Subscribe now for full access and no adverts
In recent times, artificial intelligence (AI) has seen significant advances, fundamentally altering numerous facets of human existence. In the entertainment, healthcare, finance, and communication industries, AI is transforming how we interact with one another and with technology. In the realm of archaeology, AI is being utilised to reconstruct and simulate ancient environments and landscapes, sites, artefacts, and monuments, thereby challenging conventional notions of historical knowledge and providing fresh perspectives on the past.
AI Rethinks the Past, a new exhibition pioneered at Duke University, highlights the capabilities of generative AI technology in the specific context of central Italy’s Etruscan and Roman landscapes. Initially held from 22 April to 24 May 2024, the exhibition was an innovative gathering that fused historical scholarship, scientific inquiry, and cutting-edge technology. It is now travelling to Europe. Examining how AI was used to illuminate past Italian landscapes for the exhibition provides a sense of its archaeological potential.
AI archaeology
The category of AI algorithms known as ‘generative AI’ can create new content, including text, images, and videos, by analysing data and patterns that have been learned. In contrast to conventional AI systems, which predominantly focus on classification, prediction, and decision-making, generative AI models undergo training to discern the fundamental structure and attributes of the input data. This enables the models to produce unique and genuine outputs. These models can learn to imitate the structure, content, and manner of the original data through training on massive datasets; as a result, they can generate original works that bear resemblance to the training data.
In the case of AI Rethinks the Past, generative AI is being applied to palaeobotanical and paleoenvironmental data from archaeological sites to reconstruct ancient landscapes and environments. These data, comprising pollen, charcoal, and various plant residues, offer researchers a wealth of information regarding past flora and vegetation. As a result, they are capable of meticulously reconstructing ancient landscapes.
Vulci (Latium) and Roselle (Tuscany), are two significant archaeological sites in Italy, and present the focal points of the exhibition, which ranges through about 1,300 years of human history. These cities developed during the Etruscan and Roman eras (1st millennium BC- 3rd century AD), and can today offer significant insights into the intricate relationship between human settlements, their transformations, and the landscapes in which they were situated. Vulci was located on a plateau overlooking the Fiora River basin, and renowned for its thriving trade networks and its aristocracy, which made it one of the most powerful city-states in ancient Etruria. It became a Roman colony in 280 BC.
Roselle, located in the basin of the Ombrone River, was another significant Etruscan city and evolved into a Roman colony; it featured elaborate cisterns and aqueducts, among other impressive public structures. Researchers can learn how ancient civilisations sought to influence and enhance their surroundings through the examination of the ruins of these cities. To optimise the productivity of their flocks and fields, the Etruscans, for instance, developed sophisticated systems of land management and water control.
To reconstruct the environments and landscapes of Vulci and Roselle, the exhibition team made use of a wide range of methods, including palaeobotanical analyses, virtual reality, remote-sensing surveys, and archaeological excavations, among others. Archaeological investigations have revealed an abundance of material remains, encompassing artefacts, structures, and ecological data, which impart significant knowledge regarding the daily routines and customs of the Etruscan and Roman residents of these urban areas. Using technologies such as LiDAR, drone photography, and geophysical prospection, remote-sensing surveys have uncovered a wealth of lost landscape features and patterns, including ancient roads, buildings, channels, and agricultural terraces. Researchers have been able to map the layout and extent of the ancient cities with the assistance of these surveys, which has provided a broader context for the archaeological findings.
Through the examination of ancient plant remains, palaeobotanical analyses have been especially useful in reconstructing the environments and landscapes of Vulci and Roselle in antiquity. Pollen, charcoal, and additional botanical remnants have been procured from a multitude of archaeological sites, including the urban stratigraphic deposits of Roselle and a large Roman cistern that was recently excavated at Vulci. The samples comprise significant data regarding past flora and vegetation, enabling scientists to reconstruct the ancient landscapes.
An illustration of this can be seen in the pollen analysis conducted on material from the Roman cistern at Vulci, which unveiled a wide variety of plant species, such as untamed plants, cereals, legumes, and fruit trees. The significant abundance of cereals in the pollen record suggests that the cistern was most likely surrounded by crop fields, whereas the coexistence of fruit trees and untamed plants denotes a vegetation cover that was a mix of cultivated and natural elements.
