Marble, polyamide, machine learning algorithms, custom software, original dataset,
multi-channel video installation;
Project Film: https://vimeo.com/egorkraft/casfilm
The Content Aware Studies series initiates an inquiry into the possibilities AI and machine learning technologies hold, both as tools for speculative historical investigation and means of emerging aesthetic formation. The process, developed for over a year now by an artist together with a data scientist engaged in training artificial neural networks, replenishes lost fragments of sculptures and friezes of classical antiquity and generates never before existing, yet authentic objects of that era. The research examines how our custom developed AI, utilizing the largest recent advancement in computer
vision and cognition, operates when trained on datasets consisting of thousands of 3D scans of classical sculptures from renowned international museum collections (i.e., British Museum, Metropolitan, National Roman Museum etc.). The algorithm generates models, which are then 3D printed in various synthetic materials, filling the voids in the eroded and damaged marble sculptures. Some of these algorithmic outputs are turned into new entirely marble sculptures uncanny in their algorithmic integrity. They render the work of synthetic agency that lends a faithful authenticity to the forms, while also producing bizarre errors and algorithmic normalizations of forms previously standardized and regulated by the canon of Hellenistic and Roman art.
Recent research in General Adversarial Networks (GANs, a class of machine learning systems) has shown outstanding results in hyper realistic image rendering. The technology is already in use for both investigation of historical documents (ex., Voynich Manuscript) as
well as predictive instrument for modelling futures. However, we might want to critically examine a role of such form of knowledge production beforehand; how do we distinguish between accelerated forms of empirical investigation and algorithmic bias? Will this question survive, when such forms of knowledge production become ubiquitous governing agencies?
The work examines questions and topics of bias, authenticity, materiality, automation, authorship, knowledge and history. It inspects what visual and aesthetic qualities for such guises are conveyed when rendered by synthetic agency and perceived through our anthropocentric lens. What of our historical knowledge and interpretation, encoded into the datasets will survive this digital digestion? It examines new forms of historical knowledge and artistic production and calls into question the ethical implications of such approaches in relation to culture and the notion of endangered anthropocentric world.
CONTENT AWARE STUDIES SERIES
Marble, polyamide, machine learning algorithms, custom software, original dataset, multi-channel video installation;
Machine learning assistance: Artem Konevskikh
Materiality has reappeared as a highly contested topic in recent art. Modernist criticism tended to privilege form over matter — considering material as the essentialized basis of medium specificity — and technically based approaches in art history reinforced connoisseur- ship through the science of artistic materials. But in order to engage critically with materiality in the post-digital era, the time of big data and automation, we may need a very different set of methodological tools.
We may need to address digital infrastructures as entirely physical and to reexamine the notion of “dematerialization”, by addressing materialist critiques of artistic production, surveying relationships between matter and bodies, exploring the vitality of substances; and looking closely at the concepts of inter-materiality and trans-materiality emerging in the hybrid zones of digital expe- rimentation.
The image below is a result of interpretation of antique portrait by general adversarial neural network based on the analysis of nearly 10,000 3D scans. Custom created dataset included 3d scans of sculptures from the collections of Metropolitan Museum, Hermitage, British Museum, National Museum of Rome and other world renowned collections of antiquity;
Testing Out Sticky Guy
Hellenistic Portrait [Content Aware Studies Series]
Content Aware Studies exhibition at Alexander Levy gallery, Berlin, Germany.
Content Aware Studies exhibition at Alexander Levy gallery, Berlin, Germany.
Some Blah blah blah here etc
Fragment of sculpture CAS_08 Hellenistic Ruler
Linux based server equipped with multiple GPUs performing gen- eral adversarial machine training during the exhibition ‘Conent Aware Studies’ at Alexander Levy Gallery, Berlin, Germany.
‘Wonderful exhibition of Egor Kraft, one of the best artists working with AI - Anna Nova gallery, Saint-petersburg, Russia. His series of sculptures uses machine learning trained on real clasical sculptures missing some parts. The networks reconstructs these parts resulting in delightful and friendly fantastical creatures. The results are created from real marble. This meeting of classical high culture and latest technologies is one of the things making this work unique.’
– Lev Manovich
author of books on new media theory, professor of Computer Science at the City University of New York, Graduate Center, U.S. and visiting professor at European Graduate School in Saas-Fee, Switzerland.
