NFTs and GAN in the Creative World

Original Source Here

NFTs and GAN in the Creative World

by Divasha Renata Nareswari

Art and technology are two intellectual spheres that have been increasingly fused in the past few years. This statement holds true for both its creation and distribution. The term ‘NFT’ has been popular amongst the public recently with the media heavily trailing behind its recent stratospheric rise. The idea of selling unique copies of your creation is now becoming a digitized process no longer exclusive to traditional art, evoking a spark of active conversations of its utilization in the creative space. We may want to also explore the possible entrance of AI-Generated artworks in the marketplace and its potential to extend this development even further. In delving into this newfound field, it’s essential for us to continually re-evaluate its current impacts and capacities to determine if all these new jargons surfacings will merely be a short-lived fad or be able to truly revolutionize the art scene in the long-term future.

My own GAN-Generated Images

Nowadays, more online softwares has allowed the public the opportunity to generate their own artwork using GAN models. Generative Adversarial Network models use deep-learning methods of neural networks to generate new outputs by discovering and learning the patterns identified in the original dataset. RunwayML is an example of a software that allows you to do so. These are few examples of some of the artworks I’ve created using this software.

Image 1 generated from GAN Model
Image 2 generated from GAN Model
Image 3 generated from GAN Model

The main concept of the images was to find Space within Earth. Among the pictures, you’re able to feel the atmosphere with elements of the spacescape yet combined with the familiarity of nature (trees, mountain, river, etc). As someone who loves reading about the universe and is interested in photography, visiting places with mesmerizing sceneries like this has always been a dream of mine. The images were able to simultaneously give out feelings both calm and dreamy yet intense with their strong emboldened colors. The goal is for the audience to feel as if space is tangible and within our reach.

In order to create these images, I used a dataset of around 1500 images of Aurora Borealis (the Northern Lights), Space images, and Forest landscapes. I’ve manually selected these photos through stock photo sites and google images online. I’ve also searched through personal blog sites and popular social media platforms. I used several palettes of the scenery, ranging from a green tint to a more purple hue. They have different structures and distribution of light objects and distinction of color patterns. There are a variety of elements of nature ingrained in the pictures. The following are some of the images utilized in my input dataset.

Image Dataset

I trained this model using the software’s existing landscape pre-trained model. I used around 2000 steps of training to generate around 100 photos. The process took around 2 hours. The following are the original pre-trained model images.

Original pre-trained model images from RunwayML

What is an NFT and how does it work?

More onto the distribution of art, NFT has been a revolutionary technological development in this field. In short, a non-fungible token works as a token to certify a digital object’s uniqueness and sell its ownership. Non-Fungible refers to an object being only one of its kind and has a value that cannot be replicated. It works similar to how physical art could be signed by the respective artists to create distinct value from other copies being sold. Now, NFTs allow the creation of a marketplace where digital assets can now be through the same philosophical mechanism. Except now, the authenticity is verified using an extremely powerful computer.

Difference between Fungible and Non-Fungible

NFTs exist on a blockchain, a distributed public ledger where transactions are recorded. Each chain block uses cryptography to uniquely identify a set of data in comparison to previous blocks. Paying for an NFT earns you the right to transfer the token to your own digital wallet. The most commonly used NFTs are part of the Ethereum blockchain.

However, it’s essential to remember that buying an NFT does not mean you are buying the actual digital artwork. Rather, a token consists of a link for the artwork and its description. Most blockchains are still incapable of fitting an entire image inside. The blockchain will also contain all the details and history of the asset’s owner, making it more difficult, if not impossible, for it to be replicated. In addition, different NFTs contain varying contracts ingrained when consuming. Some contracts set terms and conditions for larger monetary incentives. An example would be that for every resell, the original artist may still obtain a certain amount of profit from that transaction. It allows them to receive compensation even after the value of the artwork rises through time and resells, something traditional systems failed to do. These contracts are set per individual artist, erasing the power imbalance previously experienced where artists simply had to accept them.

