A follow-up to our previous article, this work examines AI applications that are often overlooked by the media in favor of its e-commerce, health, and advertising darlings. Here, we specifically look at new and atypical AI trends in the art world.
In the past century, a number of masterful forgeries have been wildly successful, sending shock waves through the art world upon their exposure.
Detecting them by eye, based on minute nuances in brushstroke and image details, is a difficult task, even for the shrewdest collectors, auction houses, and galleries. As a result, buyers have spent vast sums — thousands and even millions of dollars — on fakes, advertised as the lost works of old masters like Chagall, Holbein, and Vermeer.
But recent research shows that AI poses powerful new recourse. The study’s deep recurrent neural network was trained on 300 line drawings by famous artists like Picasso, Modigliani, and Matisse. From them, it was able to identify 80,000 individual strokes and their indicative features.
The creators, a team of Rutgers scientists, also created a similarly functioning machine learning algorithm, which learned to identify artists based on stroke weight in their paintings and drawings.
The systems were then tested on commissioned forgeries. In each instance, they were able to detect inauthenticity with complete accuracy — needing only to examine a single stroke.
It belongs in a museum
Worldwide, museums of all sizes — from the MET and MoMA to Rio de Janeiro’s Museum of Tomorrow and the digital Barnes Foundation — have integrated AI into their visitation, art classification, and operations systems.
And despite some belief to the contrary, many museums are continuing to explore novel applications of AI in their sphere. This has led to such projects as Seoul’s Robot Science Museum and the Musee du quai Branly’s AI-based art critic.
In the experiment, the AI art critic named Berenson wandered the Paris museum’s glass-paneled galleries, recording and categorizing visitors’ reactions to the artwork and using them as visual feedback to create preferences of its own. Based on its newly formulated attitudes, it either smiled or frowned at the artwork.
In a fully integrated project, South Korea’s robots are reportedly leading the construction of Seoul’s ovoid robotics museum. The various types of robots will guide each phase in the museum’s latent history, from design and construction to service. The work will become content for the museum’s meta curriculum on AI, robotics, and technology.
AI in visual art
With the recent years’ burst in AI art productivity and interest, many new trends have appeared. Some of those, such as general AI exhibitions and the widespread use of generative adversarial networks (GANs) and CANs creative adversarial networks (CANs), have been criticized as non-specific, mimetic, and at times, gimmicky.
Many artists attempt to avoid repetitiveness and obsolescence by staying abreast of the field's ever-evolving technologies.
But uniquely, some artists have embraced the redundant and endlessly representational aspects considered inherent to the field, in attempts to explore its mimetic limit.
Indeed, decades before GANs ushered in today’s wave of artificial intelligence art, artist Harold Cohen developed the AI program Aaron, which he iteratively trained to new artistic techniques, in a 48-year collaboration that lasted until his death in 2016.
The AI, made of about 1.5 megabytes of LISP code, learned to produce art in his style, not by looking at examples of his artwork, but by implementing a set of algorithmic rules and forms through which Cohen essentially codified his drawing processes.
Today’s luminary in the field, Helena Sarin, trains her GAN on self-made artwork, rather than the large web-based datasets widely used by GAN artists. The work of Sarin and Cohen both show individualism and personal style to a much greater degree than is common in AI art — characteristics they’ve attained through hyper-conscientious processes of iteration, refinement, and selective curation.
Trading in robot art
AI art, with its growing adoption and critical reception in the art world, is being traded in new ways. In addition to its 2018 debut on the world auction stage, the art form also entered the digital realm of cryptocurrency, or cryptoart, this year.
Throughout its history, verifying the authenticity and provenance of digital art has been a significant challenge; it is easily shared, endlessly reproducible, and therefore, quick to lose its value.
But, for many digital art creators and buyers, cryptoart is providing a solution. Like Bitcoin, it is traded via blockchain, a secure digital platform that acts as a ledger. Each piece on the platform is verified, given an indelible digital signature, or Non-Fungible Token (NFT), and tracked throughout its trading history. As a result, the pieces can be easily traced back and accredited to their originators.
AI artists, gif creators, digital illustrators, and virtual reality artists have used the platform to maintain and sometimes increase the value and scarcity of their art, as well as their visibility to art collectors.
While AI has met varied responses in the art world, it is undoubtedly finding a growing role in the field, shaping everything from the way art is made to the way it is sold and displayed.