At the edge of a generative cultural precipice

Diego Porres,Alex Gomez-Villa

2024-05-01

Abstract:Since NFTs and large generative models (such as DALLE2 and Stable Diffusion) have been publicly available, artists have seen their jobs threatened and stolen. While artists depend on sharing their art on online platforms such as Deviantart, Pixiv, and Artstation, many slowed down sharing their work or downright removed their past work therein, especially if these platforms fail to provide certain guarantees regarding the copyright of their uploaded work. Text-to-image (T2I) generative models are trained using human-produced content to better guide the style and themes they can produce. Still, if the trend continues where data found online is generated by a machine instead of a human, this will have vast repercussions in culture. Inspired by recent work in generative models, we wish to tell a cautionary tale and ask what will happen to the visual arts if generative models continue on the path to be (eventually) trained solely on generated content.

Computers and Society,Machine Learning

What problem does this paper attempt to address?

The core problem that this paper attempts to solve is: **The impact of the widespread use of generative models on artistic creation and cultural diversity**, especially with the public availability of NFTs (Non - Fungible Tokens) and large - scale generative models (such as DALL·E 2 and Stable Diffusion), the work of artists has been threatened. Specifically, the paper focuses on the following aspects:

1. **Changes in artists' behavior of uploading works**:
- Many artists have reduced their work sharing on online platforms such as DeviantArt, Pixiv and Artstation, and even deleted their past works, especially when these platforms fail to provide sufficient copyright protection.
- The study found that the upload volume of senior artists has decreased significantly, while that of young artists has increased.

2. **Quality and diversity issues of training data for generative models**:
- Text - to - image (T2I) generative models rely on human - created content to guide the generation of styles and themes. However, if in the future generative models are only trained based on machine - generated data, this will have a profound impact on culture.
- Currently, large - scale data sets (such as LAION) lack sufficient representation of niche communities (including social and artistic aspects), resulting in generative models that may amplify certain biases.

3. **The impact of generative models on future artistic creation**:
- If generative models continue to be trained only based on generated content, it may lead to "knowledge collapse", that is, the generated content tends to be "average", losing diversity and uniqueness.
- This trend may cause new artists to learn wrong geometric properties or other technical errors, thus affecting their skill development.

4. **Risks of cultural and artistic inheritance**:
- For example, the traditional fabric huipils in Mesoamerican culture, its patterns and colors not only identify the group affiliation of the wearer, but also have personal and symbolic meanings. If these non - two - dimensional techniques are replaced by pixel - generating models, these skills may disappear or be culturally appropriated.

The paper explores the impact of generative models on artistic creation and social culture by analyzing the uploading behavior of artists on different platforms, and raises concerns about future development.