As the impending tide of the fourth industrial revolution begins to engender a Gordian knot for both the judiciary and policymakers, a key buzzword that has emerged across most debates would be that of artificial intelligence (AI). The possibilities that AI promises are endless – whether in the legal, banking or medical sectors, AI has the potential to completely overhaul the status quo. The best examples of AI can be seen in travel navigation, smart home devices, and even drone technology used by large companies such as Amazon.
Yet, every rose hides its thorns. This op-ed examines the potential thorns that AI has created in Intellectual Property (IP), the legal domain which seeks to protect creations of the human mind. Intellectual Property refers to any man-made creation for which exclusive rights are recognized by the law. Yet readers may ask – what, if any, rights should be recognized for anything that is created by something that is not human? Can the definition of “man-made” be stretched so far, to the extent that anything an AI invents would be recognized to gain proprietary rights? This op-ed undertakes a deep-dive into this issue in the context of Singapore.
A study on Singapore’s intellectual property regime prepared by the Lee Kuan Yew School of Public Policy (LKYSPP) notes that one critical factor which accelerated Singapore’s growth is its push to foster “innovation, commercializing research and development … whilst striving to anchor Singapore as a global hub.”[1] Without a doubt, human capital is inextricably linked to knowledge production. Indeed, Harvard economist Claudia Goldin has attributed the growth of the global economy to “increased knowledge [created in society] and its diffusion,” viewing the increase in knowledge as “an advance in human capital,” because “knowledge creation” can add to “augmentation of the labor input.”[2] Singapore’s investment in knowledge creation—through augmenting our people with the necessary higher-level skills which fosters research and development—has pushed Singapore forward to becoming the “Silicon Valley of Asia”.[3]
In this regard, what keeps the flame of knowledge generation alive is Singapore’s IP regime.[4] The LKYSPP study states that “a strong [IP] regime influences not just the quantity of FDI (foreign direct investment) inflows, but also quality,” and is a critical factor which anchors multi-national corporations. Indeed, “stronger IP protection attracts R&D intensive activities and creates a demand for skilled workers.”[5] Currently, Singapore ranks 11th out of 53 countries in the world in the US Chamber of Commerce’s International Intellectual Property Index, due to an advanced IP framework which accelerates research, patent examination, and grants.[6]
However, the winds of artificial intelligence (AI) have begun to threaten our safe harbors. Various issues, which include questions on whether AI technologies can be patented, whether AI can be inventors, and how such strategic IP assets can be better managed by organizations, have begun to befuddle foreign jurisdictions. Indeed, Minister for Culture, Community and Youth and Second Minister for Law Edwin Tong has highlighted that whilst traditionally seen as a legal or technical function, IA/IP management is increasingly becoming an essential part of strategic management decisions as it impacts shareholder value. [7] Therefore, Singapore has released a decade-long IP policy strategy, namely the Singapore IP Strategy 2030 (SIPS 2030), to ensure its continued relevance in the global stage.
Before examining the SIPS 2030, this article first seeks to dive into the question of how AI might pose a challenge to our existing IP regime. Having done so, it will examine the likely effect of the SIPS on Singapore’s IP framework. Finally, this article will propose specific measures which might be implemented in the SIPS to address the challenges posed by AI.
AI as Inventor
In seeking a reform of Singapore’s domestic IP regime, it remains important to first examine the tides sweeping through foreign jurisdictions. Across the world, technology juggernauts such as IBM, Microsoft, and Toshiba have begun to jump on the AI bandwagon, filing voluminous amounts of patent applications to protect the IP of their AI inventions. IBM, for example, has filed over 8,290 applications while the ratio of scientific papers to inventions has decreased from 8:1 in 2010 to 3:1 in 2016, indicating that more AI technologies have been used for inventive activity.[8] Exciting as this might be, scholarly concern has begun to bubble, as the law has omitted to address whether an AI can be considered an inventor, with this issue being particularly critical for patent law given its spillover effects for the other aspects of the IP regime.
