An artificial intelligence algorithm called Botto made about $ 1.3 million in its first six jobs as NFT. The auction takes place on the SuperRare marketplace.
When creating a painting, Botto generates random lines of sentences and transfers them to the VQGAN neural network, which creates images using examples of famous works of art.
The CLIP algorithm then determines how close the result is to the original string of words. If necessary, it forces VQGAN to redo the image until it reaches a “passing grade”. CLIP also selects a two-word title for the picture.
Botto uses the OpenAI GPT-3 language model to describe the image. It generates up to ten different pieces of abstract poetry, which the development team reviews, edits and selects the most appropriate one.
The algorithm generates 300 images per day. Then the “taste model” selects 350 paintings every week and sends them to the community to vote. The highest scoring image is converted to NFT and put up for auction.
At the time of writing, the project managed to realize six paintings. The first two were sold for $ 300,000–400,000, but the cost of the last lot was $ 81,000.
The developers used the proceeds to develop the project, which issues its own Botto token, which is used by community members to vote for the best, in their opinion, options. After each auction, some of the coins are redeemed and burned.
