top of page
  • Writer's pictureRoopali Salwan

Let the Pixels Colour my Canvas

AI in reviving the dying art of India, do we need to think about copyright rather than bringing them back to life?

Artificial intelligence is fundamentally altering the essence of artistic endeavours. Computers are becoming increasingly important in creative fields including music, architecture, and the fine arts. In fact, a computer is often a modern-day canvas, brush, and musical instrument. Yet we believe that the connection between machines and imagination need to be more comprehensive for AI to be considered an artist. It might be a stretch on our intellect but we might soon have to consider a machine to be a creative genius in itself, rather than just an assistant helping us out. Computational creativity is a new branch of artificial intelligence that's been inspired by this viewpoint. But can a machine be a creative genius? Are Dall-E, MidJourney, Stable Diffusion and others creating art or are they summarizing art?

The controversy is valid, and I believe it is just the beginning of the era of court cases against Generative AI models. While Italy has banned ChatGPT for privacy concerns, I am often surprised that more and more artists and writers are not coming forward to ask for a ban/restriction on these models. A simple prompt like “Starry Night in the Style of van Gogh” will show you what I mean. It will nearly completely replicate Van Gogh’s Starry Night. This is not creation; it is barely even summarization – it is plain old plagiarism!

  Picture on Left: Stable Diffusion. Picture on Right: Vincent van Gogh.

Let us then step out of this controversy for now and consider a dying art form. The phrase "dying art" refers specifically to an artistic technique or talent that is about to become (or has very recently become) obsolete due to the lack of practitioners in the fields. Dying art forms are not necessarily irretrievable in their death. In fact, many "dying arts" are coming back to life as more and more people turn to their cultural heritage for reasons of passion or profit. But what about those art forms which are truly ‘dying’ with very few ‘takers’ who intend to carry it forward? What about languages, which themselves are the receptacles for a culture, that are dying? If it is possible for AI and digitization to help save, at least if not in human memory but in digital memory, these dying arts and languages, should we still be worried about copyrights and privacy?

How can AI help in reviving dying art?

The fact that art institutions have huge collections of photos labelled with information like the theme, period, or location of origination of an image account for a substantial chunk of the appeal of cultural heritage for machine learning. A deep neural network may acquire image features from these labels in a system of so-called "supervised" machine learning. The system can then find these patterns in photos that are not labelled, enabling an automatic categorization process. Visual search is a common tool wherein an object, person, place, or a more abstract aspect of an image (such as colour or texture) is recognized within an unlabelled image collection.

Machine-classification of visual resources is of technical importance to AI researchers, in addition to the convenience of being able to identify relevant images, as the findings offer a very concrete approach to examining how particular methods perform in practice. AI systems can also find hidden patterns, outliers, or connections in image collections that may not have been seen before. Researchers and curators can then observe how existing categorization systems are put to the test when they are applied to previously unclassified items, potentially advantageously extending such classification schemes or, alternatively, exposing their flaws, ambiguities, or prejudices.

Let’s Make it Simple!

To begin with, we focus on some of the art forms from India, some of which maybe classified as dying, while others can be considered as endangered for sure:

  • Mithila art or Madhubani

  • Warli art

  • Dhokra handicraft, Chhattisgarh

  • Rogan Painting, Kutch

The same architecture of the diffusion model was applied to two sets of data: the full dataset with 10 categories and the dataset with 5 different image categories—for the purpose of classifying images. Beyond the pretrained version of the model, a set of 50 Madhubani pictures were provided to the diffusion model to ensure it generates high-quality images.

Fig: Searching for fish figures within paintings using Convolutional Neural Networks (CNNs)

One problem faced by craftsmen and artisans of these dying art forms is that they are unaware of modern techniques used to bring these art forms to life. In short, how to make the young generation and children aware of these art forms as they are the future and last hope of these artisans. And one way of doing it is to ensure that the patterns that emerge from these are stored, may be reused and in some cases potentially utilized for more modern or different subjects. Art forms do evolve, and the latter might help these dying art forms to evolve and capture the imagination of an younger generation.    


Difference between the same art form created by human and AI

Fig: Same art form sample created by human and AI

Though artisans say it is tough to recognize because machines are able to follow the patterns exactly and properly, which humans find difficult to follow, they aren’t concerned about who is making it at all; rather, they are concerned about saving it for future generations. The fish above created by a human artist (left) and an AI (right) in the same style, reflects the ability of the AI to classify and create, and recognize the relevant patterns.

Following the same thread of pattern recognition and modernization of the subject of the art created by AI using traditional patterns, we asked our pretrained prompt engineered model to create van Gogh’s Starry Night in the style of Madhubani painters. The output, not close to perfect, did seem fairly exciting in that it picked up the right subject matter and the somewhat the patterns of the Mithila art.

The AI art transformation is also bringing to light ethical and legal concerns related to the idea of a relationship between originality and authorship. Now we need to think about new concerns, like whether AI-generated works are eligible for copyright protection. But Questions arises that whether we should worry about copyright or we should focus on saving and uplifting dying art. The debate is still going on, but it's high time we started focusing on the rich cultural heritage and tried to save it for our future generations so that they could witness the greatness of their country through its art and history.


Thanks for subscribing!

bottom of page