The tangible benefits of AI being recognized as inventors can be summed into one word – recognition. Quite simply, awarding recognition to AI would engender more breakthroughs. As Abbott argues, allowing AI to be recognized as inventors “would make inventive AI more valuable and incentivize AI development, which would translate to rewards for effort upstream from the stage of invention and ultimately result in more innovation”.[9] Though AI may still lack the self-awareness to appreciate that it has been recognized as an inventor, its human owners most definitely will, and financial rewards and industrial acclaim will naturally follow down the stream to an AI’s creator. The general implications, in contrast, for not recognizing AI as inventors would be to cause these inventions to tumble into the public domain. This would render them ownerless, or force human inventors to protect what said AI creates as trade secrets, with this result entirely against the raison d’être of innovation.
Indeed, for patent law, foreign jurisdictions remain befuddled on whether AI is itself patentable, alongside the proper role AI should play in the inventorship process. This article focuses on the latter, as this has spillover effects for the rest of intellectual property, as well as for Singapore law in general. Therefore, in undertaking this inquiry on inventorship, one must examine whether an AI ought to be an inventor and if not, who should be the proper inventor, if one were to deny inventorship rights to AI.
In a recent case brought by Stephen Thaler in the US, the US District Court held an AI system cannot be named as an inventor. Instead, an inventor must be a natural person.[10] Here, the court was insistent to adopt a natural reading of ‘inventor’. Parsing the statutory language used in the Patents Act, explanatory materials, previous versions of said legislation, alongside case law examining the meaning of ‘inventor’ in US law, the court was firm that the grant of a patent had to be premised upon an invention arising from ‘the mind’ of a natural person, and ruling otherwise was currently unsupported by existing legislative material. This holding is consistent with the position of the US Patent and Trademark Office, alongside the European Patent Office (EPO), though it stands in a stark contrast to Australia and South Africa, which hold that AI can be inventors.
For the former, the EPO viewed it preferable to bar AI systems from inventorship, because there is an internationally applicable standard interpreting “inventor” as a natural person.[11] The EPO viewed this as preferable because Article 60 of the European Patent Convention requires an applicant to have “legal capacity,” interpreted as an ability to be subject to rights and duties.[12] Likewise, the UK Intellectual Property Office stressed that absent any legislative intervention, one should not conflate the term “natural person” to include AI technologies.[13]
For the latter, the Australian Intellectual Property Act interprets an inventor as an agent, and Australian courts held that said reading of agent is sufficiently broad to include an AI system. According to the court, this interpretation respects “the reality in terms of many otherwise patentable inventions … where [it is obvious] a human cannot be an inventor.”[14] Likewise, South Africa permits inventions to be owned by an AI system, as there is nothing in the IP Act which explicitly prevents ownership.[15]
This quandary has however led to a deeper rabbit hole regarding who may claim credit for inventions created by an AI system. Lacking adequate policy guidance, inventors have omitted to disclose the role of AI in their inventive process. As an example, Thaler’s AI generated a myriad of novel inventions. However, due to uncertainties on whether AI can be inventors, Thaler omitted to mention the Creativity Machine for subsequent inventions of its AI.[16] Indeed, Thaler states that if he listed himself as an inventor for something his AI had created—“he would have two issued patents … [which] would be undesirable because it would involve misrepresentations and individuals would be claiming credit for work they did not perform.”[17]
This has yet to be addressed in Singapore. Presently, inventors are using narrow AI to perform inventive tasks which are well-structured, as contrasted to broad AI which might reach the intelligence level of a typical human. Narrow AI is AI that is focused on a specific, singular task, which lacks the self-expansion functionality that is required to unravel unfamiliar problems. In contrast, broad AI is a sophisticated and adaptative system which is able to perform any cognitive task by virtue of its sensory perceptions and data input. When narrow AI is used, who counts as the inventor? Would it be the AI system, the programmer, or the data trainer (the individual feeding the AI with data to enable it to learn the art of ‘inventing’)? Courts have yet to answer this question, nor have Singapore policymakers come to any consensus. Given that modern technological standards are still far from reaching the level of broad AI, the debate has largely focused on narrow AI.
