4 NOVEMBER TALK: KATHARINE BUTLER AND TUO ZHANG ‘CAN AI UNDERSTAND ART? RECENT RESEARCH USING VISION LANGUAGE MODELS ON CHINESE PORCELAIN’

TALK: CAN AI UNDERSTAND ART? RECENT RESEARCH USING Visual Language Models ON CHINESE PORCELAIN
KATHARINE BUTLER AND TUO ZHANG

MONDAY 4 november
3:00 pm
AT ASIA HOUSE, 63 NEW CAVENDISH STREET, LONDON W1G 7LP

LIMITED PLACES AVAILABLE

Large vision-language models (VLMs) have demonstrated remarkable abilities in understanding everyday content, however, their performance in the domain of art remains less explored. Evaluation of art, particularly in regard to its authenticity, can be approached from art historical, scientific and archival / provenance methods. This talk will discuss the advantages of applying AI to the data generated from such research and, in particular,  we will demonstrate how our VLM interprets the decoration on 3D objects, such as Chinese porcelain. Relying on Dr Ni Yibin’s vast database of images found on Chinese art objects of all media, we have focussed on three primary tasks: identifying salient visual elements, matching these elements with their symbolic meanings, and interpreting the intended messages. We will discuss the added complexity incurred by the Chinese language and the use of “pun rebuses” in its art and culture. Finally, we will show how our AI assisted programme could support museums, dealers and collectors in the future.

Katharine Butler is an entrepreneur turned researcher in Chinese porcelain. She has an MA in History of Art from Edinburgh University and, from the 1990s, worked closely with Sir Michael Butler, cataloguing and researching his extensive and comprehensive 17th-century porcelain collection. She is co-author of Leaping the Dragon Gate: The Sir Michael Butler Collection of Seventeenth-Century Chinese Porcelain (Ad Ilissvm, London, 2021). She has published articles in Arts of Asia, The Asian Art Newspaper and Transactions of the Oriental Ceramics Society, of which she has been a Council member, and has lectured widely.

Tuo Zhang is a Ph.D. Candidate in the Electrical Engineering department at the University of Southern California. Before joining USC, he received his Bachelor of Science with High Honor from the University of California, Santa Barbara. His research focuses on distributed / federated machine learning algorithms, human-centered machine learning applications, and Large Language Model understandings. In his recent work, he offered a new perspective on how AI reads and interprets Asian Arts.