According to a popular tweet, in a certain elementary school in Jasło, “Black Mass Week” was to take place instead of nativity plays, and children sang black metal songs. A lot of Internet users fell for the joke when the post and photo began to circulate on Facebook. Artificial intelligence creates more and more perfect images, but there are still ways to recognize such photos. Here are five basic tips.
“At school No. 63 in Jasło, children from classes 1-3 opted out of religion. This is thanks to Mr. Adam, a music teacher, who instilled in them a love for black metal. Instead of nativity plays, he organized Black Mass Week. The children performed a jointly prepared song ‘I worship You, my sweet Satan’ – he wrote April 19 on Twitter Bartosz Scheuer, lecturer at the Department of Economics and Environmental Management at the Wrocław University of Economics (original spelling of all posts). The title of the song can be translated as “I adore you, my sweet Satan”, and the name of the event can be translated as “Black Mass Week”.
The author attached a photo to the post to confirm the story from the tweet. We see a guitarist dressed in black with long hair and a face painted in white and black. He looks like a member of Kiss or some other black or heavy metal band. A man plays an electric guitar surrounded by children of early school age. They are dressed in black clothes, some of them also have their faces painted. The room resembles a schoolroom; strange ornaments hang behind the group: a deformed doll is leaning against the wall, next to it is a mask that resembles an African totem. The post quickly gained popularity. It has been viewed over 186,000 times. times, it was liked by 2.8 thousand. twitter users, and shared more than 290.
“Weak photomontage” vs. “where are the parents of these children”
After looking at the photo, you can notice some details that cast doubt on the authenticity of the photo. They were also noticed by Twitter users who discussed the post and the photo. “How does a guitarist have 3 hands but no right thumb?”; “There’s nothing like a guitarist without a thumb but with two left hands”; “A music teacher without a thumb, but with a double left hand. Such miracles only in Jasło.” Others also paid attention to the setting of the snare behind the man with the guitar – this one was at the height of an adult’s head. It’s too high to play on.
Many netizens claimed that the photo was not real: “And the proof of this photo generated by AI?”; “Weak photomontage”; “An interesting thread. It shows the dangers of using artificial intelligence to spread disinformation.”
Others pointed out an important detail from the content of the entry: “I don’t agree with one thing. There are not so many schools in Jasło”; “Just type in Google: ‘at school no. 63 in Jasło’ and you already know it’s fake, because there is no such school. Low quality fake.”
A screenshot of the post with a photo circled also on Facebook. There, many commentators wondered about the authenticity of the story from Jasło. From the content of the comments, it appears that some believed in it. “Let it all be f******” – wrote the author of the post on Facebook. “From extreme to extreme. Where parents. Children are looking for someone who will be interested in them, whether it’s a priest or another satanist”; “From one swamp to another… some light years need their soul light to come out of the dark”; “therefore it is known that this is the beginning of their end”; “and where are the parents of these children”; “Horror”; “Well done. The teachers did a great job.”
More attentive Internet users were right. The popular tweet is a made up story, the photo is also not real. They were created by artificial intelligence.
There are not so many schools in Jasło. “The post was highly satirical”
As noted by Internet users, there is no primary school No. 63 in Jasło, which has over 33,000 inhabitants. There are only ten elementary schools in Jasło, according to information on the city’s website.
The fact that the popular tweet and photo is a joke was told to us by its author, Bartosz Scheuer. He confirmed that the Twitter account belongs to him and that the account is “completely private” and that he posts content on it. “The entry was extremely satirical and its entire content clearly suggests it (including the deliberately selected school number, which even without checking it can be guessed that it is fictitious)” – he said in response to our questions.
“The likelihood that someone will take something like this seriously (or even that the reader of a text will read it differently from the author’s intention) is always there. However, in the case of content intended to be satirical, I do not think that the author is responsible for this different reading.” Scheuer added.
“I saw that screenshots are circulating on the Internet, but compared to the large number of previous cases, from the small number of criminal threats and death wishes received this time, I conclude that, in principle, it was read correctly (i.e. as satirical)” – noted Scheuer .
How to recognize a photo created by artificial intelligence?
At the end of March on Konkret24 we wrote about photos generated by artificial intelligence. One showed French President Emmanuel Macron joining protesters against raising the retirement age in France, while another showed former US President Donald Trump struggling with police officers who allegedly wanted to arrest him. We also published material about the photo showing the supposedly “biggest horse in history”except that the photo was created by artificial intelligence.
