To create this atlas, natural objects (rocks) and anthropogenic substances are sampled during expeditions to geological outcrops. They become specimens to train machine learning models and explore the latent space between the natural and artificial. The outcome—hybrid digital artifacts—are captioned by publicly accessible image classifier algorithms.
This atlas was initiated by New York/Berlin-based media artist Pascal Glissmann in 2018. It is in progress—new collections/ albums/ plates & minerals are generated continuously.

— Material & Existence
An Atlas of Mineralogy is an illustrated inventory of the inanimate material forming the surface of our planet.
       Human beings have always been compelled to understand this surface better as it relates directly to our being. James D. Dana describes this in his Manual of Mineralogy(1857): "the very existence of many of the arts of civilized life, depends upon the materials which the rocks afford." Yet we are not only driven by scientific curiosity. There is a deep, almost supernatural, fascination: "the student of Mineralogy (…) finds abundant pleasure in examining the forms and varieties of structure which minerals assume." (James D. Dana, Manual of Mineralogy, New Haven, 1857) Dana refers to the senses: we look, touch, smell, and even taste to learn more about the world through rocks.

       Modern technology extends our senses: satellites enable us to study our planet's surface from the orbit; x-rays and electron microscopy allow us to look inside rocks; and digital networks enable a simultaneous comparison of global shifts in geology. 

— Naturalists  & Machine Learning
Scientific Atlases of the 18th/19th century compile and circulate newfound knowledge—often driven by emerging technologies. This "Atlas of Speculative Mineralogy" employs machine learning and deep neural networks to instigate a conversation with the past and the future. It revisits the practices and visions of 17th/18th-century naturalists through today's computational algorithms.
       Naturalists—driven by a fascination to discover nature's patterns—were scientists, artists, publishers, authors, editorial designers. They were polymaths. Today, our world is a complex system of interconnected networks. While the individual might be able to understand a sliver of it, its totality is impossible to grasp. Inspired by a polymath's ability to look across disciplinary knowledge, this atlas combines diverse practices and branches of knowledge to discover unexpected relations.

The visuals introduced in this atlas are digital artifacts created and contextualized by a sequence of machine learning models. The input is a collection of specimens that I collected during micro-expeditions. These day-long journeys take me to specifically chosen geological outcrops to sample (1) rocks representing the local geology and (2) anthropogenic substances found in its vicinity depicting globally circulated human-made materials. I am not using instruments to look at these samples (though, in a way, I do) but the samples themselves become my instruments. They are the extension of my senses to look at our world—and our various histories of exploring it.
How many specimens do I need to initiate my investigation? I found the answer in Alexander von Humboldt's notes on his expedition to South America in 1799. In a diverse mix of bags and suitcases—and with a lot of help—he carried forty-two instruments: microscopes, a thermometer, a dip Circle, a sextant, a cyanometer (to measure the blue of the sky)—to mention just a few. Inspired by this collection of specialized instruments, I am sampling forty-two substances as my working objects. They are divided into two groups: twenty-one (1) rocks and twenty-one (2) anthropogenic substances.

(1) Rocks: the components of rocks are minerals. Minerals are substances formed by natural geologic processes without human intervention. An outcrop appears when a rock layer or formation is not covered by soil or other bio-material. It allows sampling the local geology that—in most cases—has a history of hundreds of million years. This atlas explores outcrops located within—or close by—post-industrial areas and transportation hubs of the 20th century which contrasts the outcrop's static geo-location with the global mobility of everyday objects.

(2) Anthropogenic Substances: our everyday lives depend on human-produced substances. They allow us to travel, carry our food, and even cure diseases. However, they are simultaneously one of the most significant threats we face. In all likelihood, they will remain on this planet as witnesses of the Anthropocene beyond humankind. For this investigation, substances are sampled based on their current location. What unites them is their vicinity to a specified geologic outcrop—not their function, origin, or chemical composition. If we deconstruct these artifacts, we might find that their compounds have traveled globally. In this sense, looking at these working objects reverses the idea of scientific expeditions: not the observer but the observed traveled.

— GANs & Image Classifier
Photographs of the specimen are used to train GANs and create visual hybrids. These artifacts are new, unknown, speculative; using James D. Dana's words, "I find abundant pleasure in examining the forms and varieties of structure."
   How would a machine that is not sharing my emotional access nor my appreciation of form read this artifact? To find out, I am feeding the visualizations to a third algorithm: an image classifier. This image reader makes decisions based on how it has been trained by other humans—and their choice of training sets, including their biases. Its response is a list of tags. Keywords. Words and letters that feel random in their collocation. While these concepts are meaningless to machines, they depict how computational algorithms are trained and which aspects of current visual culture and language find their way into training artificial systems.

— Collections, Albums & Plates
The atlas consists of albums and plates which are collated into collections. In addition, appendices offer a more experimental perspective on the specimen. The elusive and uncanny visualizations do not provide any evidence to verify a scientific hypothesis. Instead, they invite us to question our understanding of the natural and the artificial; the archive and the truth; the past and the future; our material culture and rituals. What are the things we bring into this world, and what will we leave behind?