OPERATIONAL BODIES: THE TOMATO TASTED LIKE AUTOMATION OF CARE

Every evening, the Plantalyzer robot slowly rolls along the heating pipes between rows of tomato plants in the greenhouse. During its 6 hour ride, it snaps images every 10 centimeters with its 6 cameras. After their capture, it uses AI to analyse the visual images for data, identifying the number of tomatoes and their colour. The grower receives a report containing the current harvestable yield, classified into 12 categories of ripeness. The images on display here were captured between 2 and 20 May 2022 along 60 meters of one tomato row in a state-of-the-art greenhouse. Normally only ever glimpsed by machines, now for everyone to see at the exterior of Waag Futurelab.

SKOTOPOIESIS (2014- 2022)

5-minute excerpt of 20-hour film, documenting the 20-hour durational performance Skotopoiesis, and displayed as a 2-channel video (total and close-up of cress). Filmed during OuVert festival, Bourges, 2019

Skotopoiesis, Click Festival, Helsingor, Denmark, 2017

Skotopoiesis at Metaboles festival, Marseilles, 2021

INSTITUTE FOR INCONSPICUOUS LANGUAGES: READING LIPS (2019 - 2020)

The central premise of the work Reading Lips was that the cognitive mistake (apophenia) of seeing mouths in stomata would occur both in humans as well as algorithms made by them. Indeed even early experiments with off-the-shelf applications showed successful ‘mouth’-detection (far left), which we followed up with LipNet Neural Network implementation; to get this algorithm to work, however, we had to change the colour of the stoma to match (white) skin tone. After that, LipNet had no trouble interpreting the leaf pore.

PL’AI (2020)

PL’AI, the AI-controlled robot made to enter plant time and be able to play with cucumber tendrils at plant speed, was a complex and challenging artwork because of the translations between living materiality and the abstract realm of the digital. The sensors are limited in their modalities and capacities, so we developed many tools that helped visualise what was going on in the “mind of the machine”, creating hybrid representations of video and simulation to assess what the robot was sensing and how it was making decisions. Left: side-by-side comparison of top view machine vision analysing tendrils touching the coloured balls. Right: a latent-space walk - a reanalysis of the matrix of the cummulative 3D LIDAR scans to travel along the axis of time. Below: excerpt of close-up time lapse of cucumbers playing with the robot during the exhibition.