Generate content through artificial intelligence (AI) It struck many with its high level of detail and accuracy. However, in the case of the images, there are still three things to solve to make them more realistic: hands, feet and teeth.

These body parts usually have many inaccuracies in the photos created by these tools and allow us to distinguish the image created by artificial intelligence from real people.

You might also be interested in: Five Cybercrime Practices That Could Be Augmented by Artificial Intelligence

A detail that was essential to distinguish an attempt to spread false information France In which they published false content, generated by artificial intelligence, about some protests and which they tried to pass off as real, but thanks to these errors, such as hands with six fingers, users learned the truth.

A case that creates a warning before repeating the practice, using these new tools to manipulate news or facts. However, at the moment, artificial intelligence is still unable to create hands and teeth.

Why does artificial intelligence have bugs on its hands?

To understand the origin of this failure, it is necessary to understand that all AIs work according to the model of machine learning, thanks to the data they receive and the information they receive from the Internet.

So the generated designs are not a product of imagination but are strictly dependent on the information available to them. For example, the basis ChatGPT It lasts until 2021, what was asked after that date, you don’t know yet.

You may also be interested in: Google integrates its artificial intelligence into Maps, translator and search engine

In the case of tools that create images, e.g Stable Diffusion, DALL-E 2 and Midjourney, Ask it to design a dog in a certain way, it will search its database of dog models it has and auto-train it to give you the final result.

Something special happens at a specific point on the hands and feet. “It is generally understood that within AI datasets, hands appear less prominently than faces in human images. The hands are also smaller in the source images because they are rarely seen in large form. “- the speaker Stability AI For BuzzFeed News.

This means that the information that the AI ​​has is not clear enough and therefore Amelia Winger BearArtist and Associate Professor of AI and Art University of Floridabelieves that this technology does not really understand what the hand is, or at least not as it is anatomically related to the human body.

You may also be interested in: ChatGPT recommends love letters and ideas for Valentine’s Day

In the information that feeds these technologies, the hands are always holding something or putting it on something, so it is not visible in its entirety, but only the fingers or part of them. AI senses to do the same in your content and therefore messy results.

The solution to this would be to capture this technology in which the limbs are free and open as a reference, so that it has enough information to replicate them, to understand how they relate to the body and its entire composition, however simple. It seems to us.

A point that, if resolved, would make it difficult for us to distinguish between a real image and an AI-generated image, with all the good and bad that brings.