Short Essay on GAN and NFT
Concept and Process behind the trained model
My dataset for this GAN model training included 216 pictures of sunsets. The pictures were manually selected. For example, pictures with people, dogs, shadows, etc. were not included for the training dataset. The dataset purely included pictures of sunsets that ranged from different colored skies as follows:
My inspiration for this dataset was watching the sunset after 3 long weeks of seeing dull cloudy weather in Shonandai. Recently, the sky was bright blue with beautiful white clouds which reminded me how sunny weather was supposed to be like. During evening time, I glimpsed outside my window and the sunset I saw after a while just struck me and froze me in place. Something about the pink color in the sky filled me with calmness and patience.
After training the model using the ‘scenery’ training option with no pre-processing, and using 1500 steps of training, a few images were obtained out of which, this one really stood out to me:
Most of the generated images from the dataset looked very similar to normal sunsets and some pictures in the dataset. However, this one image was very interesting to me. It somehow looks like a sunset picture that includes a mountain looking shadow on the bottom, a seashore looking boundary above it, that deepens into the sunset itself. Perhaps if more training steps were conducted, maybe the final generated images would have been much more distinctive and interesting to look at.
NFT
Being introduced to NFT the first time, all I knew was that it was just another site that uses cryptocurrency for the buy-and-sell process of digital artwork. After looking into it further, it can be understood that NFT is more like a digital certificate of authenticity and NFT and unlike fungible cryptocurrency like bitcoin, is completely unique. They exist as letters and numbers that hold information about what was sold when and to whom it was sold, etc.
NFT is very sophisticated. NFT’s add a measure of certainty of the work uploaded being considered. Putting artworks on the blockchain is like listing them in an auction catalog.
With reference to the article ‘NFTs Weren’t Supposed to End Like This’ by Anil Dash, seems like when someone buys an NFT, they’re not buying the actual digital artwork. Instead, they’re buying a link to it. Turns out, NFT’s depend on one company that stays in business to verify one’s digital artwork along with pre-blockchain internet where artworks would vanish if people would forget to renew their domain name.
Impact of NFT and AI on art/human creativity
Through NFTs, mining artists are always listed as the owner of the art linked to NFT and provide a clear proof of ownership and authorship. Due to NFTs, there has been a lot of excitement and hype around the idea of using technology to sell digital art.
NFT gives artists the incentive and motivation to create new artwork as it can be done easily and comfortably. Perhaps, it can also boost one’s confidence once the digital artwork is bought and motivates them to create more artwork hence, also improving one’s productivity. NFT also provides various artists an equal opportunity and levels the playing field when it comes to buying and selling art as everyone can upload/buy whatever art they like.
With the help of AI, people can save some amount of time that they would normally spend on creating artwork. We have a little head-start in training pre-existing pictures to get new ideas that can then help people think more on how to design something manually or through an AI models again which in my opinion, is a very good way of improving human creativity and helping them boost and brainstorm ideas for more art projects.
