Written by Samuel Chan Zheng Wen (Associate Author) | Mentored by Lenon Ong | Reviewed by Associate Professor Saw Cheng Lim
LawTech.Asia is proud to have commenced the third run of its popular Associate Author (2020) Programme. The aim of the Associate Authorship Programme is to develop the knowledge and exposure of student writers in the domains of law and technology, while providing them with mentorship from LawTech.Asia’s writers and tailored guidance from a respected industry mentor.
In partnership with the National University of Singapore’s alt+law and Singapore Management University’s Legal Innovation and Technology Club, five students were selected as Associate Authors. This piece by Samuel Chan, reviewed by industry reviewer Associate Professor Saw Cheng Lim (Singapore Management University School of Law), marks the first thought piece in this series. It examines the future of artificial general intelligence and intellectual property rights.
From the triumph of computer programme AlphaGo over the world Go champion Lee Se-Dol in 2016 to Google’s algorithm, PoemPortraits, which generates an entire unique poem from a single word of one’s choice, the feats accomplished by Artificial Intelligence (“AI”) continue to astound and impress. The growing capabilities of AI, both in terms of sheer computing power as well as (what appears to be) creativity, have increasingly resembled and even surpassed those of a natural person’s cognitive abilities. It is against this backdrop that the law has been forced to confront novel issues, for example whether rights and liabilities may be attributed to AI entities. In the same vein, this article seeks to briefly canvass the potential challenges which the rise of AI may present to the field of intellectual property (“IP”) law.
What is AI?
Before proceeding to consider the intersection between AI and IP law, it may be beneficial to elaborate on what the term “AI” refers to (apart from generic notions of its applications in Facebook advertisements and the Siri app in your iPhone). While there does not appear to be a universally accepted definition of AI, it is generally considered to be a discipline of computer science targeted at creating machines and systems capable of carrying out tasks considered to require human intelligence. Intelligence, in this regard, is taken to be “that quality that enables an entity to function appropriately and with foresight in its environment”. Of course, given that intelligence falls along a spectrum, this definition may be broad enough to include both autonomous robots as well as the facial recognition software on your phone. Nevertheless, discussions on AI tend to focus primarily on the higher-level functions of human beings, such as the ability to reason, achieve goals, and perceive and respond to sensory inputs. In this regard, machine learning and neural networks, which allow computer programmes to learn through experience (such as by analysing training examples), have attracted much attention.
At this point, it may also be helpful to make the distinction between what has commonly been termed as artificial narrow intelligence (“ANI”) and artificial general intelligence (“AGI”). ANI refers to a type of AI which focuses on achieving intelligence for the purposes of carrying out a narrowly defined task. These tasks include speech recognition, text comprehension, and pattern recognition. Most applications of AI in our everyday lives comprise the use of ANI, which commonly functions as tools which assist us in the completion of various tasks. In that sense, ANI may be thought of as Photoshop to a designer’s creations or a pen to a writer’s work.
AGI, on the other hand, aims to develop intelligence within a machine or system that is not specific to any problem, context, or task but which is able to perform any task a human being is capable of. This type of intelligence encompasses the abilities to independently reason, plan, and learn in order to achieve goals (think I, Robot-esque humanoid robots). While it is speculated that AGI remains a distant reality, the possibility of AI acquiring human-like cognitive abilities (and possibly surpassing humans in such abilities) sometime in the future raises interesting questions for IP law, and it is with this conception of AI in mind that the article will proceed.
AI and IP Rights
The potential rise of AGI raises the question of how IP regimes will respond, in particular because many of these regimes trace their roots back to relatively dated statutes. For example, Singapore’s Patents Act is largely based upon the UK Patents Act 1977, while Singapore’s Copyright Act was enacted in 1987 and was originally adopted from the Australian Copyright Act 1968, with some provisions copied from the US and UK copyright legislation. Hence, the fundamental concepts underlying these fields of IP law, such as the conditions for a patent to be granted and the types of works to which copyright may attach, were not developed with the likes of AI (much less AGI) in mind. Similarly, the justifications and objectives behind IP rights have also been left largely unquestioned in spite of the significant advancements in technology and science since.
IP regimes share the feature that they, in their respective ways, protect some product of the human mind, for varying periods of time, against the use by others of those products in various ways. This protection is accorded through the provision of IP rights which may be asserted against members of the public who seek to use or profit from the product of another’s mind. It is therefore easily observed that IP law has been founded upon anthropocentric reasoning, where protection is only accorded to human beings for their creations. This is also supported by the justifications behind these IP protections.
