, by eye, are very realistic, but when you look at them more closely, there are often problems. Text is problematic, because the AI ​​can’t render it well. Other A clue are the faces in the background, which are quite poorly done. They are blurry faces, not fully formed,” said Lise Kiennemann, a journalist from the information verification area of ​​the AFP agency.

It is not a minor fact to look at the faces of the people generated by AI. In general, they tend to be disproportionate, such as having ears at different heights or teeth, hair, and even fingers with strange shapes or malformations.

Artificial Intelligence 07042023
Another image of Pope Francis created with AI.

Beyond these details, the dizzying pace of AI progress opens up the possibility that the software will be polished and it will become increasingly difficult to differentiate the images.

The AI ​​still has room for improvement, but at the rate at which it evolves, it may soon become impossible to tell the difference between the images generated by it and the real ones. “We may think that in a few years perhaps, but I rather believe that in a few months, we won’t be able to tell the difference,” explained Guillaume Brossard, a disinformation specialist and founder of the website joke hunter.

The danger of misinformation

The use of artificial intelligence opened up a world of possibilities and threats due to the misuse of images, especially when it is done with political fines. By way of example, the image of an elderly Frenchman injured during the repression of a protest against the pension reform promoted by Macron inflamed Internet users.

Artificial Intelligence 07042023
The false image of a protest in France that inflamed critics of President Emmanuel Macron.

The user took the image to criticize the French president for his actions against what was a “peaceful demonstration”. “I shared (the publication) to show what Macron has done with our democracy,” says the text that accompanies the image.

The tweet caused a stir and indignation until some users began to question the veracity of the image, especially when they detected that the police helmets did not carry an institutional insignia.

The confusion with the images also happened in reverse, with the case of the photo of a young woman detained in Paris that was immediately identified as an AI creation until the author of the photo contributed that it was real. In order for them to believe her, she had to present other images of the arrest taken from another angle.

“We can believe that real images are actually AI and we can believe that AI images are real, so the lines are already very blurred and will become even more blurred in the coming months. But there is one thing that AIs do not know how to do and that they’re not about to know how to do, I think, and it’s playing a scene from various angles and that’s a very good clue,” added Brossard.

Artificial Intelligence: GPT is called the new “cuckoo” of the classroom

On the other hand, there are other tools, such as the application hug facewhich are used to determine the probability that an image is the product of AI, but it is not that precise, so so far the best way to check is by looking for photos taken from other angles.

Added to this is the responsibility of each user to “question all the time” the images they see, and not take them as an absolute truth. Especially those used with political fines or to to provoke a emotion. The latter, meanwhile, is one of the main sources of disinformation. “As soon as an image generates an emotion, it is imperative to ask yourself if it is not potentially adulterated in some way,” Brossard stressed.

Can an Artificial Intelligence stream on Twitch?

AI, an “existential challenge” for humanity

With the rapid improvement of AI, it is not certain that it will be enough to fight the growing influence of fake news. “Today, people just believe what they want to believe. They don’t care if what we show them is true or not and that’s the problem. This is a bit like what Trump theorized about alternate truths and the age of post-truth. We are in the middle of it and we will have to learn to live with it,” lamented the founder of joke hunter

To counteract the threat posed by the malicious use of this tool, it is essential to educate the population. However, the rapid development of artificial intelligence alerts even tech leaders and moguls, such as Elon Musk and Apple co-founder Steve Wozniack, who wrote an open letter calling for a six-month moratorium on AI, which given an “existential challenge” for the humanity.


You may also like