MICROSTORY OF ART ONLINE JOURNAL FOR ART, CONNOISSEURSHIP AND CULTURAL JOURNALISM On Why Artificial Intelligence Has No Place in Connoisseurship

(12.12.2022) Artificial Intelligence (AI) has no place in scientific connoisseurship yet. It may have a place in other cultures of connoisseurship, as for example astrology may have a place (in case people do trust in such practices), but in the context of scientific connoisseurship AI cannot have a place – yet. Why?

Because: a) in the context of scientific connoisseurship every claim has to be backed by reasons, by good reasons, by a solid reasoning. And AI tools, insofar they do not talk to the scientific community yet, cannot be included into that community. – Suggestions, based on the use of AI tools may be something else, but, again, every claim has to be explained and be backed by a solid reasoning;
b) because AI tools have to be taught what to do, and have to be taught on the basis what humans have yet found out; in Leonardo scholarship for example, we have neither a solid Leonardo da Vinci oeuvre catalogue, not any decent catalogue of the Leonardeschi; if AI tools should manage the distinction of cats and dogs, after having studied millions and millions of pictures of cats and dogs, it is unreasonable to think that AI tools can manage the distinction between Leonardo pictures and pictures by the Leonardeschi, since humans have produced a mess; and on the basis of such mess (with very few pictures anyway), a solid learning process might remain a fantasy, as long as humans have not produced something better as a starting point to learn from.

In addition to that it has to be said that Leonardo oeuvre catalogues have not even begun seriously to address the question of hybridity; any picture, on the basis of which machine learning might take place, might be a composit, a hybrid, brief: the result of many hands (that yet have to be distinguished).

And in addition to that it has also be said that AI tools, insofar such tools are mere pattern recognition systems, might be seen as an aid in the future of scientific connoisseurship (as soon as these tools are able to explain what they do), but also would seem to reduce connoisseurial expertise to pattern recognition. This might be helpful, but connoisseurship, the process of attributing of works of art to its authors, is more. It involves intellectual dimensions, reflections as to art and history interacting and producing meaning and so forth. It is a misunderstanding to think that connoisseurship is simply ›mere looking‹; it is informed looking plus further reflection. And this is something else than mere stylistic observation and mechanical pattern recognition, and it should be kept in mind, that it should not be suggested that AI tools could easily replace the human connoisseur. This is not the case – until AI tools might also manage more intellectual abilities as the interpretation of art in the context of history for example.

And in addition to that it has to be said that AI connoisseurship as well as human connoisseurship faces a logical problem: which is the problem that it, connoisseurship, does work only in the framework of what we already know. The unseen, the unexpected move of an artist, the outstanding work, does, systematically, confuse the system, since we have to compare what we know with something that does not necessarily fit into the parameters of what we already believe to know (but in truth have known not entirely). Human connoisseurship might be able to adjust the parameters of what we know, but a machine has to be taught anew. Future AI might, one day, be able to handle that problem. At the moment we see just dumb machines that not even can tell a restorer from an original artist, but are supposed – as the fuss made about these tools seems to indicate – that connoisseurship is about to be revolutionized.

MICROSTORY OF ART
ONLINE JOURNAL FOR ART, CONNOISSEURSHIP AND CULTURAL JOURNALISM
HOME

Top of the page

Microstory of Art Main Index

Dietrich Seybold Homepage

© DS