Some of the paintings looked realistic, but upon closer inspection, you can see elements that raised doubts. Many English-speaking editorial office fact-checking in recent weeks explained how to recognize images generated by artificial intelligence (AI). Here are five, in our opinion, key tips and ways to recognize such images.
1. The probability of the situation from the photo
It is worth starting with the question: is it likely that the situation depicted in the photo took place? So, how likely is it that a “Black Mass Week” was organized in a Polish elementary school instead of a Nativity play? It is not impossible – unusual and surprising situations do happen, but this one seems extremely unlikely.
What’s more, if the photo raises your doubts or strong negative emotions, it’s worth being careful and not trusting the message it carries. It’s worth taking a closer look at it.
2. Visual inconsistencies: AI has a problem with human hands and reflections
Although in recent months more and more advanced tools based on artificial intelligence have appeared – the photos generated by some of them (usually available for free) are full of errors. Looking for them is a good step in recognizing a fabricated image.
Many English speakers journalists and verifiers information, returned note that currently (March and April 2023) image-generating artificial intelligence systems have some problems with generating realistic human hands. And so in the case of the photo that circulated on Twitter and Facebook, Internet users quickly noticed that the guitarist’s right hand was missing a thumb, and the guitar’s neck was held by … two left hands. In addition, another hand appeared behind his left shoulder holding a second guitar fretboard.
Moreover, such photo generators often create asymmetries. And so the faces of people in such photos can be disproportionate, ears can be of different sizes, and the eyesight of the person in the photo can be strangely spaced, as if they were looking in two different directions at the same time. Also, the arrangement of teeth or hair, e.g. their contours blend together, or a strange blurred texture – may prove that the image is not real. Some elements may be poorly integrated into the photo, such as glasses that “blend” into the wearer’s face. For example, in one of the AI-generated photos of Pope Francis, his corrective glasses strangely connect to the eyebrow, and the next ones “flow down” to their own shadow – explained the editors at the end of March American magazine “Time”.
What other elements should you pay attention to? “Currently, AI images also have a big problem with generating reflections,” Vincent Terrasi, co-founder of Draft & Goal, a startup that created an AI detector for universities, told AFP. “A good way to notice the AI is to look for shadows, mirrors, water, in addition to that we can zoom in on the eyes and analyze the pupils, because there is usually a reflection in them when taking a picture. We can also often notice that the eyes are not the same size, sometimes have different colors.”
3. Smoothed image. Pay attention to the background and look for logical errors
Is the texture of the photo smoothed, almost blurred? Faces are almost flawless or look blurry, devoid of detail? This is a common effect achieved in image-generating AI applications. Many AI-generated images feature deformed-looking objects or figures in the background. And so in a popular Twitter photo, the silhouettes of children in the background are very blurred, their faces seem to be deformed.
AI can make logical errors – generate elements in the photo that should not be there, e.g. road sign nonexistent in the country where the photo was supposed to be taken. In the photo from the post in question, we can see that the snare drum on the left is too high to play. Moreover, his stand seems to be “broken” – this is another error in generating the photo. The number of tuners on the headstock also does not match (in most guitars there should be three tuners on each side of the headstock; in the photo there are three on one side and only two on the other).
4. Search for information about the photo in reliable sources. Use image search
Looking at the picture is a starting point, but it is not always enough. Annalisa Verdoliva from the University of Frederic II in Naples in an interview with the editors of AFP Fact Check she warnedthat “generation methods improve over time and show fewer and fewer synthesis artifacts, so I wouldn’t rely on visual cues in the long term.”
Therefore, it is also recommended to carry out the so-called side search to confirm or disprove the authenticity of the material – gives Politifact editorial office. Using your search engine of choice, check to see if other credible sources are reporting the same photo or event. Read what Internet users write about it in other posts. Do they provide evidence of the authenticity of the material, or do they also have doubts?
Just searching for the phrase “school no. 63 in Jasło” in the selected web browser will confirm that such a school does not exist in this city. This already casts doubt on the whole story from the popular tweet.
It’s also a good idea to use an image search in your browser. This way you can check where and when the photo was published. This feature will allow us to complete an internet query and possibly reach the oldest available version of the photo.
5. Check the photo with the detector
Currently, there are free tools available on the web that assess whether a given photo was generated by artificial intelligence. One of them shares Hive Moderation, an American company dealing with artificial intelligence. We used the tool in a photo from a popular tweet. Hive Moderation estimated that the probability of generating photos by artificial intelligence is 99.8 percent.
Using another similar tool Maybe’s AI Art Detector we found out that this probability is 83 percent. These results confirm that the photo was most likely created using artificial intelligence.
Main photo source: twitter.com