Artificial Intelligence and Societal Roles
The use of artificial intelligence (“AI”) continues to expand into aspects of our lives. From shopping to health care, everyone has benefited from the implementation of such technology. Artificial intelligence is the ability of a machine to display human-like behavior in areas such as learning, reasoning, and creativity. Should machines have such a predominant role in our lives? What are the legal implications of such use? Other blog posts and websites have covered these and other questions.
The arts have also been fertile ground for research on the use of artificial intelligence. The issue is a fundamental question of who is the artist? The programmer? The computer that generated the work? Is the computer-generated work creative or original enough for copyright registration under copyright law? Can a computer be an artist? There is no simple answer, and artists, computer engineers, and copyright offices worldwide are wrestling with these questions as well.
AI-generated artistic works raise the question of who owns the intellectual property rights to the work. In the U.S., the Constitution provides the basis for copyright law¹. The writers of the Constitution presumed that authors are human and not machines, although clearly, the Constitution does not make that distinction. The New Oxford American Dictionary defines an “artist” as “a person who produces paintings or drawings as a profession or hobby” or “a person who practices any of the creative arts such as sculptor, novelist, poet, or filmmaker”². Furthermore, a “person” is defined as “a human being regarded as an individual”³.
Copyright law, for instance, raises the question about the originality and creativity of the work. In the U.S., for copyright protection, the work must be “an original works of authorship fixed in any tangible medium of expression”⁴. One example of a type of artificial intelligence is machine learning. The idea behind machine learning is that it receives training data to learn and make predictions. Training data may consist of the sample output data and the corresponding sets of input data that influence the output. The training model runs the input data through the algorithm to correlate the processed output against the sample output. In general, there are two types of learning: supervised and unsupervised. Supervised learning occurs when the training data contains both the input and output values already. Unsupervised learning involves determining patterns in the data where there is no preset output.
Use of AI in Music
One example is the use of artificial intelligence in completing Beethoven’s unfinished 10th symphony⁵. The 10th symphony not only used recordings by Beethoven but also pieces of music that may have influenced Beethoven at the time, including pieces by Johann Sebastian Bach. The artificial intelligence repeatedly reassembled fragments from the existing pieces of music, superimposed them, and mixed them. In the end, this resulted in a score for Beethoven’s 10th “AI” Symphony by the composer Walter Werzowa and the Beethoven Orchestra Bonn.
As shown in the 10th symphony example, AI music is based on the data collection of pieces of music, symphonies, and operas – created by another party. For instance, a machine-learning system may use known music by a composer to train a machine learning model to conceivably generate new musical work. Originality could become an issue; however, how those pieces come together with the aid of an algorithm may show originality for this newly created piece. Also, the algorithm itself, authored by a person, may also be the subject of copyright protection.
Use of AI in Literary Works
Should it be possible for an author to receive copyright protection for a work generated with the assistance of artificial intelligence? As seen earlier, copyright law requires an identifiable human. The Copyright Office requires not only an identifiable human author for registration but also human authorship of the protected work⁶.
AI is in use in industries such as journalism. For instance, Narrative Science®’s patented Quill™, a software which produces news stories and reports⁷. As described on their website, authoring a story is as simple as selecting the data someone wants to write about⁸. The power of such tools is that “It explains your data in a way everyone can understand – with stories”⁹.
For general writing, AI assistance such as Sudowrite provides an artificial intelligence engine that may “bust writer’s block with our magical writing AI”¹⁰. Sudowrite uses a GPT-3 or “generative pre-trained transformer 3” as its base, created by OpenAI, an artificial intelligence research laboratory in San Francisco, and acquired by Microsoft® in 2019. The 175-billion parameter deep learning model can produce human-like text and learn from large text datasets with hundreds of billions of words.
Copyright law may not necessarily exclude the works that used AI writing assistants from copyright registration. Although there is no case law currently indicating which direction the courts may opine in this area, there still may be a level of creativity provided by a human author as the author may just want AI assistant inspiration rather than using its words verbatim. The Copyright Office may require statistics indicating what percent of the work used the AI assistant versus the human author. However, such metrics may be difficult to obtain from such assistants as they may not be capable of such operation.
