AI art, how is it created?
How? How does a machine create Art? The answers vary. Algorithmic Art has evolved over time. Harold Cohen, an artist and engineer, is known to have been one of the first people to recognise and connect technology and Art. His program AARON in 1973 was created in a way that produced drawings based on the set of rules programmed into it and in its lifetime, remained a program that performed creative tasks under the supervision and direction of the artist. However, as AI has developed, so has autonomy in algorithmic art.
Now, these artists create algorithms that use a large group of images as their data set in order to identify or “learn” a style. It then attempts to generate its own images based on this skill. These range of outputs have to be then looked at and chosen by the Artist. In this, the creation of the new painting/image/art is entirely depended on the program being used but the Artist has the chance to give their insight into the creative process by being able to curate the input data themselves as well as choose their desired output.
Usually, these images created don’t end up perfect, a lot of them create weird images. An AI artist called AICAN for example, has been inputted with thousands of images of Art from the west spanning over several centuries. This system than uses this data to generate a new image. What makes this unique is it’s “faceless portraits”. Faceless portraits are the result of the algorithm’s poor imitation of a face however, this distorted result has formed its own audience. The magnificence of something so random is now growing to be appreciated.
GANs fully known as generative adversarial networks have also gained popularity for use in Art over time. In fact, these have ended up creating some of the most realistic artworks. Here are some created by NoArtist, a collective that generates art through AI:
GANs were first introduced by a computer scientist called Ian goodfellow. This model uses two neural networks . One acts as a generator while the other acts as a discriminator. A generator would create these outputs based on the data set inputted while the discriminator will use the input to identity a real and fake (generated image) with the use of trends and patterns. These two together make it seem like a supervised machine learning process. Lastly feedback is given to the models so that they can improve. This improves the quality of images produced with usage.
You yourself can create AI Art too! Yes, you! There are several websites and softwares that use techniques like this to generate their own artwork. The following link lists some that you can explore on your own.
This sums up the basics of how AI Art is really produced. I hope this was insightful to anyone interested in the combination of these two areas.
Anushri Iyer is a Student Ambassador in the Inspirit AI Student Ambassadors Program. Inspirit AI is a pre-collegiate enrichment program that exposes curious high school students globally to AI through live online classes. Learn more at https://www.inspiritai.com/