SIPS 2030
The SIPS undergirds Singapore’s response to the above-mentioned changes. Singapore aims to: (i) ensure a world class IP regime by supporting technology advances through policy reviews by empowering innovators through digitization; (ii) positioning Singapore as a node connecting ASEAN and the world; and (iii) growing the international IP dispute resolution in Singapore.[18]
How might Singapore address the challenges heralded by AI in particular? Prong (i) sets out a long-term study to examine the interface between AI and IP policy.[19] According to the SIPS, “AI has … raised important issues for [Singapore’s] IP regime, particularly regarding whether the regime is conducive for AI innovation and adoption.” Noting increasing public consultations on AI-related topics in other jurisdictions, the Intellectual Property Office of Singapore (IPOS) acknowledges Singapore’s own move through the National AI Strategy 2019, which recognizes the important role of the IP regime in supporting the development and commercialization of new AI technologies, by providing a top-class IP regime alongside accelerated patent initiatives. Yet, the IP Strategy remains uncertain as to the concrete way forward. Although the report cites a recent Information Note addressing this specific issue, this note merely restates the problem, explaining the “key issues that AI innovators ought to be aware of.”[20] Therefore, Singapore finds itself back at square one, with policymakers left without clear guidance on how to tackle this vexing question.
Nevertheless, the other components of the SIPS deserve commendation for being forward-looking. The proposal to create a “next-generation IP filing system” is likely to incentivize business activity. The IPOS is keenly aware of the importance of user experience in the filing process; and making IP easier to register will ultimately engender knowledge production.[21] Likewise, the “Patent Prosecution Highway” proposal is commendable, this being a “work-sharing” arrangement which allows businesses to use patent search and examination reports from an earlier office in another jurisdiction to accelerate patent prosecution. This arguably reduces costs, thereby increasing the value of holding a patent in Singapore.[22] Furthermore, the IPOS’s proposal to leverage Singapore’s position as the dominant international dispute resolution hub in the world, which includes amending the Arbitration Act to promote IP dispute resolution, will establish Singapore as a prime location for international IP disputes.[23]
Charting the Way Forward
Singapore must come to a consensus on whether to allow AI systems to be recognized as inventors, given the significant impact AI will have on our economy in time to come. For Singapore’s manufacturing industry alone, AI adoption is set to increase projected growth by US$101.1 billion. [24] Lacking any incentive for AI, numbers have instead been projected to be around US$71.3 billion.[25] An IP framework that discourages the use of AI will then hamper innovation and affect our economy. How should policymakers then proceed in the regulation of AI for inventive activity?
It is argued the Australian position is preferable for Singapore. This is because favoring AI systems being recognized as inventors would increase the value of inventive AI and encourage AI development in general, with this providing a mode for inventors to get recognition without tumbling into the rabbit hole of inventions falling into the public domain. Given the infancy and authority of the Australian decision, there has yet to be evidence in Australia where AI systems have been recognized as inventors, though the future will certainly tell. The result would be an upstream reward system for inventors, which leads to a domino effort encouraging inventors to create better AI. Indeed, this is consistent with Singapore’s National AI Strategy—compared to other jurisdictions such as the US or the UK which are relatively hostile to the patentability of computer applications, Singapore has adopted a policy stance favorable to the patentability of AI through its Accelerated Initiative for Artificial Intelligence (AI2). As Lindsay highlights, the Singapore Patent Examination Guidelines refer to the solution of a specific problem as being the key factor in determining whether an AI related application is patentable subject matter. This is much wider than the position in the UK, which requires a technical subject matter to be solved.[26] Given this liberal position, a regime favoring recognizing AI systems as inventors would be favorable for Singapore, encouraging human talent to be groomed in AI development. By focusing on the problem being solved, rather than what exactly is being used to solve the problem, the confusion between narrow and broad AI can be resolved, cementing the way for a more forward-looking regime.
The SIPS is the correct step forward. With minor tweaks, particularly through equipping its AI proposals with more substantive teeth, Singapore will be able to lead the global charge and emerge as an AI juggernaut of tomorrow.
