Gradient Descent

From neural to railway networks

Gradient Descent’s adventure delivers unusual results across various mediums, oscillating between artificial intelligence and intelligent artists. Graffiti writers are, by nature, creative beings and masters of repurposing whatever they find. Whenever circumstances do not look promising, writers are keen on finding ways to modify and hack conditions to their favor. Looking for tools to reappropriate does not necessarily have limits in the physical realm. The software is used the same way the founders re-contextualized spray cans from an industrial environment and transcended the tool into the art world. This project unveils lots of parallels to the approach of the graffiti pioneers in New York back in the 1970s. Their playful approach draws much strength from an uncompromising purity and naivety, which Gradient Descent tries to re-discover with the help of new technologies.

The process of painting based on previously accumulated data reveals a lot about human learning, not only in the field of graffiti. It points to how refreshing a partial unlearning of these traditional norms can be. Reflecting on both artificial and human iterations of designs, the outcome is exciting and uncertain. While exploring, this method poses questions of agency and man-machine cooperation in creative visual media. Hence, Gradient Descent is translating old-school aesthetics into the modern-day digital age while preserving the fundamental structures. The output of this research is the book’s content, paired with an audiovisual installation by Theodor Guelat.