Understanding the Invisible

Part 1 - The Dataset

This first video shows the NFT dataset collected for the project. 12 million NFTs represented by 3 million images appearing on the screen at a variable rhythm: speed depends on the level of market instability. Images follow each other according to the time an NFT appeared on the blockchain. We observe the historical evolution of the phenomenon, including the different aesthetics and collections that monopolized the market in the beginning, to the rich variety of collections populating the NFT market today.

Part 2 - GAN Models

After observing the training data, GAN models are used to reproduce NFTs that appear to belong to their original collections. These fake NFTs are the first example of recycling and reusing this huge amount of NFTs produced by the artist community. Exploring the most capitalized collections in the market allows us to observe the small differences between the 10,000 "cards" in each collection.

Part 3 - Diffusion Models

In the final part, the collection of 12 million NFTs is used for fine-tuning a Diffusion model. This second example of recycling the creativity of the NFT community leads to a very different aesthetic. The images are not remotely similar to any of the training data collections—the Diffusion model produces a new aesthetic that often resonates as already seen, but somehow absorbs their style without tracing it. The exhibition closes by showing both the aesthetics of the NFT world and the consequences of having so many high-resolution images available to train new AI models. An endless path of creation and reuse. This was the first AI text-to-video film ever presented in an Italian museum.

The Most Extensive Exploration of the NFT Phenomenon

Understanding the Invisible is an immersive, site-specific installation that presents five years of NFT market and aesthetic evolution. This work represents the most extensive exploration of the NFT phenomenon conducted to date, combining scientific rigor with artistic creativity to reveal a world that is anything but in crisis.

"A world effervescent, vivid and in continuous evolution... an ecosystem interconnected and nuanced, with significant cultural and economic implications."
— Maria Grazia Mattei, MEET Founder

Data as Creative Medium

Using deep-learning analysis tools and artificial intelligence algorithms, the project transforms extensive collected data into what I describe as "a data-film suspended between reality and imagination." Rather than displaying traditional artwork, it creates an emotional visualization that blends data analysis with AI models.

The installation visualizes relationships and connections among millions of recombined NFT images through artificial intelligence, employing innovative text-to-idea methodology — positioning it as one of the first examples of digital art using this approach.

Research Foundation

This work draws from over three years of scientific research, resulting in publications and reports examining how museums and galleries utilize NFTs and their broader impact on cultural and economic sectors. The project serves dual purposes: examining the NFT revolution's practical implications while simultaneously reflecting philosophically on its potential and the metaverse's space-time continuum. Read the full scientific paper published in Nature Scientific Reports.

Exhibition at MEET

The installation was presented at the MEET Digital Culture Center in Milan from September 6 to September 29, 2022, with support from Fondazione Cariplo. The exhibition showcased the work in an immersive environment that allowed visitors to experience the data visualization in its full scale.

Ars Electronica Festival 2023

The project was also presented at Ars Electronica Festival 2023 in Linz, Austria, under the theme "Who Owns the Truth?" The work was showcased through a dedicated presentation examining how AI-driven storytelling can communicate findings about digital asset ownership and blockchain technology adoption patterns.