CAS_08 Hellenistic Ruler; 2018
Marble, Polyamide; Machine Learning Algorithms Dimensions: 19x26x21;
Courtesy of the author & Anna Nova Gallery
CAS_09 Colossal head of Hercules; 2018
Marble, Polyamide; Machine Learning Algorithms Dimensions: 24x32x20;
Courtesy of the author & Anna Nova Gallery
One of the first friezes form the series originates from two marble blocks of Parthenon Frieze joint by machine learning generated fragment (in the middle), in which machine suggested to merge two horses into a single creature with many feet;
CAS_04 Parthenon_South_XI_31; 2018
Carrera Marble, Machine Learning Algorithms Dimensions: 120x100x10cm;
Technical and artistic assistance: Matthew Lenkiewicz Courtesy of the author
CAS_07 frieze originates from dataset based on Pergamon and Telephos friezes datasets;
CAS_07 Telephos Frieze; 2018
Botticino marble, machine learning algorithms 67 x 94 x 10 cm
Egor Kraft – Chinese Ink Generative Installation Electronic ink screens, custom software, machine learning algorithms, liquid cooled server Technical support: Artem Konevskih, Anna Demidova Hermitage Museum, St. Petersburg Recent studies in the field of artificial intelligence (in particular generative adversarial networks) have demonstrated outstanding results in the synthesis of hyper-realistic imagery [e.g. neural network Style Gan 2, https://arxiv.org/abs/1912.04958]. Along with the rendering of photorealistic images, data scientists and machine vision specialists have demonstrated extraordinary capabilities of aforementioned class of artificial neural networks in simulating artistic techniques and style transfer. The degree of quality and accuracy of those algorithmic outputs strikes imaginations. These developments and the emerging prospect of their further applications and calibrations do pose new challenges in the fields of media, journalism and, of course, artistic production, raising a set of new aesthetic issues. The work Chinese ink is meant to facilitate an inquiry into traditional Chinese ink calligraphy technique; my interest is not related to the imagery, visual style or iconography in Chinese painting, but rather to the physical properties, material specificity and historical connotations of industrialisation of technic itself as seen through the lense of the present context. The way in which the ink behaves - as a material produced from soot and glue of animal origin or sometimes graphite based mineral types, in contact with a special coarse-grained and pre-moistened paper, still remains superior in certain qualities as opposed to European inks. Radiating black lacquer sheen Chinese inksticks are rubbed and diluted with water to thick or thin liquid consistency, which allows to achieve a wide range of shades of black and grey, such depth and tone richness had hardly been achieved with European inks. In China the ink is considered the cult of tradition and state of the art technology. The work Chinese Ink visually examines applications of generative-adversarial network in synthetic simulation of original technique. The neural network is being trained on thousands of ink blots images – a dataset, especially developed for the project; An involved machine is capable of rendering thousands of images per minute, similar to those it analysed in the dataset, yet each being unique in its algorithmic authenticity. The resulting images are generated in real time and sequentially displayed via the means of electronic ink displays. The latter play a crucial role in conveying politics between visual and conceptual contents of the work. Electronic paper (also e-paper, electronic ink or e-ink) are display devices that mimic the appearance of ordinary ink on paper. Unlike conventional backlit flat panel displays that emit light, electronic paper displays reflect light like paper, involving particles. These hard pigmented grains distributed across micro-structural material. There is no surprise that such displays are fabricated in modern China, – the country which also occupies one of the leading positions in application, development and research in the fields of machine learning and AI. The economic, political and industrial conditions of neoliberal globalisation under which the above-described technologies are developed in modern China regulate another pace and purpose as opposed to those, at which production and application of traditional ink technology was maintained for centuries. How such conditions reconvey visual aesthetic qualities? - an issue raised through this work; it suggests to traceback a chain of links between tradition, technology, time, economies and techno-political processes leading to automation and new emerging aesthetics and tools that enable them on a material level. The main focus of the research around the work is preoccupied with the processes of formation of algorithmic aesthetics and their links to anticipated them traditional visual languages. E. Kraft
↑ Back to top
Copyright © 2005 – 2020. All rights reserved.
Site build 24 September 2020 at 5:51am