Throughout the years, AI has managed to infiltrate all areas of life. But can it make a significant difference in creating or reconstructing paintings and thus contributing to the world of art? 

In 2015, a team at Google tested neural networks on their ability to create images independently. Then the AI networks were trained using a database of a large number of different pictures. However, when the machine was "asked" to depict something on its own, it turned out that it interpreted the world around us in an unusual way.  

Organizations like Microsoft use AI to protect cultural heritage. In early March 2019, Microsoft announced an artwork-based image generation project. Developers used a deep neural network microservice architecture, Azure services, and BLOB object storage to create the service. According to Rohini Srivastava, National Technology Officer at Microsoft India, "We don't realize that many pictures, artefacts, sculptures, or cultural remnants can be both preserved and enhanced digitally. We can also piece together components in museums worldwide using AI and gain many interesting insights, especially from ancient scripts".

In early 2016, the painting "The Next Rembrandt" was created. The researchers of the project analyzed about 350 paintings of great artists over the course of their work. They used 3D scanners, which allowed the neural network to capture even the smallest details of each work and copy the style of Rembrandt's painting.

AI techniques 

ML algorithms help to solve both routine computational tasks and non-trivial creative tasks, helping artists handle the "fear of a white sheet". In combination with the artist's creativity, ML makes it possible to obtain interesting results. Many artists working with neural networks finds their unique approach and develop a recognizable personal style. 

1. Neural Style Transfer: It is the most accessible and popular form of AI used in art. The model is based on image stylization and CNNs. It is embedded in mobile applications like Prisma and DeepArt. The network includes two images, a styled template and an original. The system optimizes the parameters, ensuring that the template transformation and the original results are as close as possible in the intermediate CNN layers.  
2. Generative Adversarial Network (GAN): Most artists use this algorithm when first trying AI techniques in their works. GAN subdivides into two neural networks- one that generates pseudo-random images from a given set of distributions and the other determines the plausibility of the image based on the training set. 
3. Non-Fungible Tokens: NFTs can sell almost any virtual object-image, music, text, or 3D models. But objects of digital art are the most commonly discussed topic.  

Future with AI 

It is hard to deny the growing influence of AI on art development. It helps to restore fragments of historically important paintings that have been lost over time and create new art that keeps up with the digital era.  

Some artists and researchers argue whether it is acceptable to consider new AI-made paintings as art pieces. However, in the vast majority of cases, AI is a tool controlled by people. Nevertheless, modern artists are learning the possibilities of neural networks and using AI to make memorable statements.