Similarly, charcoal analysis in Roselle has shown that the oakwood forest found there was not uniform. It had holm oak, common oak, and hornbeam trees, as well as Mediterranean shrubs and plants. According to these data, there is evidence of human activities such as wood harvesting and land clearance in the surrounding landscape, which comprised a mosaic of woodland, shrubland, and grassland habitats.
Virtual pasts
Generative AI models have been trained using this palaeobotanical data to simulate and visualise ancient environments and landscapes from the complex taxonomy of the samples, which is to say long lists of Latin/scientific names. The exhibition team has successfully integrated scientific data with AI algorithms to produce a sequence of interactive and immersive installations that investigate the notion of a ‘mindscape’: this is an essential concept that relates to the way individuals or collectives perceive, comprehend, and interpret their surroundings.
The mindscape approach assumes a heightened level of intrigue when applied to ancient landscapes, given that it signifies the convergence of physical reality, historical understanding, cultural convictions, and the human imagination. Fundamentally, a mindscape is an intellectual depiction or cognitive cartography of a setting, influenced by a multitude of components, including individual encounters, societal heritage, and obtained information. It is a method of comprehending and establishing a connection with a landscape that transcends its tangible attributes, incorporating the sentimental, spiritual, and symbolic connotations associated with it. The mindscape concept becomes more intricate when ancient landscapes are considered, as the physical reality of these landscapes has frequently been altered or lost to time. A multitude of factors have, for example, contributed to our comprehension of primordial landscapes: archaeological findings, historical narratives, and inter-generational cultural practices. Thus, in addition to the physical remnants of the past, the collective knowledge, beliefs, and interpretations of the societies that inhabited or studied these landscapes also contribute to the formation of the mindscape of an ancient landscape.
By using generative AI to reconstruct these mindscapes, the exhibition encourages a more immersive and interactive experience with the past. The AI-generated visual representations are designed to evoke emotion rather than be precise replicas of the ancient landscapes: they aim to encapsulate the atmosphere and substance of these environments. A variety of AI-powered installations are displayed, including interactive touchscreens, video simulations, and a one-of-a-kind ‘AI Puzzle’ that generates real-time visualisations of ancient Roman landscapes through the manipulation of 3D-printed objects (see box below).
Large-scale video simulations projected throughout the exhibition space illustrate the generative capacity of AI in the reconstruction of landscapes. Vulci and Roselle’s landscapes can be seen undergoing gradual transformation due to the impact of weather fluctuations, shifting seasons, and human activity, all of which contribute to an environment that is intricate and occasionally unexpected. Using interactive touch displays, visitors can investigate the ancient landscapes in greater depth through a more tactile experience. They can explore, for example, the Roman cistern at Vulci and the important data that it contained. Alongside the AI-driven installations, the exhibition showcases a selection of 3D-printed pollen grains, magnified 3,500 times their actual size. By utilising microscopic and CT-scan imaging and 3D-printing technology, these intricately detailed models highlight the diversity of ancient plant life, establishing a concrete link to the environments and landscapes of antiquity.
An AI chatbot was also designed in the likeness of Pliny the Elder, a celebrated Roman author who lived from roughly AD 23 to 79, when he died during the eruption of Vesuvius. Pliny’s extensive Natural History, a 37-volume compilation that distils ancient knowledge about subjects such as geology, botany, and zoology, was employed in the training of the chatbot. It was developed by ChatGPT via the Social Intents platform. Subsequently, instruction phrases were employed to modify the automaton, while maintaining its integrity and imparting any information that was omitted from the training material. This ensures that the bot operates within the designated timeframe, being apprised of Pliny’s demise, and furnishes precise responses as opposed to generic ones.
The chatbot offers visitors a distinctive and captivating method of acquiring knowledge regarding historical subjects by enabling them to pose inquiries and obtain responses in plain language (English, Latin, French, and Italian). Visitors can delve into various subjects about ancient environments and landscapes via their engagements with the chatbot. These subjects may include Roman agricultural methodologies, the therapeutic applications of plants, and the impact of the Etruscans on Roman understanding of the natural world. The automaton additionally offers visitors the chance to contemplate how the sources and perspectives at our disposal influence our comprehension of the past, as well as the significance of subjectivity and interpretation in historical scholarship. All the AI simulations in this exhibition reflect a new research perspective in the study of the past as a ‘multiverse’ rather than an objective and univocal narrative. The idea of multiplying interpretation and how we perceive ancient environments poses new challenges and opportunities: is this an extension of human knowledge or a risk of misinformation?
