Introduction to AI Art


AI Art is a current trend in the world of art that is gaining popularity day-by-day. The concept is based around the idea that machine learning algorithms are capable of producing original images as output, when appropriately trained using large computational resources on a vast amount of image data. These machine learning algorithms are based on the technology called General Adversarial Networks or GANs for short.

This seemingly simple process is actually more complicated than one might think since it is highly dependent on the size and what’s most important – quality of the dataset as well as time spent on training the algorithm. All of these variables, in turn, depend on the power of the graphics cards used (we use a grid of industrial grade GPUs), e.g. – the more computational power, the larger the dataset and the lesser the time it takes for every model to train. In our context, “training the model”, in simple terms, means that the algorithm continuously generates images through the generator component and the other component of the model called the discriminator “decides” whether the output of the generator is close to the original dataset or not. Hopefully, you are not too much bored by the slightly technical terminology.

After countless hours of training, the model, based on what it learnt during the training process, is then able to generate something completely new, something that has never been drawn before – original AI art.

Fascinating, isn’t it?

It is beautiful how technology has advanced to the point where it is able to engage in art creation. Only 15 years ago, this was barely considered a possibility and today AI Art Shop displays a collection of over 50 AI Portraits with human faces so realistic that it is almost impossible to tell the difference between a human-made portrait and an AI portrait. So, are we entering the new era where robots are going to completely replace human artists? The answer is probably no, at least in the near future. AI algorithms, with appropriate training, are indeed able to generate original images, however a lot of the output, even with a lot of training, is either noise or art that believe me you would not like to see. So, this is where human involvement comes in. There needs to be a team of people who understand art and are able to select and classify only the best images. The next step once the paintings have been classified and named, is to use a different group of AI algorithms to increase the resolution of every selected image to allow for the best quality of paintings.

We can see that there is definitely a degree of human involvement, but the creative process is done completely by Artificial Intelligence. As of the time of writing of this article, we offer custom made one-off exclusive AI canvas prints. Our AI paintings are brought to life using a state-of-the-art Giclee Print technology with an Ink Jet machine, guaranteeing 99% colour accuracy as well as outstanding colour depth and longevity. One interesting trait of AI paintings is that on some of them we are able to observe brush strokes and certain 3D elements that the AI overlaid on certain parts of the painting. Coupled with our vector correction and resolution adjustment AI technologies, this allows to create AI paintings with an oil-painting effect. For some of our paintings it is hard to even tell whether the painting is a canvas print or an oil painting when looking from a distance - all because of the ultra high resolution (1000 dpi) and 3D brush stroke effects that are the product of AI imagination.