Use of AI in Visual Arts
During the summer of 2021, engineers used artificial intelligence to restore Rembrandt’s The Night Watch¹¹. Scientists used neural networks to scale the missing elements from the copy to the original. Another set of algorithms helped them matchRembrandt’s signature light-and-shadow style as they “extended’ the piece back to its full form.
In addition, artists created a work called “Portrait of Edmond De Belamy” with a computer-generated algorithm:
The piece sold for $432,500 – 45 times its expected price¹². The painting includes a signature at the bottom right that includes a piece of the algorithm that produced the work. The algorithm is made up of two parts: the generator, which makes a new image based on the set, and the discriminator, which aims to tell the difference between a human-made image, and one created by the generator¹³. The generator learns from its mistakes; when the discriminator is fooled into thinking that a new image is a real-life portrait, the portrait results. The algorithm was trained on a set of 15,000 portraits from the online art encyclopedia WikiArt, spanning the 14th to the 19th century.
The visual arts are another group where there is no case law that would help guide any copyright ownership question about the work. Should the programmer, or programmers that developed the code, or the algorithm, be the owners of the registration? The work itself is not subject to registration due to the human authorship requirement described earlier.
The US Copyright Office has provided initial guidance on this matter. On February 14, 2022, the Copyright Office refused to register artwork called “A Recent Entrance to Paradise” which was created entirely with an AI machine called DABUS which stands for “Device for the Autonomous Bootstrapping of Unified Sentience.”
The Copyright Office stated that “[c]opyright law only protects ‘the fruits of intellectual labor” that “are founded in the creative powers of the [human] mind,’” noting that it was accepting “as a threshold matter” that “the picture was autonomously created by artificial intelligence without any creative contribution from a human actor”¹⁴. The board also cited the Supreme Court case called Burrow-Giles Lithographic v. Sarony in which Justice Miller had, limited copyright protections to “artists made of flesh and bone”¹⁵.
Conclusion
The use of artificial intelligence in the arts continues to grow, and as such, the legal ramifications of such use, especially in copyright law, are still not determined. Congress and the courts should provide guidance as to a path forward in this area. This guidance should be consistent with the rest of the world to make copyright registration under the Berne convention consistent.
1. “To promote the progress of science and useful arts, by securing for limited times to authors and inventors the exclusive right to their respective writings and discoveries,” US Constitution, Article 1, Section 8, Clause 8.
2. New Oxford American Dictionary, 3rd ed. (New York: Oxford University Press, 2010), s.v. “Artist.”
3. Id., s.v. “Person.”
4. 17 U.S.C § 102(a)
5. https://www.smithsonianmag.com/innovation/how-artificial-intelligence-completed-beethovens-unfinished-10th-symphony-180978753/
6. U.S. Copyright Office, Compendium of U.S. Copyright Office Practices § 306 (3d ed. 2021) (“The U.S. Copyright Office will register an original work of authorship, provided that the work was created by a human being”).
7. https://narrativescience.com/
8. https://narrativescience.com/quill
9. Id.
10. https://www.sudowrite.com/
11. https://www.bbc.com/news/technology-57588270#:~:text=The%20missing%20edges%20of%20Rembrandt’s,been%20restored%20using%20artificial%20intelligence.&text=But%20computer%20software%20has%20now,trained%20by%20processing%20existing%20information.
12. https://www.christies.com/features/A-collaboration-between-two-artists-one-human-one-a-machine-9332-1.aspx
13. https://www.scmp.com/magazines/style/tech-design/article/3017387/would-you-pay-us432500-blurry-painting-created-ai
14. Copyright Office, Correspondence ID 1-3ZPC6C3; SR # 1-7100387071, February 14, 2022.
15. Id.
Source: https://www.jdsupra.com/legalnews/application-of-artificial-intelligence-7871054/
