Seizing the Means of Production: Exploring the Landscape of Crafting, Adapting and Navigating Generative AI Models in the Visual Arts

Overview

• This paper explores the creative possibilities for visual artists in using deep generative models, including the design and manipulation of these models as a creative activity in itself.

• The authors argue for revisiting "model crafting" - the process of designing and manipulating generative models to achieve creative goals - as a valuable area of study and practice for artists.

Plain English Explanation

• The paper provides a roadmap of the different ways visual artists can leverage advanced AI models to create personal artworks. This includes not just using pre-trained models, but also crafting, adapting and navigating deep generative models.

• One key idea the authors highlight is the concept of "model crafting" - the ability for artists to actively design and modify generative AI models to suit their creative goals. They argue this should be seen as a creative practice in its own right, worthy of further study and exploration, rather than just a technical process.

• By understanding and shaping human-technology assemblages in the age of generative AI, artists can push the boundaries of what's possible with these powerful AI tools, moving beyond just using them as black boxes.

Technical Explanation

• The paper provides a comprehensive overview of the different approaches visual artists can take to leverage deep generative models, including using pre-trained models, fine-tuning them, and even designing custom generative AI architectures.

• A key focus is on the concept of "model crafting" - the ability for artists to actively design and manipulate the underlying structure and training of generative models to achieve specific creative objectives. This goes beyond just using models as fixed tools.

• The authors make the case that model crafting should be viewed as a creative practice in its own right, akin to other artistic disciplines. They suggest studying the cognitive and creative processes involved, as well as developing automated ML tools and workflows to support model crafting.

Critical Analysis

• The paper provides a well-reasoned argument for exploring model crafting as a valuable creative activity, but acknowledges that this is a relatively new and underexplored area. More research is needed to fully understand the cognitive and creative processes involved.

• Some potential limitations include the fact that model crafting may require a high degree of technical expertise that could exclude many artists. There are also open questions about the extent to which generative models can truly be "shaped" by humans versus the inherent biases and limitations of the underlying technologies.

• Additionally, the paper does not delve into the ethical considerations of using generative AI in architectural design and other creative domains, which will be an important area for future research and discussion.

Conclusion

• This paper makes a compelling case for revisiting model crafting as a valuable creative practice for visual artists working with deep generative AI technologies. By actively designing and manipulating these models, artists can unlock new possibilities for personal expression and push the boundaries of what's possible.

• Exploring model crafting as a creative discipline in its own right could lead to important insights about the interplay between human and machine intelligence in the creative process. This has significant implications for the future of art, design, and other creative fields as generative AI becomes more ubiquitous.