Though the idea of spending extra millions on artwork may seem strange to many, it inherently feeds onto a human’s innate desire towards ownership. UCLA Professor, Valentin Haddad, pointed towards the endowment effect, which states ‘people value things more when they own them, simply because they own them’.

Impacts on Human Creativity


The creation of this new marketplace essentially democratizes the art scene. The traditional art scene is often perceived as elitist. Its accessibility is highly skewed to the rich, with the consumption being concentrated with auction houses and gallerists. Now, consuming and selling art has become a much more open and accessible space that erases the need for an intermediary ‘middle-man’. The technology directly connects artists with potential buyers in each respective platform available. Its presence on the internet, a web with interconnected audiences, allows artists to create viral community fan bases with their artworks.

Ultimately, this accessibility generates more incentive for artists to produce more artworks. It allows existing voices to be louder and empowers newer creatives to enter the marketplace. Not only are is the contemporary scene filled with elitism, but this elitism also naturally comes with biases within the consumption of individual artists’ works. Delocalized platforms allow racial and other aspects of an artist’s identity to be excluded from the process of purchasing artwork, and thus once more, creating more incentive and freedom.

Artworks produced across markets are also often highly influenced by the type of market that is currently prominent. With the rise of a digitalized market, we could expect artists to start investing in more digital tools to create their artwork and alter their main medium. With the rise of NFTS, there are also more Generative Artworks being produced. Thus, it all creates more capacities for human creativity to expand.


The same impact could be predicted with the use of AI models with Art. As of currently, computers are still unable to replace higher-order levels of creativity. However, they can create environments that can optimize and catalyze our creative output. Instead of disparaging its role in the creative world, we should view them as our collaborators.

An example of GAN Models, we no longer have to create art from scratch, but can build onto the idea of art itself and feed data needed to create it onto our computers. Neural networks can output novel patterns that have never existed in the world prior based on our input. And they can do so autonomously. This creation of an automated creative process machine learning offers has allowed the art-making process to be easier and faster, along with extending the current possible shapes of narratives artists can undertake.

Instead of replacement, they essentially allow the augmentation of human creativity. Automating proportions of creative tasks have freed up time for artists to increase productivity in other areas of the creative process. When machines generate new images, artists are able to spend time planning and creating new ideas and concepts.


There are still shortcomings in this system. Is someone an artist if they feed photos onto a pre-existing trained model? I say, to some degree, the artistic value could still be found in providing training data. However, so with the easiness of Generative Art, does come with possible implications of mass production and blurring the relative line of value that can come from selling these works.

There may be copyright issues that arise from using AI-generated models. When we put datasets into the model, will an artist know whether their creation is being utilized as an input to the model? Transparency can be more difficult to track and easy to hide.

Moreover, NFTs have sparked much negative feedback as more concerns start to arise against their ecologically destructive nature. Blockchain technology has a large amount of network traffic and storage needed, thus much more electricity consumption.

According to Digiconomist, a single Ethereum transaction has an estimated footprint of around 35 kWh on average. An individual act of sale in the network can sum up to 87 kWh. A comparison set by writer Memo Akten describes the average power to be equivalent to an EU resident’s electric power consumption for four days. Its carbon footprint is still yet to be fully analyzed to this day. There needs to be a transformation in power consumption for these transactions or many protests that the technological advancement is not at all worth its damaging impacts on our environment.

Final thoughts

In conclusion, the presence of NFT is still very early in its run. So far, it has shown much potential to revolutionize the art scene we currently consume. Moving forward to the future, we could expect the growth in research for using technology that utilizes renewable energy sources and more environmentally friendly hardware to exist within this space. Though AI has been around for a much longer time in comparison, its maximal potential is still nowhere near our current foresight. Instead of outrightly rejecting the implications of technology in the art space, we ought to re-focus on our discussions on the ways for refinement and fixing the system’s current mistakes.



Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot

%d bloggers like this: