Skip to content

Fuelling Apocalypse

2024
Digital prints on paper mounted on acrylic glass with aluminium backing
150 × 60 cm (each)

As is well known within the industry, Singapore’s oil refineries boomed through the 1960s and 70s largely by supplying the Vietnam War. As recounted in Fuelling the War: Revealing an Oil Company’s Role in Vietnam, a book by Louis Wesseling, who headed Shell’s Vietnamese operations during the war, both the American-South Vietnamese forces and the Viet Cong were relying on oil sourced from Singapore’s refineries, with the latter obtaining the fuel either by capturing enemy supplies or through networks of small traders.

However, outside the industry, the fact that Singapore’s oil industry has historically profited from the ravages of war is rarely discussed or even acknowledged in the public discourse. Fuelling Apocalypse addresses this lacuna by turning to cinema in search of visual traces of Singapore’s involvement in Vietnam. The process involves training an AI model on film stills of the 1979 Vietnam War film Apocalypse Now and getting it to generate images based on quotes from Wesseling’s book, from which a set of images is selected by the artist to be each paired with the relevant page from the book. Given that like the actual war itself, Apocalypse Now, which was filmed in the Philippines with US military equipment, also relied on oil sourced from Singapore’s refineries, these images can be approached not merely as visual descriptions of the Singapore connection but also as the end result of a signifying chain literally fuelled by oil refined in the city-state.

Yet, such a connection is also partly rendered oblique by the generative AI process itself: similar to how the industrial practice of blending oil from different sources can obscure the origins of the oil that is traded on the market, machine learning involves reinterpreting the training data as probabilistic distributions of noise, which is then redistributed to produce “original” images wherein no definite trace of the training data can be discerned.