The portrait that Rembrandt never painted

and How Artificial Intelligence (AI) created a new masterpiece

The great painter Rembrandt died on October 4, 1669 in the city of Amsterdam, and almost 400 years later he has returned to paint a new masterpiece.

It is a portrait that, as it could not be otherwise, has been created by a Generative AI (GenAI) "trained" with 346 paintings by Rembrandt, which allowed the system to identify the style (the patterns) of the painter to create a new work with the same technique and appearance.

The artwork took almost 18 months to complete, consists of 148 billion pixels and for greater realism, it was 3D printed.

?

The process

As with any AI system, the first step was to obtain the data needed for the AI to "learn" the painter's style. In this case, this phase was simple as it was a matter of compiling the reproductions of all the painter's works.

A simple initial statistical analysis indicated that, if the painter were alive, the new work would probably be a portrait of a Caucasian man, between 30 and 40 years old, with facial hair, dressed in dark, with a white cloth necklace and looking to the right.

The AI algorithm then studied the 346 works to identify the geometric patterns that the painter followed in his portraits. Common features were found in the eyes, noses, or mouths he painted, as well as in the geometric proportions of faces.

And in a similar way, the system discovered the most common volumetric and light compositions used by the master.

Finally, to bring the painting to life, a specially designed 3D printer was used to make high-quality reproductions of existing artworks. In the end, it took printing 13 layers of ink, one on top of the other, to create a realistic paint texture.

?

But who should be considered the author of the work, the owner, and the beneficiary of its exploitation rights?

All countries establish, in their respective Copyright Law, the Author is the person (natural or legal) who makes the original creation.

The difficulty begins when the new Generative AI is part of the creative process, since then it is not at all clear who really brings creativity to the work: the person or the AI.

?

To solve it, we have two approaches to this problem:

·???????? On the one hand, we can consider that the intellectual property of the work is held by the person(s) who has developed the AI algorithm. We are referring to the programmers or ultimately to the company that owns the AI system. Because it's the AI that is really creative.

·???????? Or, on the contrary, we can consider that the person who provides the creativity is the person who gives the instructions or uses the AI system to achieve the desired result. The copyright should belong to the user who defines the prompts.

?

Neither option is foolproof.

In the first case, it would be like accepting that the creator of the tool is the author of any creation in which the tool is used. Is Microsoft the intellectual author of all the books written with MsWord, or is Canon the intellectual author of all the photographs taken with its cameras? (In this second case, the doubts are more evident because current cameras have many automatic options that allow the final result to be improved without the participation of the photographer.) (1)

But the other option isn't perfect either.

If we recognize as the author the one who uses the “tool”, in the case of using Generative AI it may seem that the one who provides the creativity is the one who writes the “prompt”. Although with a simple experiment we can dismantle this theory.

If 10 people write the same prompt with the same Generative AI tool, we will get 10 different results (similar, but not identical), so the one who is providing a differentiating creativity factor is the algorithm, not the person who authored the prompt.

Then we go back to the first option where it is the AI tool that should have the copyright, but we have already seen that it is not a perfect option either.

So far, court rulings have established that applying current laws it is not possible to grant copyright to any of the parties, so works created with Generative AI are orphaned of copyright and are free to use by anyone.

Faced with this new paradigm that opens up with Generative AIs, some voices have been raised calling for laws to be adapted to the current reality. Copyright laws were born in the United Kingdom in 1710, and since then each country created its own version, which has been modified over the years.

Laws are not set in stone, and the time has come to update the law of Copyright ... But which way should the balance tip? The debate is open.

?

Notes:

(1)??? Nowadays all cameras include some assistance system, such as autofocus, ISO adjustment, etc ... and lately with AI for much more advanced processing. Where do we draw the line to accept 100% human authorship?

Another related case that became famous a few years ago was with the copyright dispute of the selfie taken by a monkey. Who should have the copyright: the monkey (he took the photo), the owner of the camera (he arranged the camera), or the camera manufacturer? https://en.wikipedia.org/wiki/Monkey_selfie_copyright_dispute

(2)??? The project "The new Rembrandt" has been made possible thanks to the sponsorship of ING, together with several companies J.Walter Thompson and Microsoft, as well as with the advice of the Delft University of Technology (TU Delft), The Mauritshuis and Museum Het Rembrandthuis.

(3)??? Intellectual property protects both copyright and industrial property, although neither the idea (in the case of copyright) nor the procedures, methods of operation or mathematical concepts (in industrial property) are protected. www.ayudaleyprotecciondatos.es

#Art #Arte #AI #ArtificialIntelligence #Copyright #Law #GenAI #Creativity #DerechosdeAutor

?

World Intellectual Property Organization – WIPO 荷兰国际集团

智威汤逊 微软

荷兰代尔夫特理工大学

@The Mauritshuis and Museum Rembrandthuis