Introduction:
The dominance of AI in the contemporary world is such that if it were to disappear into thin air one day, mankind would find itself devastatingly crippled. While it looks like Artificial Intelligence is on the track to obliterate human intervention; the things which it cannot do are those which come most naturally to humans. Morals, ethics, cultural dynamics, and social reasoning are just some things that cannot be fed into an algorithm. Programmers feed a set of rules into the computer, set the parameters till which it can run, and each time it is processed, the computer develops enhanced expertise by building up from what it has learned. Artificial Intelligence is being applied to work in diverse sectors, even those without arithmetic parameters. While it can be said that a machine would not be able to match the aesthetic style or the symphonic sequence made by an artist who has channeled years of experience into his work; an arts-collective company called Obvious in 2018, made 'a generative adversarial network portrait painting' titled - Edmond de Belamy, using AI. More recently, AI was also credited with restoring the edges of Dutch artist Rembrandt's painting. Statutorily, it is clear that computer programs and software are protected under Copyright Law, but who exactly would own the copyrights to works generated using Artificial Intelligence? This question has been around since the influx of mechanically generated works and poses copyright issues to not only works of art, but also musical and literary works.
The Issue Of Ai Authorship:
In most traditional works, copyright protection is automatically afforded to the artists as soon as it is expressed in tangible form. So it is only fair that for works generated mechanically using artificial intelligence, the copyright protection would vest with the person who writes the algorithms which were used to generate the work. As simple as it seems, the process of creating an AI tool is not easily discernible. The programmers who write the algorithms only set the parameters within which it has to function, and once it is processed AI does most of its own work to generate the desired result. If this argument is to be purported, then AIs could just as easily be granted the same status as human authors since it satisfies the two most important requirements for copyright protection i.e. originality and creativity. Once it is clear that copyright protection may be given to AI, the next issue that would arise is ascertaining who exactly would be given the rights to such work. Since computer algorithms are sets of instructions that are independently processed and such algorithms are often scattered among various programmers. Additionally, AI also depends on several other external and internal determinants to function. The fusion in the work generated by AI is so unclear that it becomes difficult to ascertain where exactly they have been referenced from, or to determine each contribution in a quantifiable manner.
In the curious case of Edmond de Belamy, the artwork was created by French developers and auctioned off in England. At the bottom, it was signed with a part of the algorithm code that was used to create it, to perhaps match the style of a conventional painting. The portrait was created using GAN- something that was developed by an independent researcher called Ian Goodfellow in 2014. The code for the program was borrowed from a student called Robbie Barrat who posted them to the popular code-sharing website- GitHub. The case in question here has three elements to it- the first being the artwork itself, the second is GAN and the third one is the algorithm used to train the GAN. Since each of these elements owes its origin to a different creator, the copyright implications on works generated combining all three would be almost too difficult to ascertain. In another instance, Artificial Intelligence helped restore the missing edges of Rembrandt's painting called The Night Watch. The edges of the original painting had been trimmed to fit in the city halls and have since been restored by using AI tools. The machine was trained to learn the Baroque style of Rembrandt's work and also referenced an original copy of the full painting made by Gerrit Ludens. The digital restoration resulted in not only an off-center perspective which was originally intended by Rembrandt but also attracted high praise from art critics coming close to what it could have been if it were never cut-off. The underlying principle in most of these AI tools used for generating artwork is that they can be programmed to pick up and build upon what already exists. It may be trained to mimic the style of a certain painter, emphasize certain brush stroke techniques, embody certain patterns and shapes, or encapsulate all of it in the same work. This ensured that there remained a touch of human intervention and control over the machine. In all of the above instances, the works continue to exist in the public domain, be it owing to the expiry of copyright protection or the total lack of it.
A Case For Ai Authorship & Conclusion
It is clear that any work produced by AI, whether literary, musical, or artistic, is not devoid of its own input. In the Feist Publications case, the US Courts held that creative choices visible in selection and arrangement were necessary to generate sufficient originality to warrant copyright protection.[1] The working of an AI Tool designed especially for creating art, the creativity in the result is ensured by the incidental extraction and arrangement. In Bleistein, v. Donaldson Lithographing Co., it was held that the measure of copyright was not the end-use or aesthetic value of the work, but rather the introduction of a unique element by the author. [2] Every new element introduced by AI into the resultant artwork can be deemed as an expression of its artificial personality. With the ingress of this element of its personality, even though artificial, works produced by AI become unique. In Alfred Bell & Co. v. Catalda Fine Arts, Inc. the Court while discussing engravings, found originality in "distinguishable variations" from the original public domain works. [3] The same can be said about works produced by AI, as anything with substantial similarity could be considered as an infringement but distinguishable variations combined with random extraction and rearrangement would make the work qualify the threshold for creativity. As observed in the Google Books case, it can be suggested that the use of "copyrighted works for the non-expressive purpose of training AI models amounts to fair use". It is noteworthy that Japan has amended its copyright laws and included "exemptions of the use of copyrighted works for machine learning".[4] With both creativity and uniqueness garnered through random extraction, granting copyright protection to AI would only be logical as it seems to satisfy the most basic requirement for copyright protection.
Author: Divya Singh, Student at School of Law, KIIT University, in case of any queries please contact/write back to us at via email chhavi@khuranaandkhurana.com & IIPRD
References:
[1] Feist Publications v. Rural Telephone Service Co., 499 U.S. at 348.
[2] Bleistein, v. Donaldson Lithographing Co 188 U.S. 239 (1903).
[3] Alfred Bell & Co. v. Catalda Fine Arts 191 F.2d 99 (2d Cir. 1951).
[4] Karen Robinson, "Copyrights in the Era of AI", Adobe Blog, February 27, 2020, available at: https://blog.adobe.com/en/publish/2020/02/27/copyrights-in-the-era-of-ai.html#gs.opdukw (last visited on June 30, 2022).
No comments:
Post a Comment