The beauty in GAN-generated images
My progress/ concept behind my work:
For collecting my images I used a chrome extension called ‘Image Assistant Batch Image Downloader’ and it allowed me to download a large number of images in one go, and it allowed me to filter out the pixel resolution and frame size. I was able to collect a batch of images that I wanted to use when training my GAN-generated images. For this round, I decided to collect images of landscapes of mountains, skies, and sea. The direction I was going for was nature, and I wanted to see how this would turn out if I combined it with images of invertebrates particularly jellyfishes, therefore, I collected a batch of jellyfishes as well. I always thought jellyfishes had really elegant and beautiful shapes and colours especially the ones that glow in the dark. Hence, why I was super excited to see how the GAN-generated images would turn out when combining nature images and jellyfishes! For the training option, I selected scenes because they best fit with the theme and images that I collected. I collected nearly 2000 images to be trained and used the process of 3000 steps that lasted approximately 3 hours. After the training process, it generated many images and a video which then I downloaded.
Concept: Landscape + Jellyfish
After training my model of landscapes mountains/skies/oceans and jellyfish it generating many images, and I could tell that there was a certain pattern throughout my images. I found the beauty in each of these patterns and was astonished at the variation of each generated image. Just comes to show how much more creative we can be with the combination of AI, and we can see the beauty of combining human creativity and artificial intelligence into an art piece.
Patterns:
- Watercolour and tie-dye-like effect where it looks like the ink is seeping through, so there’s a gentle blend and transition between the images that are being combined.
- Cell and bacteria look like shape and form. The structure of the images looks like something you would see under the microscope in a biology lab. I believe that these image shapes were more greatly influenced by the outline shape of the single jellyfish.
- Landscapes, because some look like real landscapes. Several of the images looked like snowy mountains due to the combination of light colours with mountain landscapes. I could also observe that several of the landscapes had also looked like they were cloudy and foggy as if it was high up the mountain.
- Crystal-like images because of the similarity in structure and colour. The sharp edges that could be depicted as cracked glasses, as well as the colour choices of bright blue, have a depiction of crystal rocks.
NFT in general
A non-fungible token (NFT) is digital assets like music, photos, games, videos, and art turned into an AI where it can be bought and sold online. I believe that this concept is a very modern and future cryptocurrency. It’s like collecting digital fine art, so having the ownership and copyrights of the original work can bring personal satisfaction and pride. NFT is an incredible system and network where artists and buyers can be supported, and gain profit. However, the NFT system isn’t perfect, and like with any system there are holes and loops. For instance, there have been times when cryptocurrencies have been stolen before. Sometimes the image quality deteriorates or certain file formats can’t be viewed anymore, websites that no longer exist, people forgetting their password to their wallets. Unfortunately, NFTs use blockchain technology for cryptocurrencies, so they use up a lot of electricity. This has a rolling effect of generating a lot of greenhouse gas emissions, which directly has effects on climate change. So NFTs are at risk of worsening global warming. This makes me rethink if NFTs is such a good platform to develop.
The implication of AI and NFT on art/human creativity:
With the implication of AI and NFT on art/human creativity, it lets everyone create and sell their work. It questions: “What is art? “ and “Who can be an artist?”. The concern of copyright becomes an issue because in this platform there is mass production of art where thousands of people post their work. Over saturation of the market can bring copyright issues, as to the originality of one’s work. The question of how original the work is to the artist is one of those tricky things to track and modulate.
There are issues regarding environmental impact and biases (I mentioned briefly in the previous paragraph on NFT): How energy consumption supplied by non-renewable sources isn’t sustainable and has negative impacts on the environment. Bitcoin is controversial due to its legitimacy and environmental impact, and now we live in a century that must be sensitive to climate change and global warming since it has an immense harmful direct effect on our lives and the rest of the entire world.
Biases, Art made by AI might have less representation if the data sets have less diversity, so the data sets need to have a range of data that doesn’t specify a specific category.
Lastly, who will own the art, only the rich? Ultimately you must pay to purchase the artwork, so artworks that are insanely priced can be only be afforded to the rich can be limited. As well as buying several artworks at once can limit other buyers to purchase the artwork, therefore, buyers who have a lot of currency may dominate the market.
Citation:
Droitcour, B. (2021). GANs and NFTs: AI Artists in the Crypto Space — ARTnews.com. Retrieved 10 July 2021, from https://www.artnews.com/list/art-in-america/features/gans-and-nfts-1234594335/holly-herndon-mat-dryhurst-crossing-the-interface-dao/
Lewis, S. (2021). NFTs may be the future of art — but are they threatening the future of the planet?. Retrieved 10 July 2021, from https://www.cbsnews.com/news/nft-art-environmental-costs/#:~:text=The%20growing%20awareness%20of%20NFTs,countries%20like%20Qatar%20and%20Hungary.
NFT art made by Artificial Intelligence. (2021). Retrieved 10 July 2021, from https://aimade.art/