Training computers
Archaeology and cultural heritage are fields where the application of generative AI presents both promising new opportunities and significant methodological and ethical concerns. Assuring the authenticity and precision of AI-generated reconstructions, which are highly dependent on the quality and representativeness of the training data, is one of the greatest obstacles. Such imperfections could result in distorted or insufficient representations of the past – although it is important to note that this is also something that can be said about human attempts to understand archeological material. AI can extend the range of human interpretation, but the results depend on the training methods. To mitigate these concerns, researchers must ensure that the AI models undergo training using a wide array of dependable data sources, including archaeological, historical, and scientific evidence.
This approach entails furnishing comprehensive details of the data sources, methodologies, and assumptions that underpin the development of the visualisations. Additionally, it entails drawing attention to domains where specialists disagree or are divided. Another crucial ethical factor to contemplate is the potential ramifications that AI-generated reconstructions may have on the represented communities and cultures. When AI is used to reconstruct ancient landscapes and environments, cultural appropriation or misrepresentation could happen, especially if the knowledge and perspectives of indigenous or descendant communities are not fully included in the process.
Notwithstanding these obstacles, the prospective advantages of employing generative AI in the fields of archaeology and cultural heritage are substantial. By facilitating the reconstruction and visualisation of the past in advanced and inventive ways by researchers, AI can challenge conventional notions of expertise and authority, while aiding in the democratisation and pluralisation of historical and archaeological knowledge.
Because AI-generated reconstructions can present a more immersive and engaging way for the public to interact with and learn about the past, the approach promises to increase the accessibility and applicability of archaeological knowledge to a broader audience. In addition, by bringing together specialists from natural sciences, computer science, archaeology, and history, among other fields, the application of AI in cultural heritage and archaeology can facilitate new forms of interdisciplinary collaboration and expertise sharing.
In summary, the AI Rethinks the Past exhibition developed at Duke University provides an intriguing insight into the revolutionary possibilities that generative AI technology may bring about regarding our comprehension and interaction with ancient environments and landscapes. By pushing the limits of what is possible from studying and interpretating the past, exhibitions such as AI Rethinks the Past will have a significant impact on the future of the exciting and swiftly developing field of AI archaeology.
Researchers will ultimately determine the efficacy of AI in the fields of archaeology and cultural heritage by successfully addressing the intricate ethical and methodological obstacles that arise, all the while embracing the opportunities for collaboration and innovation. By encouraging people from different fields to work together and getting people from the communities and cultures that are being represented involved, we can use AI to build a dynamic, all-encompassing, and diverse historical understanding that reflects the complexity and depth of human history and culture.
Shaping virtual environments
The ‘AI Puzzle’, which merges generative AI and 3D-printing technology, is arguably the most inventive and captivating installation developed for the exhibition. The puzzle comprises an assortment of 3D printed elements that symbolise different components of the ancient Roman topography, including structures, roads, buildings, and ecological formations. Participants are encouraged to organise the aforementioned items on a table that is furnished with cameras that record the spatial connections among the items (below). The AI system produces real-time visualisations of the corresponding landscape that are projected on to the walls of the exhibition space as visitors manipulate the objects. The visualisations undergo modifications based on the orientation and positioning of the objects on the table, demonstrating their dynamic and responsive nature. This approach gives rise to an intriguing dynamic between the tangible and digital domains, as individuals actively reshape the virtual environment through their decisions and actions.
Acknowledgments:
• Rethinking the Past is the vision of Maurizio Forte, Director of Dig@Lab (https://diglab.duke.edu) at Duke University. See https://rethinkingthepast.org for more details on the exhibition.
• Exhibition design: Maurizio Forte, Augustus Wendell, Caitlin Childers
• AI Puzzle: Felipe Infante de Castro, Distillery.dev, Maurizio Forte
• Pollen analysis and palaeobotany interpretations: Laboratory of Palynology and Palaeobotany, Department of Life Sciences, University of Modena and Reggio Emilia (Anna Maria Mercuri, Assunta Florenzano, Eleonora Clò, Giovanna Bosi, Paola Torri, Eleonora Rattighieri, Jessica Zappa, Elisa Furia, Cristina Ricucci, Lorenzo Braga)
• Emptyscapes Initiative and archaeological investigations in the site of Roselle: Stefano Campana, Laboratory of Landscape Archaeology and Remote Sensing, University of Siena, Italy
You must be logged in to post a comment.