From data to stone: How AI is rebuilding Armenia's lost architecture for the future

The National

8 hours ago

Rows of engraved rock structures line a factory floor in Yerevan, bound for the Armenia Pavilion at the Venice Architecture Biennale. They look ancient, though they are still damp and fresh from the cutter.
As the team prepares the shipment, a factory worker stops in front of one of the arches. 'That's from the church in my grandmother's village,' he says. 'The church by the forest. This was over the entrance.' He is certain.
The arch is not from that church, but that moment of recognition, however misplaced, is proof the project had worked. It made memory, with all its flaws and reconstructions, tangible.
The slippage between memory and material is at the core of Microarchitecture Through AI: Making New Memories with Ancient Monuments, the Armenian Pavilion's presentation at this year's biennale.
The pavilion is being led by Armenia 's Ministry of Education, Science, Culture and Sports, in collaboration with the Tumo Centre for Creative Technologies and Electric Architects. It was developed with the support of CALFA, MoNumEd, and US artist Ari Melenciano.
'We wanted to see whether we can use technology to emulate the act of remembering,' says Marianna Karapetyan, the pavilion's curator and co-founder of Electric Architects. 'If there was an option of downloading the distorted image of a memory, and printing it, then maybe it will look something like this. Because you won't be able to remember it in detail.'
The exhibition takes over a sprawling warehouse space with a series of small-scale structures: arches, columns, capitals (the top section of columns) and fragments. Each is carved from tuff, the volcanic rock central to Armenian heritage and used in its ancient monuments and modern architecture.
At first glance, they could be mistaken for genuine artefacts. But a closer look reveals the uncanny: engravings are ill-defined, crosses twist out of form, the Armenian etchings are incomprehensible, and there are plenty of motifs and designs where they shouldn't be.
The forms were generated using AI, trained on real data and 3D scans of historical Armenian sites. That data set is part of Tumo's continuing initiative to digitally preserve the country's cultural heritage.
In recent years, the centre has undertaken the scanning and archiving of endangered architectural sites across Armenia – monasteries, churches, khachkars (carved stones bearing crosses) and vernacular structures – many of which are threatened by conflict or environmental degradation. That growing digital archive became the foundation for the pavilion's presentation this year.
'From the beginning, we knew we wanted to present a project based on the Tumo heritage scanning initiative,' Karapetyan says. 'The initiative began after the 2020 war, in the interim before Karvachar (Kalbajar), Kovsakan (Zangilan) and other areas were surrendered to Azerbaijan. The team went in and, within two weeks, documented everything they could using laser scanning and photogrammetry. The scans are extremely high resolution – accurate to the millimetre.'
Several of the sites scanned by Tumo have since been destroyed or altered, Karapetyan notes, including Saint John the Baptist Church (Kanach Zham) and Ghazanchetsots Cathedral.
'Many of these sites only exist digitally,' she says. 'We knew the Tumo team had done an incredible job and wanted to present the implications of their work at the biennale in some way.'
The Tumo team has scanned more than 260 monuments and the centre is developing an open-access platform to serve as a repository for Armenian heritage. The platform will feature immersive virtual tours and scholarly resources, aiming to make Armenia's centuries-long cultural legacy accessible to academics and the public alike.
While Tumo's platform focuses on preservation, the pavilion took a different path. The 'artefacts' on display do not reproduce any specific monument. They are not reconstructions, but reimaginings – sculptures that speak to the erasure of history, the instability of memory and the possibility of new forms of preservation.
The designs were machine-engraved, without human adjustment. 'We wanted to avoid human interference altogether. We didn't meddle with the designs, and there were no prototypes,' Karapetyan says.
'The sculptures are presented as the AI designed them and as the machine engraved them. We identified a handful of typologies of ancient Armenian architecture, their essence or DNA, so to speak, and let the AI come up with new designs.'
The idea was to explore how scans of endangered heritage – whether threatened by conflict, neglect or climate change – could be used in a new way – 'to give new life to that information,' she says.
The geopolitical stakes the exhibition alludes to are urgent and thought-provoking, but the pavilion does not pretend to offer solutions. Instead, it asks how cultural memory may persist when the physical world is no longer accessible. Can heritage survive as suggestion rather than structure?
In a biennale dedicated to architecture, the Armenia Pavilion stands out not for what it builds, but how it remembers what was lost and how it reimagines the space left behind.
The result isn't mournful or didactic but reflective. The proposal it offers is even uplifting: memory, like architecture, doesn't have to remain fixed to be real.