Scrapblog

My conversation with ChatGPT

Can you think of any court precedent where automatic algorithmically generated digital derivatives of visual artworks were judged to be transformative and not infringing the artworks of original input images?

I am not aware of any specific court precedent where automatic algorithmically generated digital derivatives of visual artworks were judged to be transformative and not infringing the artworks of original input images. However, this is an area of law that is still being developed and there are likely to be future court cases that will address this issue. In general, whether a particular use of an copyrighted work is considered to be a fair use, which would not infringe on the copyright, would depend on a number of factors, such as the purpose and character of the use, the nature of the copyrighted work, and the amount and substantiality of the portion used in relation to the copyrighted work as a whole.

You mention "the portion used in relation to the copyrighted work as a whole". I can think of music sampling as an example of small pieces of media which were variously judged to be infringing, or not infringing. Can you think of any other similar examples of marginal cases where the amount of copyrighted work included in the new work was a deciding factor in court outcome?

Yes, music sampling is a good example of a case where the amount of copyrighted material used in a new work was a deciding factor in a court outcome. In some cases, the use of a small sample of copyrighted music in a new work has been found to be infringing, while in other cases it has been found to be a fair use. Another example where the amount of copyrighted material used in a new work was a deciding factor in a court outcome is in the case of Google Books. In this case, Google scanned millions of books without permission from the copyright holders and made snippets of the books available for search. The Authors Guild, which represented authors, sued Google for copyright infringement. The case went to the court of appeals and the court found that Google's use of the copyrighted material was fair use because it was transformative, and the amount of the copyrighted work used in relation to the copyrighted work as a whole was minimal. In addition, in the field of journalism, the use of copyrighted quotes or images in news articles is often considered fair use as long as it is limited to the purpose of commentary, criticism, or news reporting. These are just a few examples, but it's important to note that the outcome of each case would depend on the specific facts and circumstances and the court's interpretation of the law.

Your Google Books example is interesting because there's no human author choosing which passage to quote. In that way it resembles the "algorithmic collage" of AI Transformer networks like Stable Diffusion, ChatGPT, Midjourney, and automatic machine translation systems. Can you think of any more examples like that, where a machine is sampling or quoting copyright material without human intervention?

There are a number of examples where machines are sampling or quoting copyrighted material without human intervention, some of which are similar to the Google Books case you mentioned. One example is the use of content scraping, where a machine automatically extracts and republishes content from a website without the permission of the copyright holder. This has been a contentious issue in the context of online news aggregation, where some publishers argue that aggregators are profiting from their content without permission. Another example is the use of automated text generation software, such as GPT-3, which generates text based on a given prompt or input. This has led to concerns about the potential for the software to generate text that infringes on copyrighted material. In the field of music, there are examples of generative music software that can compose new music based on input samples or algorithms, which might raise copyright concerns if it generates songs that are substantially similar to existing copyrighted songs. In the field of visual art, there are examples of generative art software that can create new images based on input samples or algorithms, which might raise copyright concerns if it generates images that are substantially similar to existing copyrighted artworks. All of these examples, as well as the one you mentioned, are cases where a machine is sampling or quoting copyrighted material without human intervention and they raise legal questions that have yet to be fully addressed by courts. The interpretation of copyright law in these cases will likely depend on the specific facts and circumstances, as well as the court's interpretation of fair use and the transformative nature of the work.

It's interesting that all of the Transformer AI examples you mention tend to use much smaller samples than Google Books. In that way, In that way, the courts might side with AI image generators in lawsuits about their copyright status. But on the other hand, Google Books arguably helped to sell more books, helping the very authors who were suing them. It's much less clear if transformer image generators are beneficial or harmful to existing skilled artists overall. Which of these factors do the courts tend to consider more important when deciding cases? The size of the sample, or the financial impact on existing industries?

The courts tend to consider a variety of factors when deciding cases involving the use of copyrighted material, and the relative importance of each factor will depend on the specific facts and circumstances of each case. The size of the sample used is certainly one factor that the courts will consider when determining whether a use of copyrighted material is a fair use. In general, the smaller the sample used in relation to the copyrighted work as a whole, the more likely it is to be considered a fair use. However, the size of the sample is not the only factor, and courts will also consider the other fair use factors such as the purpose and character of the use, and the nature of the copyrighted work. The financial impact on existing industries is also an important factor that the courts will consider when determining whether a use of copyrighted material is a fair use. In general, if a use of copyrighted material harms the market for the original work, it is less likely to be considered a fair use. However, if a use of copyrighted material helps to promote the original work or creates new markets for it, it is more likely to be considered a fair use. For example, in the Google Books case, the court found that the use of copyrighted material in relation to the copyrighted work as a whole was minimal and that the use was transformative, in that it provided new information and new insights. Also, it helped to promote the original work and create new markets for it, such as by making it easier for researchers and other users to discover and purchase books. In the case of AI generated images, it's likely that the court would consider the degree of creativity of the AI-generated images, the degree of similarity to the original, and the intent of the use as important factors. However, since this is an area of law that is still being developed, and it is uncertain how the courts will rule in specific cases involving AI-generated images. It's important to note that fair use is determined on a case by case basis, and each case will depend on its specific facts and circumstances.

Thank you, this has been an interesting and illuminating discussion.

You're welcome! I'm glad I could help. Copyright law can be a complex and nuanced area, and the use of copyrighted material by AI systems raises unique questions that are still being explored. If you have any more question, feel free to ask.