Manipulated AI / machine learning generated images. 2023.

When AI large language models generate false information, it’s called hallucination. The same term applies to the inflated promises the tech industry makes about these systems’ transformative potential.

Working against the scale that defines contemporary AI, this work trains Generative Adversarial Networks on intentionally minimal datasets — a handful of nature photographs or just one or two solid colors, rather than the millions or billions of images typically used. With such limited training data, the system produces images that expose both the digital artifacts of its processing and its attempts to construct meaning from insufficient information.

These AI-generated images are then manipulated with custom code, channeling my emotional responses to the tech industry’s false promises. The manipulations become gestures of anger and skepticism, searching for and amplifying patterns that may be apparent to the machine but remain invisible to human perception. The resulting works preserve both the original computational grid and traces of the system’s constrained understanding, now layered with critical intervention. Each piece expresses the gap between technological reality and industry mythology.

The titles are quotes from tech leaders promoting the revolutionary promise of their systems — statements often as detached from reality as the hallucinated outputs of the algorithms themselves.

---

Press

---

Related Work

Tabula Rasa - AI / machine learning generated images

Manipulations - Manipulated AI / machine learning generated images

Flag - Manipulated AI / machine learning generated images