AI and Technology Unite to Preserve Armenian Heritage: New Memories for Ancient Monuments

Armenia has been grappling with preserving its cultural heritage, especially after the loss of Artsakh (Nagorno-Karabakh) in 2023, when it lost control of the region to Azerbaijan. Though many organisations in the country are making efforts to preserve heritage in several ways, Armenia’s display at La Biennale di Venezia offers a unique perspective - using AI to reimagine cultural heritage and preservation.

The Venice Biennale is a major international cultural exhibition dating back to 1895. Held in Venice, Italy, biannually, this exhibition encompasses various art forms like art, architecture, cinema, dance, music, and theatre. This year’s exhibition, being run from Saturday 9 May to Sunday 22 November 2026 focuses on the topic of 'Intelligens: Natural. Artificial. Collective.'

It explores how different forms of intelligence, including Artificial Intelligence, can work with natural and collective intelligences to address the challenges of a rapidly changing world.

Following the theme, the Armenian pavilion focused on using AI to explore the concept of how memory distorts but preserves the many monuments seen around Armenia, which might look different from each other but have elements that are common and unique to Armenian heritage.

Electric architects, an architecture and urban planning team that focuses on preserving and reimagining Armenian cultural heritage, in collaboration with TUMO centre, CALFA and MoNumEd, produced an innovative display in Venice.
TUMO centre for creative technologies is a free-of-charge educational program for teenagers, but it is also involved in digitally preserving Armenian cultural heritage sites by using 3D scanning and AI. During the war with Azerbaijan in 2020, TUMO scanned around 250 monuments in very intricate detail out of fear of losing those territories. Their teams went to the sites and conducted very high-resolution scans even under the threat of being attacked. By using the many scans from TUMO’s archives, the AI model was trained by CALFA and MoNumEd to produce 3D images that were carved into stone by Electric Architects.

This project kicked off last year when they entered a competition by Armenia's Ministry of Education, Science, Culture, and Sports, which decided to choose the Electric Architects team and also contributed towards funding the project.

Marianna Karapetyan, founder of Electric Architects, the curator of the project, spoke about how the idea was brought to reality. She reflected on how collaboration with TUMO brought depth to the initial idea, which was to carve into stones the scans TUMO had. However, by utilizing their 3D scans from Artsakh and around Armenia, they decided to take it a step further and train an AI model to reimagine what Armenian architecture looks like. That is when CALFA and MoNumEd, came on board to work on the AI aspect.

The main elements the team focused on while training the AI model were windows, entrances, Khachkars (crossstones), capitals and niches. For each of the five elements, the team looked at thousands of images, found each element, cut it out, and fed it to this AI model for it to learn and understand what a cross meant, what a window meant, what an element meant, so it could learn. It took a few months of this continuous process. The generative AI was also made from scratch, which rendered images for weeks.

Hulé Kechichian, a senior project manager at TUMO, said training an AI model in such a short amount of time was extremely tenuous since it required a great deal of data to understand a single element. They annotated each element they wanted the AI model to learn so it could identify them. They drew around the specific elements manually to feed the AI. After feeding AI numerous images of specific elements, they were able to ask it to produce 3D files based on what it had learnt.

She said that it was important to train an AI model from scratch because existing AI does not know much about Armenian architecture, as she tried to produce something on existing AI, and it was inaccurate. The images generated by AI, which captured the essence of Armenian heritage sites, were used to carve tuff stone, a volcanic rock widely used in Armenian architecture. There were 57 pieces produced.

The deeper meaning behind this is how memory preserves this kind of architecture, where artificial intelligence acted as a memory agent, as a way of remembering Armenia after looking at a lot of heritage, monasteries and monuments. The details that are remembered in the mind are not the same after seeing all of them.
“And if there was a tool to bring out your memory back to the physical form, that would be exactly this thing,” Maria Karapetyan said.

She mentioned how they realised the model worked because they heard construction when they were packaging the stones. Workers were talking about how they recognised the carvings from their village or some other place in Armenia, not realising that the carvings were not specific at all.

Sipana Tchakerian, CEO of MoNumEd, an initiative that focuses on using Artificial Intelligence to revitalise architectural monuments, spoke about how, as an archaeologist, the idea of monuments is closely related to the idea of memory "The term monuments comes from the Latin word "monumentum", which derives from a verb that means 'to remind' - monere." So, the memory is the concept itself of the monument. So it's really interesting - in the axis we were following for the project.”

This ambitious project came to life in the span of several months, but required a lot of funding, which was handled by TUMO through fundraising. This project was supported by several funders, from the training of the model to the execution of the project, to the transportation of stones to Venice and setting up the biennale pavilion. It took several boats to send it to the location. Karapetyan mentioned that the project cost several times more than they had initially calculated.

One of the funders was Team Telecom Armenia, which is the strategic partner of TUMO. Since the acquisition of Beeline by an Armenian company, as part of Team’s rebranding, the social direction of the company was decided to focus on projects that support national values and preserve national culture. Grigory Yurin, head of the public relations department at Team Telecom Armenia, noted, as the company aims to support initiatives that bring together national culture and digital technologies they were delighted to support the idea of the Armenian team to present microarchitecture through AI.

“It stood out the most to them was the potential of the project because it not only contributes to the development of technology, but also has national significance for the preservation of Armenian identity. The Venice Biennale this time effectively unites both technology and the preservation of Armenian content”, Grigory Yurin adds.