There are three commonly accepted justifications for the protection accorded by IP law. The first justification is founded upon the labour theory, which asserts that people are entitled to own IP rights, and indeed to enforce them on infringing members of the public, based on the efforts they have invested into creating a particular good. Human endeavour is thus rewarded with the protection of the resulting product from unauthorised use by others. The second justification posits that protection stems from the fact that the resulting product is an extension of its creator’s personality. This is based upon theories by, for instance, Kant and Hegel, that people express their wills and develop as persons through their interactions with external objects. This moral justification therefore views interferences with a resulting product, which is an extension of an individual’s personhood, to be unjust. The third justification involves the provision of economic incentives – protection functions as a reward for the availing of a resulting product which benefits society. This prevents individuals from withholding products of potentially great utility to society for fear of losing the ability to profit from these products.
Having considered the anthropocentric justifications behind IP law, it is not difficult to see the potential dissonance which may be generated by the rise of AI, which one would struggle to accept possesses personhood of the same nature as a human being. For example, an AI entity would not need to be provided with economic incentives for it to continually generate products for the welfare of society. In addition, the moral justifications which reward human endeavour and the breadth of human personality seem to fall short when applied to AI entities, which carry out their processes tirelessly and, for the most part, rather mechanically. This difficulty is exacerbated when one considers the limitations which AI entities presumably face regarding emotional and social intelligence. As such, under current IP laws around the world, AI entities are typically not allowed to own IP rights, and are, for example, not legally recognised as the authors or inventors of AI creations.
How, then, should AI and its creations be regulated? Should the longstanding concepts of IP law, founded upon anthropocentric reasoning, be nevertheless thrust upon this burgeoning and novel field of technology? Should patent and copyright protection, for example, be accorded to creations by AI entities? These are the very questions which policymakers, academics, and lawyers have assiduously directed their minds to in recent years.
Possible responses to the challenges brought about by AI
While the frequency of discussions on these issues has increased in recent years, there has not been a clear consensus on potential solutions to address these future challenges. To this end, consultative exercises have been initiated by the US Patent and Trademark Office, the European Patents Office, the UK Intellectual Property Office, and the World Intellectual Property Organisation in the last few years. There appears to be some agreement that the IP regimes in the various jurisdictions remain robust enough to handle the current developments in AI. With regard to how future developments in AI may be handled, several interesting perspectives have been offered.
For one, a study commissioned by the European Patent Office contemplated the possibility that AI entities may be allowed to obtain legal personhood, and thereby be allowed to own IP rights. In particular, the study considered the use of a “functionalist” approach, where the output generated by the AI would be deemed sufficient to accord it IP rights if the same output would have accorded a human being the same rights. While the study seemed to ultimately reject this approach as it would first require a larger revision of the legal system to accord AI entities legal personhood, the study raised the alternative possibility that the natural person who “realises the significance and utility of the output produced by an AI system may be considered as an inventor”, and thereby own the patent rights to the output. In other words, instead of according an AI entity rights directly, AI-generated output may be protected by IP rights held in a natural person. This may, at present, serve to bypass thorny issues concerning the personhood of AI.
Such a solution seeks to extend the justifications for IP protection to AI-generated output by vesting IP rights in a natural person who is most closely connected to the creation of the output. Specifically, these rights would reward the natural person for developing the AI system and/or its output, and to incentivise him to make such output available to the public. This approach has also been echoed by several academics. For example, Pamela Samuelson, a professor at the University of California, Berkeley, School of Law, agreed that creations by autonomous machines should be protected by IP law, although these IP rights should be vested in a natural person. She argues that according IP rights to a person even for a creation completely generated by a machine will, at the very least, reward the circulation of the output (if not its creation) to the general public, which will promote the progress of science and the arts.
However, this raises yet another question: if IP rights are granted to protect AI creations, in whom should these rights be vested? This question is a tricky one because an AI’s output may well depend on some input by a human user, depending on the specific application of AI. For example, Google’s algorithm, PoemPortraits, creates a poem from a single word provided by a human user. As such, IP rights could potentially be vested, exclusively or jointly, in (1) the owner of the AI machine or system, (2) the creator/designer of the AI machine or system, (3) the user of the AI machine or system, or (4) the AI entity itself.
Academic views remain largely divided on this question. Some observers argue that these IP rights should be vested in the owner of the AI system as this would be most consistent with the rules governing the ownership of property and would most incentivise innovation. Others believe that such rights should be vested in the designer of the system as he would be the only person who understands how the output has been arrived at, given that he set the parameters within which the AI machine or system worked; any other individual would be “completely unnecessary and irrelevant to the final result”. In contrast, some academics assert that such rights are best accorded to the user of the system as vesting such rights in the owner or designer of the system would over-reward them, given that they may not be able to anticipate the AI’s output better than anyone else.
