Fermynwoods Focus – Zhengzhou and Zhengyang Huang

Art critic Lucy Lippard may be best known for her writing on the emergence of conceptual art in the 1960s. Following her thoughts on The Dematerialization of the Art Object to their natural conclusion it may be unsurprising that this has led to a more recent focus for Lippard on Environmental Activism.

Spiral Jetty, Robert Smithson, 1970. Photo by Retis, CC BY 2.0.
Spiral Jetty, Robert Smithson, 1970. Photo by Retis, CC BY 2.0.

After making Spiral Jetty – a work which Lippard described as “the best known earthwork in the world”, artist Robert Smithson declared himself a ‘geological agent’. Susan Ballard and Liz Linden have suggested that the work could be the first artistic marker of the Anthropocene, beginning to articulate human relationships with environmental transformation. Created at a time when the water levels of Utah’s Great Salt Lake were unusually low, the work now draws particular attention to the climate crisis as years of drought continue to expose the structure from receding water levels.

Zhengyang and Zhengzhou Huang’s Fire is a more recent work, which speculates on the environmental impact of Machine Learning and graphics processing units (GPUs), firmly locating the energy required by contemporary AI models participating in the materialisation of art within the context of our burning planet.

Commissioned for Xylophobia: Online, our exhibition of digitally based work which takes its name from the fear of wooden objects, forests, or wooded areas, whilst inquisitive and mindful of the environmental impact of creating and presenting artworks using digital technology. Zhengyang and Zhengzhou describe Fire as a wood-based GPU kit outside of computers for use in a currently conceptual scenario where mineral-based GPUs run out and Fire as a wood based GPU is the only type available for any further AI training.


Presented initially as a website element, the work proposes that each speculative Fire GPU kit contains four wooden triangular panels to be assembled into a tetrahedron, more commonly known as a pyramid (another geometric symbol and man-made landmark much like Smithson’s Spiral Jetty).

On the surface of each panel algorithms, models, and datasets are to be etched. To run code requires power – in this case the wood is to be burned by fire. A single use GPU kit that places a more finite emphasis on the idea of ‘executing’ code. The artists note that Fire can hold 196, 324, or 784 pieces of data, with the capacity corresponding to the size of the wood to be burned.

To test the capacity the artists part realised the work as a series of laser cut panels each bearing real world code now in use in the public domain. The smallest pyramid contains the training and model scripts based on GPT2, a precursor to the more familiar ChatGPT currently making headlines. (GPT2 has an estimated 1.5 billion parameters whereas the latest GPT4 release has a number of parameters estimated to be in the trillions.)

FIRE324 contains training and model scripts based on DCGAN, or Deep Convolutional GAN, a generative adversarial network architecture used to generate images. The largest pyramid contains training and model scripts based on Open Jukebox – an impressive neural net that generates music, including rudimentary singing, as raw audio in a variety of genres and artist styles. Where these products already exist, Fire reveals hard limits to these often invisible processes.


Fire also consists of an interactive downloadable demo that places users within a bucolic forest resembling the aesthetics of an animated Disney film, where the burning of the etched wooden pyramids disrupts the pristine landscape. By making the work downloadable the artists ask participants to consider their role in activating the work. Ultimately Fire provides a situation where data collection, analysis and consumption ― essential in forming powerful and expensive AI models ― should be carefully considered. Echoing Lippard’s call to consider the social, political, and ecological ramifications of artistic creation.

James Steventon

The final sentence of this text was provided by ChatGPT.

This work took 75kg of carbon to produce, according to The Networked Condition carbon calculator tool.

Zhengzhou and Zhengyang Huang are a group of artists working with alternative techniques, speculative systems and imaginary avatars. With their practice they try to situate themselves in the current reality layered with data tracking, digital policies, and algorithmic decision making. Throughout their works they hope to navigate through layers of mediation that segment, percolate but may also connect our identity, knowledge and memory. They have shown their most recent series of works about alternative technologies, Air, Water, Fire at Society for Literature, Science and the Arts Conference at University of Michigan and the Last Online Show at University of California, Los Angeles. Zhengzhou and Zhengyang are currently based in Los Angeles (US).

Fire was commissioned by Fermynwoods Contemporary Art for Xylophobia: Online – an exhibition of digitally based work which takes its name from the fear of wooden objects, forests, or wooded areas. Responding to the increased needs and fears of outdoor activities in the wake of the Covid-19 pandemic, the featured work presents explorative, experimental and conceptual spaces at this intersection, whilst inquisitive and mindful of the environmental impact of creating and presenting artworks using digital technology. Funded by Arts Council England.