The Recent Development of AI tools such as DALL-E[i] and ChatGPT[ii] is the offspring of two important developments, firstly the improvement and variety of training models, second but most importantly, the large available training datasets. The first source of works stems from open access or public domain works, these are sources that are licensed under permissible licences such as Creative Commons,[iii] or they are works that are in the public domain such as Flickr.[iv] But of course, the amount of such datasets is limited, so researchers can have access to many other datasets, some are even free such as Common Crawl.[v]
What’s important here is how creativity works? Creativity is the ability to produce novel and valuable ideas. Creativity involves using one’s imagination to come up with new ideas and to think outside the box. Is Artificial Intelligence here having imagination or is just creating something out of data that is fed into it?
ANALYSIS OF AI AND COPYRIGHT’S FUTURE
The world can plainly see the creativity and intelligence that AI systems have displayed, and those who are in charge of upholding intellectual property rights are starting to worry about how to secure such rights. Thus, in relation to AI systems, we need to investigate more deliberate purposes for copyright laws. These days, robots are capable of producing really imaginative works of art that, if they were made by people, would be covered by copyright. This necessitates a global review of copyright regulations governing AI systems.
The diffusion model is a popular approach to training AI systems and has been used in some of the most successful recent examples of AI, such as DALL-E 2, Stable Diffusion, and Midjourney. The diffusion model works by taking an input, such as an image, and adding noise to it to create a corrupted version of the input. The AI system is then trained to “reverse” the corruption process and produce the original input from the corrupted version.
The most important takeaway from the perspective of a legal analysis is that a generative AI does not reproduce the inputs exactly, even if you ask for a specific one. For example, If we ask DALL-E to generate “CAT in Style of Vincent Van Gogh” it will generate an artwork which will highly resemble the character and uniqueness of Artist Vincent Van Gogh.[vi]
DALL-E is a large-scale AI language model developed by OpenAI. The model was trained on a dataset of text and images, which allowed it to learn the relationship between language and visual concepts. This allows DALL-E to generate images based on natural language descriptions, such as “a drawing of a three-eyed purple elephant wearing a tutu.” The training data used to develop DALL-E likely included a variety of text and image sources, comprising books, articles, and other publicly available data. The exact details of the data used to train DALL-E are not publicly available however OpenAI has clarified the Pre-Training Mitigations of DALL-E.[vii]
If an AI system were to commit copyright infringement, it is not clear how the law would treat it. Some argue that because AI lacks the ability to have original thoughts or creativity, it should not be liable for copyright infringement. Others argue that the person or entity that created or trained the AI system should be held responsible for any infringement committed by the AI. Ultimately, the legal treatment of AI-related copyright infringement will depend on how the law evolves to address these complex issues.
The laws regarding copyright can vary from country to country. In general, however, most countries have laws that provide protection for original works of authorship, such as literary, dramatic, musical, and artistic works.[viii] These laws typically give the owner of a copyright the exclusive right to reproduce, distribute, perform, and display the work, and to authorize others to do so.
The specific details of copyright law can vary depending on the country. For example, the duration of copyright protection, the scope of the exclusive rights granted to the copyright owner, and the limitations and exceptions to those rights can all differ from one country to another. Additionally, some countries may have specific provisions in their copyright laws that address the issue of AI-generated works, while others may not. It is important to consult the specific laws of a particular country to understand the specific details of its copyright regime. Particularly talking about US and India.
The Copyright Act of the United States[ix]safeguards original creative works of authorship But the US Copyright Office has solidified its position on the matter, stating that human authorship is a requirement for awarding copyright ownership in the US. Feist Publications v. Rural Telephone Service Company, Inc.[x] established that copyright law exclusively safeguards “the fruits of intellectual labour that are founded in the creative powers of the mind,” which is where this strategy originated.
“Author” is defined in India as “the person who causes the work to be formed” in regard to “any literary, dramatic, musical, or aesthetic work that is computer-generated” under the Indian Copyright Act of 1957. The copyright Act’s architects’ strong stance—that copyright protection may only belong to a natural human—is encapsulated in Section 2(d). The Delhi High Court expounded on the definition of “author” in the case of Camlin Pvt. Ltd. v. National Pencil Industries.[xi] The court determined that the “mechanically duplicated printed carton” was not protected by copyright since the originator could not be recognised. Regarding the authorship of an AI-generated work that was produced without human intervention, the Indian Copyright Act’s provisions are still unclear.
In India, the “Modicum of Creativity” test is used to determine whether a work is eligible for copyright protection. Under this test, a work must exhibit a minimum level of creativity in order to be protected by copyright. This means that the work must be having some originality and cannot simply be a copy of another work.
The question of whether AI-generated works are eligible for copyright protection is still a matter of debate in India. Some argue that because AI lacks the ability to have original thoughts or creativity, it should not be entitled to the same rights as human creators. Others argue that the works produced by AI can still have value and merit protection under the law.
[vi] Guadamuz, Andres. “Copyright Infringement in Artificial Intelligence Art.”TechnoLlama, 11 Aug. 2022, www.technollama.co.uk/copyright-infringement-in-artificial-intelligence-art.
[vii] Nichol, Alex. “DALL·E 2 Pre-Training Mitigations.”OpenAI, 28 June 2022, https://openai.com/blog/dall-e-2-pre-training-mitigations.
[ix]United States Code: Copyright Office, 17 U.S.C. §§ 201-216 (1958)
[x]Feist Pubs., Inc. v. Rural Tel. Svc. Co., Inc., 499 U.S. 340 (1991)
[xi]Camlin (P) Ltd v. National Pencil Industries, 1985 SCC OnLine Del 378