|
The survey first aims to acquire the profile of the artist, such as the medium they work in and their online presence or lack thereof. It then asks whether or not artists use any currently available protection on their art and why or why not. The survey was run over the period of a month, and received 39 responses. Figure 1. shows the percentage of people who used available protective resources, 25.6%, compared to the percentage who did not, 69.2%. Alongside this, there were two AI users, accounting for the remaining 5.1%. Figure 2. represents the protection methods chosen by artists who answered “Yes” on the previous figure. Some artists used multiple methods, however there were a total of 11 respondents. The majority of respondents used Glaze. Despite this, only 3 people utilized Nightshade, created by the same group as Glaze. Two artists rely on Cara, which is a portfolio sharing platform for artists that prohibits the sharing of AI-generated images. Similar to Artstation, Cara uses the “NoAI” tag that notifies AI scrapers to not scrape tagged images, but this system is built on the honor code, meaning that AI scrapers are able to use tagged images regardless. At last, a number of artists also chose to not upload their work, preventing any AI scrapers from reaching it in the first place. Figure 3. represents the reasons listed for not using protective resources by artists who selected “No” in figure one. Some artists listed multiple reasons, however there were a total of 26 respondents. First of all, some artists do not know where to find protective resources. Others do not have the available income to afford protection. Some stated that they tried to use resources, but they caused technical difficulties: “I tried to use a protecting tool twice (I don't remember the name), but it crashed my computer and the artworks they put as examples didn't [have] noticeable changes, so I thought it didn't work well.” Other artists also saw a worsened image quality and differences within artworks. Another inconvenience is that it is another process in the step, with one respondent saying they would use these resources “if it was automatic on export from [Adobe] Lightroom.” Lastly, most answered in an indifferent fashion, whether that be apathy towards their works being used in AI or a pessimistic outlook on trying to defend themselves from AI, leading to non-action. Survey responses include “It is a futile effort. All information can and is accessed by someone with [the] ability to do so,” and “...there's a plethora of content they can scrape already, mine literally won't make a difference.” These indicate an idea of helplessness throughout artists dealing with generative AI, that their work is a drop in an ocean of other content. Contrasting this, another says “I don't care if my art is used in training datasets.”
1 Comment
With certain methods, data can be removed and forgotten by a machine, called machine unlearning. Largely, this has been theorized for the purpose of correcting models which are biased or for security. In this context, however, it can be useful for removing large portions of copyrighted artwork, through comparisons of AI-generated images and copyrighted non-AI-generated art. This can be done through several methods, such as training non-copyrighted works of a similar nature or removing the influence of copyrighted works on the model. Despite this, the main goal of unlearning is to retain as much model usability as possible, which would still mimic and regurgitate artists’ styles. The same issue of contacting all artists whose works are in a model’s dataset would also be relevant here, as creative works created recently are technically protected by copyright by default (though this is not the same as registering ownership). To continue, it has been shown that despite a concept being erased from a model, it’s still possible to generate something directly related to that concept. Machine unlearning technology is still incredibly finicky as of now, with many people receiving mixed results.  |
RSS Feed