The Importance of Curation in AI Art
Balancing Creative Freedom and User Experience
Introduction
As AI art continues to evolve and become more accessible, the importance of curation in this space cannot be overstated. With the rise of platforms like Civitai, creators are now able to share their work with a global audience, but this increased visibility also brings its own set of challenges. One of the most pressing issues facing the AI art community is the need for responsible curation, balancing creative freedom with user experience.
The Problem with Hodgepodge Curation
The issue with excessive mass style versioning in one model card is that it can be overwhelming for end users to navigate. When using a GUI like Automatic 1111 or even the Civitai browser, it's easy to get lost in a sea of unrelated styles and concepts. I'm not alone in this; I've made the same mistake with my own embeddings due to misunderstandings of the system. While we have the freedom to do as we please, it's essential to consider the user experience. Multiple versions that lack cohesion can lead to two outcomes: either they get buried and receive fewer downloads, or they receive tons of downloads, but users may question the creator's motives. As artists, we're in this for the creative journey, not to game the system. However, as someone with a graphic design background, I believe it's possible to upload multiple styles at once while still considering how to market them effectively. It's about finding a balance between creative freedom and user experience.
The Frustration of "ARTIST CLAIMING"
As AI art continues to evolve, the issue of artist claiming has become a growing concern. With the rise of ethically varied AI art, copyright laws are being pushed to their limits. Unlike music and movies, art styles themselves are not copyrightable, but frustrated independent artists may claim ownership of entire sets of models, even if they don't belong to them. False DMCA claims are illegal, but third-party platforms like Civitai are often forced to navigate these disputes, leading to a complex and frustrating situation.
The question remains: who owns what? As creators, we train on various datasets, but what happens when someone tries to claim ownership of our work? For instance, I've trained on multiple versions of my own art, but someone could potentially claim it's not mine. They might even try to take down all the Loras based on my art, citing copyright infringement. Would I stand a chance in court? Probably not, even if I could prove it's my art and I trained it. The problem lies in the fact that art styles can't be copyrighted, and while Stable Diffusion may not be illegal, it's often viewed as unethical.
So, how do we navigate this complex issue? It's not just about hodgepodge model cards; it's about all of us. I've seen excellent trains that have been partially claimed, and it raises questions about where we draw the line. Is it ethical for someone to use someone else's art, especially if it's marked "NO SELLING IMAGES"? What about fair use? If someone took claim over my own art, where would I stand?
Ultimately, we need to be aware of these issues and exercise caution. As creators, we must be mindful of the ethical implications of our actions and strive to find a balance between creative freedom and respect for others' work.
The Impact of Excessive Versioning
Another issue facing AI art platforms is the problem of excessive versioning. When creators release frequent updates and versions of their models, it can lead to a lack of cohesion and user experience. This can be particularly problematic when models are dumped into a single collection, making it difficult for users to navigate and find what they're looking for. Furthermore, excessive versioning can lead to a sense of fatigue, as users become overwhelmed by the sheer volume of content.
However, it's essential to strike a balance. While 100+ updates to a singular model might be seen as excessive, it's not our place to judge. I've had models that split into 4 styles, which eventually made no sense on Stable Diffusion 1.5. So, how do we navigate this in the grand scheme of things? The volume of content is a problematic terminology here, as we're all contributing to the overwhelming amount of content. The real issue lies in the impact on the future of AI art, where experienced creators can guide new users with educational resources.
With my graphic design background, I believe it's time to create articles on "DESIGN THEORY FOR AI." While I don't have a teaching background, it's clear that these versioning issues are frustrating middle ground problems between leaderboard visibility and assumed spam. There's a need to smooth out the process, and while we can't fix everything, we can take steps towards creating a more cohesive and user-friendly experience.
