Emma Watson and Chris Evans through the eyes of artificial intelligence

Photo: Midjourney/CzechCrunch

Emma Watson in an AI-generated image by Midjourney

Celebrities have a hard time with privacy. They are constantly under scrutiny and the sound of camera shutters is just noise for many of them. Lenses follow them almost everywhere. However, no one took pictures of the images discussed in this text. We had images of actress Emma Watson and actor Chris Evans created by the Midjourney image generator.

What exactly are image generators? The key to understanding their operation is the English word diffusion – dispersion in Czech. Their principle lies in the ability to recognize the meaning of the connection between the image and the text that describes the images. An artificial intelligence neural network can do this by learning from a very large database of images accompanied by a text description. He explained in detail what was captured on the visual basis.

Today, this service is offered by Elon Musk’s OpenAI company with the DALL·E 2 program, the Midjourney company and the Stability AI company with the Stable Diffusion project. To generate the images in this article, CzechCrunch used Midjourney, which is partially available for free to anyone who uses the Discord communication service. All three tools are currently in beta testing.

The image databases on the basis of which the mentioned programs learn are really massive. DALL·E 2 works with four hundred million images, Midjourney with tens of millions and Stable Diffusion even with two billion images. The result can be a realistic image of Emma Watson aka Hermione Granger from the Harry Potter movies or Chris Evans aka Captain America from the Marvel Avengers.

Photo: Midjourney/CzechCrunch

Hyperrealistic photography of Emma Watson presented by Midjourney.


The program starts with a pattern of random dots and gradually creates the specified work from them. Imagine a sculptor who wants to create a statue of a person. He works on a stone block, which thus acquires increasingly clear outlines, until the work really resembles a human figure.

In the case of image generators, the result can be almost anything. You start the program by typing in English what you want it to create. In the case of actress Watson, we entered: “Hyperrealistic 4K photo of Emma Watson posing in natural light with a happy expression with a center composition.” It was done within a few tens of seconds. Judge the result for yourself.

In addition, you can have Midjourney make a new version of an already created image that is inspired by the old one. DALL·E 2 then allows users to edit the resulting image in any way. So the possibilities are basically endless.


Emma Watson in an AI-generated image by Midjourney

However, all three mentioned programs differ in certain functions. Neither DALL·E 2 nor Midjourney allow you to generate nudity or violence. Stable Diffusion is more benevolent compared to them, although its creators encourage users, among other things, to have only such images generated that “they wouldn’t be ashamed to show their own mother.” Furthermore, DALL·E 2 does not generate a specific personality image for you, Midjourney and Stable Diffusion do.

However, new technology and the ability to generate visual content with almost complete freedom today create an environment in this area where everyone can do whatever they want. But it is not that the authorities are not aware of the rapid rise of machine learning and artificial intelligence and are not preparing for it.

In 2020, the European Union issued an extensive report entitled Artificial intelligence in the audiovisual sector, in which experts highlight potential problems associated with the rise of artificial intelligence, from freedom of expression to the right to privacy. And that brings us in a roundabout way back to celebrities.

If you use the right generator, you can imagine basically any shot of anyone in the world. Is Elon Musk in a rocket headed for Mars? Not a problem. Or a picture of a politician shaking hands with a representative of an enemy state? Coming up. What about a totally made-up scene where a famous person is in a humiliating situation? That is also possible.

This raises a number of questions – does the person on it, the creators of the program or you have the rights to the resulting image? After all, it was you who entered the instruction.

Media and technology law expert Kelsey Farish writes in the aforementioned EU report that, from a legal point of view, it resembles the situation surrounding artificial intelligence “a dimly lit nightclub where intellectual property rights, trademarks, advertising, freedom of expression or consumer protection dance together but are not necessarily in harmony.”

However, Farish adds that any case involving publicity, privacy and an individual’s dignity must be considered separately and will always involve balancing different laws against each other. At the same time, law is not the only sector in which new problems and questions arise in connection with artificial intelligence. Another example is art.

As with any emerging technology, people are emotional. We are amazed and afraid. We are excited and angry. However, image generators based on artificial intelligence – as well as other programs relying on machine learning – are not going away. On the contrary, they will only get better and in the future we will see how we can get along with them.

The article is in Czech

Tags: Emma Watson Chris Evans eyes artificial intelligence

PREV The heat pumps for Europe will be from Pilsen. Panasonic invests 3.5 billion
NEXT The US wants to supply Taiwan with weapons and military services worth more than a billion dollars