Food has always terraformed, and landscapes have always created recipes. As instruments of this feedback, humans have been making active correlations between ingredients and terrain for millenia. We shape landscapes to live in and shadow landscapes to feed from. This landscapes overlap in what Europeans still call ‘the countryside’. We began this project by asking how the ritual acts of eating and shaping landscapes can be put in more explicit interplay to the benefit of both. We first visited an agri-tourism operation near Moscow that builds pleasant and productive landforms together. We discovered that it was unprofitable and fantastically inefficient at both. Very difficult to picture at platform scale.
Platforms already do food. IBM’s Chef Watson is an AI that produces quirky recipes based on patterns of ingredients that tend to appear together in its dataset of dishes. Pairing companies like 'Deliveroo' link establishments to individuals. Blue apron penetrates further into the private domain, delivering ingredients and recipes to dormant kitchens and podcasts. They pitch a more direct connection between food manufacturers and urban eaters, but at the same time retain the integrity of recognisable ingredients and the aesthetics of provenance. We were fascinated by their puriton and bucolic representations of source and site, best exemplified by the slogan: “food is better when you start from scratch”
The history of Russian agriculture teaches us that there will be no starting from scratch. Our research took us through the Stolypin reforms that privatised feudal land-holdings and led to the Bolshevik revolution. We visited the Soviet five year plans that forced pre-determined reaping patterns on the Russian landscape and led to mass starvation. In a meeting with the Minister for Agriculture of Moscow Region, we learned about current government incentives to develop vast abandoned growing sites that are surprisingly close to the city, and still very cheap. Low-capital-investment monocultures are still standard practice in Russia, just like in other landed states - wheat, meat and dairy. The sanctions of 2014 put pressure on the foreign supply of fresh fruit and vegetables, which hold a special luxuriant status in Russia The following year, a 31 hectare greenhouse complex was built on the outskirts of Moscow at a cost of $500 million US dollars The cucumbers within grow quickly and predictably. They are monitored intensively around the clock and throughout all seasons. Agricultural interiors such as these are becoming standard fixtures of European landscapes.
The lingering conflation of nature and agriculture has been a disaster for both and their separation must be completed for the sake of other remaining ecosystems. The energy-intensiveness of bringing crops indoors may be justifiable if they are located close to sites of consumption, and the intensity of their outputs is dialled right up. The trophic cascades of forests and the backyard growing movements that mimic them work too slowly and at too small a scale to be of much help to us now. If new ecosystems are synthesised indoors, they do not have to follow the same steps, rules and pace. If new ecosystems are synthesised, then so should their products be. Our attachment to the consistent identities of what we eat is so strong that entire biomes have been demolished for it.
We are proposing bounded landscapes that are sensed and expressed by an AI that manages their intensity and complexity. Artificial intelligence is already widely used in agriculture, mostly for tracking and monitoring. Machine vision is used to identify defects and diseases in plants 24 hours a day. It can determine when a goat is ready to mate from the way it walks when it goes to be milked. Soil sensors already measure chemistry, water content and acidity. Robotics are used to weed and plant and pick produce. IoT devices already address boxes and crates. Location trackers are tinier than ever. It is a very small step to imagine the management of whole sites by AI systems that index the growth, interaction and formation of multiple hybrid species at once. This slightly echoes the permaculture principles of companion planting and food forests, but at a much faster pace and with much stranger outputs. And strange they must be. AI that senses a landscape must also express that landscape’s propensity for experimentation. It may be the human taste for predictability and equilibrium that destroys ecologies. Taste, after all, eats itself.
The name of this project is tuda syuda, which means ‘back and forth’ in Russian. It has slightly dirty, hustling connotations. It represents the central idea of two AIs that negotiate food and landscape priorities between them. In testing the conversations they might have, we started to see an undeniably social interaction forming, stripped of culture and relentlessly co-operative. If one proposes the addition of something new into their exchange, the other may reject thousands of iterations of it until its own oblique priorities are satisfied. As millions of these negotiations take place in both directions at once, the pace and complexity of cooperation increases geometrically. The AIs do not see the whole picture though, because in isolation or totality they would approach vanishing points of artificial idiocy. Their dialogue of misunderstanding mimics the experimentation of an ecosystem: the optimisation of inter-relations rather than of fixed outcomes.
As humans are necessary instruments in the interplay of food and landscape, they must be immersed in the conversation too. Image and speech have become the common ground for testing machine learning outputs today, but AI has proven itself as a master creator of new textures and mistaken identities. Machine-generated textures can express the surface qualities of construction materials and edible substances together. They can be built and ingested on the same plane. Textures have long been an interchange between food and landscape aesthetics, from still life, to impressionism, to shutterstock. As an interchange between senses, textures can be understood by touch and taste as well as sight and sound. For our purposes, texture is a rich and fuzzy interchange that can be juxtaposed against the accountability of Natural Language Processing.
The video that follows is the demonstration of an immersive virtual space that is both system diagram and interface. It can be navigated as a three-dimensional time machine that displays the history of the dishes for purchase at one and their accompanying landscapes at the other. In the middle are the textures of the negotiation between them. The interface becomes the decor for new dining rituals that take place in the display their effects, products and origins. A really augmented reality. An augmented reality in a meaningful sense