Gaming the System
I've noticed that there's been a surge in new content, which some community members believe may be attributed to mass updates, versions, and potentially unethical practices aimed at staying visible within the leaderboard. While I can't claim to know the inner workings of the platform or the intentions of individual users, it's clear that many people are frustrated by the lack of clarity around these issues. The presence of known botters only adds to the complexity of this topic. Allegedly, this approach has been employed by some creators, but it's essential to acknowledge that this is a community opinion and not necessarily a fact. Regardless, this perceived behavior not only frustrates users but also erodes the trust and credibility of our community. As we move forward, it's crucial that we find ways to encourage responsible curation practices and prevent this type of behavior from undermining the platform's integrity.
Potential Solutions
To address the issues facing our community, several potential solutions can be explored. One approach is to implement a bulk notification system for new releases, allowing users to stay up-to-date with the latest content without feeling overwhelmed. Another solution could be to introduce a new model-collection feature, enabling creators to showcase related models together in a visually appealing way. Additionally, reevaluating the way models are sorted and ranked on leaderboards could help promote more responsible curation practices.
Furthermore, we could explore innovative ways to educate our community about design theory for AI, providing valuable insights and resources to help creators navigate the complexities of AI art. By fostering a culture of creativity, safety, and responsibility, we can create a more engaging and appreciative community.
Conclusion
In conclusion, the importance of curation in AI art cannot be overstated. As our community continues to evolve, it's essential that we find a balance between creative freedom and user experience. By promoting responsible curation practices and addressing the challenges facing our community, we can create a more vibrant and supportive environment for all.
Let's come together to foster a new era of AI art, one that values education, ethics, and creativity. We don't need a three-year degree to get started, and we certainly don't need to rely on misinformation from social media. In fact, we can learn from the mistakes of the past, such as the "AI poison" tools like Glaze and Nightshade, which were ironically created by educational outlets and are simply programs that are AI machine learning tools themselves. We can become the universal education that gets studied, where knowledge isn't gatekept and is accessible to all. Let's stop worrying about high and low culture and foster a better future. Ethics is ethics, but we don't have to muddy ethics with emotion and personal feelings. Moreover, we must recognize that copyright in 2024 is largely gatekept by rich companies that want to control content for their own interests. While corporate control may be necessary in some cases, it's not always a good thing, especially with the onset of AI lawsuits. Let's prioritize our creative freedom and work together to create a better future for AI art. Whatever your goal is, make it an amazing one. I believe in you, and together, we can make this awesome.
Disclaimer
I want to make it clear that the views expressed in this piece are my own and do not reflect the opinions of Civitai or its staff. I am not a member of the Civitai staff, but rather a singular individual with a passion for AI art and a desire to spark meaningful conversations.
I also want to acknowledge that I have Dissociative Identity Disorder (DID), which means that I may express myself in different ways or use different voices (or an LLM to rewrite my poorly thought out text). However, I assure you that the thoughts and opinions shared here are my own and not intended to represent any other individual or entity.
This piece is not targeted at any specific person or group, but rather aims to address ongoing issues and concerns within the AI art community. I believe that open and honest discussion is essential for growth and progress, and I hope that my words will inspire constructive dialogue and reflection.
Thank you for taking the time to read my thoughts, and I look forward to hearing your perspectives as well!
About & Links
About Us
We are the Duskfall Portal Crew, a DID system with over 300 alters, navigating life with DID, ADHD, Autism, and CPTSD. We believe in AI’s potential to break down barriers and enhance mental health, despite its challenges. Join us on our creative journey exploring identity and expression.
Join Our Community
Website: End Media
Discord: Join our Discord
Backups: Hugging Face
Support Us: Send a Pizza
Patreon: https://www.patreon.com/earthndusk
Community Groups:
Subreddit: Reddit
Extensions
ADetailer: ADetailer GitHub
Usage: Use this extension to enhance and refine images, but use sparingly to avoid over-processing with SDXL.
Batchlinks: Batchlinks for A1111
Description: Manage multiple links when running A1111 locally or on a server.
Addon: @nocrypt Addon (The link is broken for now i'll find it later OOPS)
Additional Extensions:
Backups for Loras on SDXL & Pony XL:
2024: https://huggingface.co/EarthnDusk/SDXL_Lora_Dump_2024/tree/main
2023: https://huggingface.co/EarthnDusk/Loras-SDXL/tree/main


.jpeg)