On the occasion of next week's meeting of the minister of cultures at Cyprus we offer here a selection of papers not to forget the importance of how AI is reshaping culture beyond buzzwords, gaining some tangible insight that will allow you to see the world with an effective logic.
Large AI models are cultural and social technologies
Henry Farrell, Alison Gopnik, Cpsma Shalizi et al.
This paper introduces AI as a cultural and social technology, moving beyond the classic opposition of human made vs AI made. LLMs should not be seen as a substitution for human creativity but as part of an ongoing transformation to make human information accessible to the widest possible public.
SIGNIFICANCE deep learning based platform to fight illicit trafficking of Cultural Heritage goods
E.S. Malinverni, D. Abate, A. Agapiou et al.
Not all AIs are born equal. This paper demonstrates how playing to the strengths of different AI systems is essential when fighting illicit trafficking of Cultural Heritage goods. It presents an easy-to-use LLM based interface built on a capable machine learning tool that analyzes thousands of images to identify relevant listings on online marketplaces.
Art forgery detection using Kolmogorov Arnold and convolutional neural networks
S. Boccuzzo, D. D. Meyer, Ludovica Schaerf
Developing and effectively implementing AI are two different paths towards the same result, a more effective and reliable system of checks and balances in the art world.The system was implemented to detect forged paintings, tested on works known to have been counterfeited by Wolfgang Beltracchi.
Algorithmic Ways of Seeing: Using Object Detection to Facilitate Art Exploration
L. Meyer, J. E. Aaen, A. R. Tranberg et al.
This paper explores how object detection AI can unlock museum collections for broader exploration. Still it is important not to forget that computer vision is limited by cultural and gender biases. Concepts that may appear to us simple and straightforward may carry different meanings across time and cultural contexts. A mislabeled artifact is not just an error, it can erase or distort cultural meaning.
Rethinking Cross-lingual Alignment: Balancing Transfer and Cultural Erasure in Multilingual LLMs
H. HyoJung, S. Agrawal, Eleftheria Briakou
Complementing the previous paper, this research names and formalizes the encounter between AI and culture through the concept of cross-lingual alignment (CLA). The final goal is to align multilingual representations so that LLMs can seamlessly transfer knowledge across languages, while being “aware” of the culturally situated nature of responses in relation to the language in which the question was asked. This paper in particular should highlight why "multilingual AI" is not the same as "culturally equitable AI”.