AI “art” and uncanniness
🌈 Abstract
The article discusses the complex issues surrounding AI art, including the technical and legal aspects of training AI models, the potential impact on creative workers, and the broader implications for art and aesthetics. The author presents a nuanced perspective, acknowledging the valid concerns of creative workers while also highlighting the potential benefits and limitations of copyright law in addressing these issues.
🙋 Q&A
[01] AI Art and Uncanniness
1. What are the author's views on AI art?
- The author is, on balance, opposed to AI art, but acknowledges that there are important nuances to this position.
- The author finds AI art to be "anodyne, boring" and believes it lacks the intentionality and emotional resonance that characterizes human-created art.
- The author likens AI art to the "spooky-action-at-a-close-up" phenomenon described by Mark Fisher, where the output appears to have a life of its own without stemming from a conscious organizing intention.
2. How does the author differentiate between AI art and human-created art?
- The author argues that art speaks to us at a non-rational, sub-rational level, transmitting "big, numinous irreducible emotions" from the artist's mind to the viewer's.
- In contrast, the author believes that AI art lacks this emotional and intentional quality, as AI systems do not have a mind or agency of their own.
3. What are the author's thoughts on the potential for AI art to be considered "real art" in the future?
- The author acknowledges the possibility that they may look back on AI art and see it as a meaningful art form, similar to how sampling was initially dismissed but later recognized as an important artistic technique.
- However, the author expresses skepticism about the likelihood of this, stating that they "wouldn't give odds on it."
[02] Copyright and AI Training
1. What is the author's perspective on the legal aspects of training AI models?
- The author argues that, from a technical standpoint, training an AI model does not constitute copyright infringement, as the process of making transient copies, performing quantitative analysis, and publishing the model's conclusions are all considered fair use.
- The author acknowledges that there may be instances where a model "memorizes" portions of the training data, which could potentially be infringing, but suggests that this can be addressed by post-processing the model to remove any such memorizations.
2. How does the author view the potential for new copyright laws to address AI art?
- The author cautions that any new copyright laws aimed at regulating AI art would need to be carefully crafted to avoid unintended consequences, such as limiting beneficial activities like scholarly analysis, archiving, and critical commentary.
- The author suggests that even if new copyright laws were enacted, they may not necessarily improve the material conditions of artists, as large corporations that control copyrights could still leverage AI models to their advantage and erode the bargaining power of creative workers.
[03] Alternatives to Copyright for Protecting Creative Workers
1. What alternative approaches does the author propose for addressing the concerns of creative workers?
- The author suggests that focusing on labor law and collective bargaining may be more effective than relying on copyright law to protect the interests of creative workers.
- The author points to the success of collective licensing schemes, which can provide more robust protections for creative workers than individual copyright.
- The author also highlights the potential for the US Copyright Office's stance that AI-generated works are not eligible for copyright protection to create challenges for companies seeking to commercialize such works.
2. How does the author view the role of labor organizing and sectoral bargaining in improving the conditions for creative workers?
- The author argues that rather than lobbying for expanded copyright, creative workers could join forces with workers across industries to demand the right to sectoral bargaining, which could lead to more meaningful improvements in their material conditions.
- The author suggests that this broader coalition-building approach could be more effective than relying on copyright law, which the author believes ultimately benefits large corporations more than individual artists.