The diverse range of views in this area not only demonstrates the complexities surrounding this issue, but also shows how similar considerations (such as pragmatism and the desire to reward human endeavour) in IP law may lead to well-substantiated contrasting views. These difficulties are exacerbated by the lack of clarity on how AI systems work, and the fact that AI-generated output may potentially span a wide range of creations, such as creations where the AI only serves as a tool to a human creator, creations where the AI works jointly with a human, and creations which do not involve humans at all. This suggests that a one-size-fits-all approach towards the question of who should be accorded IP rights for AI-generated output may be largely inappropriate as the role of AI may differ greatly from case to case. Thus, even if IP regimes decide to accord IP rights to protect AI-generated creations, deciding who to vest these rights in will pose yet another thorny issue – one which has yet to be resolved among academic circles.
Concluding thoughts: a state of premature alarm?
The rise of AI, in particular that of AGI, will necessitate a rethink of the objectives behind IP law, which have traditionally been focused on protecting the product of an individual’s mind. The tension which lies at the heart of IP law is one between providing adequate reward (both pecuniary and moral) to the individual and maximising the benefits to society. But what happens when the individual is now replaced (at least in part) by an AI entity? IP law will thus be forced to consider the new role(s) of the individual, and whether the traditional tension will remain relevant as a key guiding consideration in the development of IP law where AI is involved. This will require relevant stakeholders to keep abreast of the developments in AI and to understand the precise role of both humans and AI in applications of AI technology. To this end, the various legal review exercises undertaken by IP offices and law reform bodies around the world, including the Intellectual Property of Singapore and the Singapore Academy of Law’s Law Reform Committee, represent a welcome first step in the right direction.
In any case, it should be noted that many of the difficulties which AI poses to IP regimes are associated only with the rise of AGI, which has yet to arrive. In fact, some experts predict that AGI may take several decades to realise its full potential. Other observers remain convinced that AI will never be able to match certain aspects of human cognitive functions, such as creativity. Given that the current state of AI remains largely that of ANI, where AI functions primarily as a tool to assist humans in the accomplishment of specific tasks, current IP regimes should face little difficulty according various IP rights to creations, as humans still remain firmly in the driver’s seat. Nevertheless, the IP regimes governing AI will not only affect the future development of such technologies, but also the potential utility which such applications may bring to society. As such, it may be prudent to embark upon a careful examination of the moral, social, and economic dimensions to AGI sooner rather than later, to ensure that the relevant IP frameworks remain nimble and responsive to (potentially) exponential developments in this field.
 Choe Sang-Hun, “Google’s Computer Program Beats Lee Se-dol in Go Tournament”, The New York Times (March 15 2016). Accessed at <https://www.nytimes.com/2016/03/16/world/asia/korea-alphago-vs-lee-sedol-go.html> on 7 December 2020.
 Es Devlin, “Create a personalized poem, with the help of AI”, Google Blog (May 2 2019). Accessed at <https://www.blog.google/outreach-initiatives/arts-culture/poemportraits/> on 7 December 2020.
 Nur Adlin Hanisah binti Shahul Ikram & Mohd Yazid bin Zul Kepli, “Establishing Legal Rights And Liabilities For Artificial Intelligence” (2018) 26(1) IIUMLJ 161, p 174.
 World Intellectual Property Organisation, WIPO Conversation On Intellectual Property (IP) And Artificial Intelligence (AI), WIPO/IP/AI/GE/19/INF 4 (31 October 2019), ; United States Patent and Trademark Office, Public Views on Artificial Intelligence and Intellectual Property Policy (October 2020), p ii.
 Stuart J. Russell & Peter Norvig, Artificial Intelligence: A Modern Approach (Prentice Hall, 2009, 3rd Ed), p 2.
 Nils J. Nilsson, The Quest for Artificial Intelligence: A History of Ideas and Achievements (Cambridge University Press, 2010), p xiii.
 David Weinbaum & Viktoras Veitas, “Open Ended Intelligence: The Individuation of Intelligent Agents” (2017) 29(2) Journal of Experimental & Theoretical Artificial Intelligence 371, p 371.
 Id, p 372.
 Ng-Loy Wee Loon, Law of intellectual property of Singapore (Sweet & Maxwell, 2014, 2nd Ed), p 450.
 Id, p 63.
