AI and the art world: New opportunities unfold

Earlier this month I listened to an interesting panel discussion about the relationships between the art world and the rise of AI. The event was organised in partnership with AWITA and the launch of the book by Jo Lawson-Tancred “AI and the Art Market” which offers a comprehensive introduction to the potential impacts of artificial intelligence on the art world. 

The book examines AI’s role in two key areas: as a business tool for processes like authentication and valuation, and as a medium for creating art. Some key points emerged in the book and as part of this ongoing discussion:

1. Industry practices: AI can enhance art market operations but faces challenges such as data scarcity, quality, and the "black box" nature of its algorithms. Transparency and using AI to complement human expertise, rather than replace it, are advocated as effective approaches.
2. AI as an artistic medium: The book explores how galleries are experimenting with AI to create art while addressing challenges like copyright issues, ethical concerns, and the technical maintenance of AI-generated works.
3. Ethical and practical limitations: The text addresses unresolved questions, including the ownership of AI-generated art and the use of data sets for training AI models. These issues highlight the need for clearer guidelines as AI’s influence grows.

Lawson-Tancred combines her expertise in journalism, machine learning, and the art market to provide a balanced, accessible entry point into this emerging field. Her book is timely, given the increasing presence of AI in art auctions and exhibitions, and serves as a guide for navigating the complexities of integrating AI into the art world.

From the art market perspective, there have been recent headlines about the work created by AI at auction. The AI art robot AIDA achieved a significant milestone in the art world. A portrait titled "AI God. Portrait of Alan Turing (2024)", created by the humanoid AI artist Ai-Da, was presented at a Sotheby’s auction. The work was initially estimated to sell for between $120,000 and $180,000. However, it exceeded expectations, reportedly fetching a record-breaking $1.08 million at Sotheby’s Digital Art Day auction in New York on November 7th, making it one of the highest-priced AI-generated artworks sold to date.

Ai-Da, designed with advanced AI and robotics, uses cameras, algorithms, and a robotic arm to create artwork. The sold painting honours Alan Turing, reflecting on his contributions to AI and raising philosophical questions about the interplay between technology and creativity. The sale symbolises a turning point in the acceptance of AI-generated art, fostering debates about authenticity, creativity, and the evolving definition of art in the digital age. 

The first AI-generated artwork sold at auction was Portrait of Edmond de Belamy, created by the Paris-based collective Obvious. This piece was auctioned at Christie’s New York in October 2018, fetching an impressive $432,500 despite an initial estimate of $7,000–$10,000. The portrait is part of a series depicting a fictional "Belamy family," created using a Generative Adversarial Network (GAN). The GAN was trained on a dataset of 15,000 historical portraits spanning the 14th to 20th centuries. 

The work stirred debates about the role of AI in art, as critics highlighted its technical simplicity and questioned the originality of the algorithm used. Obvious acknowledged that their approach was built on existing AI models and technologies, particularly drawing from prior innovations by other AI researchers. 

This landmark sale demonstrated the growing interest in AI art and marked a pivotal moment for machine-created works entering the mainstream art market.

The integration of AI into the art world has sparked significant debates around artists' rights, focusing on issues of copyright, authorship, and creative agency. In most jurisdictions, copyright laws require a human creator. This raises questions about AI-generated works: Should the programmer, the user of the AI, or the AI itself be considered the author? For example, U.S. law specifies that only human-made works can qualify for copyright.

AI tools often generate content by analysing and reinterpreting vast datasets of existing works. This leads to concerns about whether such outputs constitute derivative works and whether the original creators of the training data should be compensated.