What is Artificial Intelligence (AI) Art?

Artificial intelligence (AI) art is a new form of art only surfacing in recent years, with the first-ever AI artwork sold in 2018. AI artworks utilise various algorithms and techniques based on artificial intelligence. In essence, any form of artworks that is not drawn purely by humans and rely on an algorithmic system can be considered a form of AI art. AI Art falls under the umbrella of generative art, art which has been created with the use of an autonomous system.

History of AI Art

AI Art was said to have surfaced in 2014, where Ian Goodfellow discovered the Generative Adversarial Network (GAN). Goodfellow’s friends sought his help regarding their project: a system that could create its own photos. Initially, Goodfellow and his friends planned on breaking down photographs to aid machines in creating images by themselves using complex statistical analysis. However, after some pondering, Goodfellow told them that it simply was not going to work.

While that was infeasible, Goodfellow concocted an idea whereby he would pit two neural networks against each other instead. This gave rise to what we all know as the Generative Adversarial Network (GAN). Goodfellow is now widely known as the GANfather of AI Art.

Over the past few years after the creation of GANs, there has been significant progress in the area of AI art, or AI generated artworks. Improvements in the field of deep learning, whereby the neural networks used simulates that of a human mind, have made it possible to create ultra-realistic forms of art indistinguishable from a human-painted one.

How is AI Art Created?

Since the discovery of GANs by Ian Goodfellow, it has been used in many forms of deep learning such as image creation, image-to-image translation, and photograph editing. AI art is one of the applications of GANs.

Firstly, a training set of images is created. For example, a collection of abstract art pieces if you want to create abstract art. These images are fed into the GAN. GANs consist of 2 networks, a generator and a discriminator.

They are both involved in a competition whereby the generator tries to fool the discriminator by generating artworks similar to those in the training set, and the discriminator tries to identify which images are ‘real’ and ‘fake’. For example, the generator creates images of human faces after learning from the training set. Afterwards, images of real and fake human faces are fed to the discriminator, and it is up to it to identify real human faces from fakes.

This process repeats iteratively, over and over, with the generator trying to fool the discriminator until its generated artworks become indistinguishable from genuine artworks. The GAN is now completed and is able to create AI generated artworks.

Advantages of GAN Artworks

Using GANs to generate AI artworks allows for the creation of unique forms of artworks never seen before. It ensures the creation of a unique, one-of-a-kind artwork similar to a human-painted one. Additionally, once the GAN is trained, it is able to create many different artworks faster than a human.

Disadvantages of GAN Artworks

Currently limitations in technology has led to the actual artwork becoming lacklustre and of poor quality. Many artworks produced from GANs look warped and unrealistic. Due to inherent limitations in deep learning and machine learning, an easy-to-fool discriminator would lead to unrealistic outputs. The artworks produced by the GANs may then look distorted, such as a dog with 3 heads or a half-head-half-hand human. Regardless, there appears to be some sort of novelty and uniqueness to these.

Implications of GANs

While the discovery of GANs in deep learning has accelerated the development of AI, many people have used GANs for nefarious purposes. A machine capable of creating indistinguishable fakes could potentially have disastrous consequences. The creation of fake news, deep fakes where they put pictures and videos of people’s faces onto porn stars, as well as words into the mouth of politicians. Extortion of money from deep fakes. A video of a president declaring war on another country. The consequences are unimaginable.

Conclusion

AI Art is a newly discovered form of artwork, and there appears to be much potential in its development and future. However, there appears to be difficulty training realistic GAN models as it requires 2 networks which makes it unstable. Who knows what the future holds for AI Art.

Leave a comment

Please note, comments must be approved before they are published