 It is noted that although the Copyright Act and the Patents Act have gone through several rounds of amendments, none of these have specifically targeted developments in AI or autonomous technology.
 David Vaver, “Intellectual property: the state of the art” (2000) 116 LQR 621, p 621.
 Reto Hilty, Jorg Hoffmann & Stefan Scheuerer, “Intellectual Property Justification for Artificial Intelligence”, Oxford Business Law Blog (30 June 2020). Accessed at <https://www.law.ox.ac.uk/business-law-blog/blog/2020/06/intellectual-property-justification-artificial-intelligence> on 7 December 2020.
 Ryan Abbott, “Artificial intelligence, big data and intellectual property: protecting computer generated works in the United Kingdom” in Tanya Aplin (ed), Research Handbook on Intellectual Property and Digital Technologies (Elgar Publishing, 2020), p 333; see also Ng-Loy Wee Loon, Law of intellectual property of Singapore (Sweet & Maxwell, 2014, 2nd Ed), p 17.
 Ng-Loy Wee Loon, Law of intellectual property of Singapore (Sweet & Maxwell, 2014, 2nd Ed), pp 15-16.
 Mark Perry & Thomas Margoni, “From Music Tracks to Google Maps: Who Owns Computer-Generated Works?” (2010) 26 Comput L Secur Rev 621, p 627.
 Ryan Abbott, “Artificial intelligence, big data and intellectual property: protecting computer generated works in the United Kingdom” in Tanya Aplin (ed), Research Handbook on Intellectual Property and Digital Technologies (Elgar Publishing, 2020), pp 329-330.
 Noam Shemtov, A study on inventorship in inventions involving AI activity, commissioned by the European Patent Office (February 2019), p 11. See also Stephen Thaler v The Comptroller-General of Patents, Designs and Trade Marks  EWHC 2412 (Pat).
 See United States Patent and Trademark Office, Public Views on Artificial Intelligence and Intellectual Property Policy (October 2020), p iii; Noam Shemtov, A study on inventorship in inventions involving AI activity, commissioned by the European Patent Office (February 2019), p 7.
 Noam Shemtov, A study on inventorship in inventions involving AI activity, commissioned by the European Patent Office (February 2019), p 28.
 Id, p 33.
 Id, p 35.
 Pamela Samuelson, “Allocating ownership rights in computer-generated works” (1986) 47 U. Pitt. L. Rev. 1185, pp 1202-1204; see also Peter Blok, “The Inventor’s New Tool: Artificial Intelligence – How Does It Fit In The European Patent System?” (2017) 39(2) E.I.P.R. 69.
 Id, p 1226-1227.
 Ryan Abbott, “I Think, Therefore I Invent: Creative Computers and the Future of Patent Law” (2016) 57 B.C.L Rev., 1079, pp 1113-1114.
 Mark Summerfield, “The Impact of Machine Learning on Patent Law, Part 3: Who is the Inventor of a Machine-Assisted Invention?”, Patentology Blog (4 February 2018). Accessed at <https://blog.patentology.com.au/2018/02/the-impact-of-machine-learning-on.html> on 7 December 2020.
 Pamela Samuelson, “Allocating ownership rights in computer-generated works” (1986) 47 U. Pitt. L. Rev. 1185, p 1208.
 Id, p 1190 (see footnote 15).
 Ryan Abbott, “Artificial intelligence, big data and intellectual property: protecting computer generated works in the United Kingdom” in Tanya Aplin (ed), Research Handbook on Intellectual Property and Digital Technologies (Elgar Publishing, 2020), pp 323-324.
 See, for example, Singapore Academy of Law Law Reform Committee, Applying Ethical Principles for Artificial Intelligence in Regulatory Reform (July 2020). Accessed at < https://www.sal.org.sg/sites/default/files/SAL-LawReform-Pdf/2020-09/2020%20Applying%20Ethical%20Principles%20for%20AI%20in%20Regulatory%20Reform_ebook.pdf> on 7 December 2020.
 Katja Grace, John Salvatier, Allan Dafoe, Baobao Zhang & Owain Evans, “Viewpoint: When Will AI Exceed Human Performance? Evidence from AI Experts” (2018) 62 Journal of Artificial Intelligence Research 729, pp 730-731.
 Katherine Schwab, “3 reasons why AI will never match human creativity” (25 April 2019). Accessed at <https://www.fastcompany.com/90339590/3-reasons-why-ai-will-never-match-human-creativity> on 7 December 2020.
This piece was written as part of LawTech.Asia’s Associate Authorship Programme.
Featured image credit: Mobile World Live