AI‑generated images blur lines between fiction and reality

When AI image and video generators became widely used, their creations often became the subject of jokes and ridicule. Indeed, the initial attempts looked like a drunk robot's dream that didn’t quite understand what a human should look like. Strange, small teeth that were usually too many, or the wrong number of fingers, and sometimes even extra hands or legs, were just some of the flaws in early AI-generated graphics.

A famous video purportedly showing actor Will Smith eating pasta became well-known. It looked like something we might see in a horror movie. However, a year later, AI no longer has any problem generating reasonably realistic-looking people who are eating:

How quickly artificial intelligence develops is best seen in the images that recently went viral online. Until now, even if a person was perfectly recreated, minor details gave away the AI, such as mistakes in the background, clothing, or eyes. Images generated using Lora AI software amazed observers because, at first glance, it's virtually impossible to distinguish them from photographs. Even on closer analysis, it’s hard to pinpoint anything that indicates they were not taken and retouched by a photographer:

Following the images came the videos, and here, the limitations of artificial intelligence are more visible. It looks very realistic, but the facial expressions still have something artificial about them. But if someone added voice and subtitles to the video, would it be easy to immediately recognize that AI-generated clip on a small smartphone screen? Not.

Artificial intelligence only took a few months to cross a certain Rubicon of realism, and there is probably no turning back. Observing this development, two feelings accompany me: on the one hand, amazement and admiration for the creators because the speed of development of these tools is incredible. On the other hand, threats, because we cannot ensure the ethical use of AI.

Thousands of social media profiles already use generated graphics to gain popularity and often spread disinformation. The same goes for videos. The more realistic the effects, the easier it will be to manipulate public opinion, especially in politics. Well-known people who influence public opinion, like Elon Musk or some politicians, spread fake AI-generated videos.

They often don't label them, so people who are less informed or don't have time to verify them may believe the presented content and form an opinion. Elon Musk was sharply criticized for sharing a video pretending to be a campaign ad for US Democratic presidential candidate Kamala Harris.

Just on Thursday, Donald Trump posted a video on his profile on the X platform claiming he and Elon Musk were dancing to a Bee Gees song. It is clearly an unimpressive work generated by AI, but at this pace of development, we will presumably see the same thing in half a year, only looking entirely realistic.

It should be emphasized that some of these tools are genuinely accessible to anyone. Yes, creating a very realistic image, like those presented in a few paragraphs above, still requires some effort. One needs to take care of the details. Sometimes, it is also necessary to fine-tune what the AI generates. But a few days ago, Elon Musk's X platform released version 2.0 of its AI, called Grok. Using it, I generated the images below - it took me literally 3 minutes. There was no need to try hard; I directed Grok to create a "realistic photograph."

At first glance, it isn't easy to distinguish them from stock photos, especially the first image. In my opinion, the only detail that allows identification of AI involvement is the man's not-quite-successful thumb – something is off. However, the differences are not noticeable as long as no one looks closer. In the second case, it's a bit easier because while the graphic looks realistic, there are at least odd features in the details, such as a tunnel leading to nowhere in the foreground.

This, however, shows how quickly the boundary of realism is shifting in the case of AI-generated images, moving or still. There are millions of scenarios where someone uses fake, AI-created videos to attack a political rival or business competitor. Much has also been said about so-called "revenge porn," or pornographic films where real faces of existing people are embedded. The cheaper and better the technology, the easier it is for anyone to take revenge, threaten, or destroy.

Can we defend ourselves against this? Yes, in one essential way: don’t believe everything we see online, especially on social media. Whenever possible, try to verify and seek the source, especially in the case of content depicting politicians or situations with a political context. It is precisely on social media that people seeking applause or manipulating public opinion reap the greatest harvest. That is where disinformation spreads most effectively, as various videos and photos spread on Facebook, Instagram, or TikTok at lightning speed.  

The European Union recently adopted the AI Act, which is intended to limit the dangers arising from artificial intelligence. However, I don't think they can effectively protect us from the spread of "fakes" generated using AI. The problem concerns not only images but also sound, and citizens may now encounter scams where someone uses AI to generate statements that sound like the voices of our loved ones. Regulations cannot keep pace with the speed of AI development and will be very challenging to enforce.

I must emphasize that I am not an opponent of technological development and do not believe the world will collapse because of artificial intelligence. Once, people feared electricity, then the internet; today, these fears can be attributed to AI. As with many technologies, we must educate and hope that we learn to use this technology responsibly and with a sense of detachment.

For now, we must realize how rapid the progress is. And we must be mentally prepared that in the near future, we will be flooded with images and videos that are not real. Their sole purpose will be to deceive and manipulate us, push us into the arms of this or that politician, or extort money or data from us. This is a threat we must learn to live with.