Featured photo from Tata Communications.
[1] Lee Kuan Yew School of Public Policy, The Development of Singapore’s Intellectual Property Rights Regime, https://lkyspp.nus.edu.sg/docs/default-source/case-studies/lkwms_series01_sg_ip.pdf?sfvrsn=5135960b_2 (2014).
[2] Claudia Goldin, Handbook of Cliometrics (C. Diebolt, M. Haupert (eds.)) (2016), p.56-57.
[3] Catherine Shu, Singapore is poised to become Asia’s Silicon Valley, TechCrunch, https://techcrunch.com/2020/12/14/singapore-is-poised-to-become-asias-silicon-valley/ (2020).
[4] Intellectual Property Office of Singapore, Who We Are, https://www.ipos.gov.sg/who-we-are/introduction (2022).
[5] Supra n3 at 13-14.
[6] International Trade Administration, Singapore – Country Commercial Guide, https://www.trade.gov/country-commercial-guides/singapore-protecting-intellectual-property (Aug. 13, 2021).
[7]Government of Singapore and Intellectual Property Office of Singapore. “Singapore IP Strategy 2030.” Singapore: Intellectual Property Office of Singapore, April 2021. https://www.ipos.gov.sg/docs/default-source/default-document-library/singapore-ip-strategy-report-2030-18May2021.pdf.
[8] Campbell Kwan, IBM leads Microsoft, Toshiba, and Samsung in AI-related patents: UN, ZD Net, https://www.zdnet.com/article/ibm-leads-microsoft-toshiba-and-samsung-in-ai-related-patents-un/ (Jan. 31, 2019).
[9] Abbott, Ryan. The Reasonable Robot: Artificial Intelligence and the Law. Cambridge University Press, 2020.
[10] Thaler v. Hirshfeld, No. 1:20-cv-903, 2021 WL 3934803 (E.D. Va. Sept. 2, 2021).
[11] European Patent Office, EPO publishes grounds for its decision to refuse two patent applications naming a machine as invention (Jan. 28, 20202), https://www.epo.org/news-events/news/2020/20200128.html
[12] See Generally, EP3564144, F3305 Communication of the Board of Appeal (ex parte/ inter partes) (Jun. 21, 2021)
[13] Thaler v The Comptroller-General of Patents, Designs And Trademarks [2020] EWHC 2412 (Pat)
[14] Thaler v Commissioner of Patents [2021] FCA 879 (Austl.).
[15] ZA 2021/03242
[16] See U.S. Patent No. 5,852,815, and World Economic Forum, Artificial Intelligence Collides with Patent Law (Apr. 2018), http://www3.weforum.org/docs/WEF_48540_WP_End_of_Innovation_Protecting_Patent_Law.pdf.
[17] Thaler v. Iancu, et al., 1:20-cv-00903 D.I. 1 (E.D. Va. Aug. 6, 2020) at [39].
[18] Intellectual Property Office of Singapore, Singapore IP Strategy 2030 Report (2021), https://www.ipos.gov.sg/docs/default-source/default-document-library/singapore-ip-strategy-report-2030-18May2021.pdf.
[19] Ibid., p.23.
[20] Intellectual Property Office of Singapore, IP and Artificial Intelligence Information Note (2021), https://www.ipos.gov.sg/docs/default-source/default-document-library/ip-and-ai-info-note.pdf.
[21] Intellectual Property Office of Singapore, Singapore IP Strategy 2030 Report (2021), https://www.ipos.gov.sg/docs/default-source/default-document-library/singapore-ip-strategy-report-2030-18May2021.pdf, p.24.
[22] Ibid., p.25.
[23] Ibid., p.35.
[24] Charlotte Trueman and Cristina Lago, How Singapore is using Artificial Intelligence, CIO, https://www.cio.com/article/221994/how-singapore-is-using-artificial-intelligence.html (Mar. 12, 2019).
[25] Ibid.
[26] Marks & Clerk, Singapore grants first fast track AI patent, https://www.marks-clerk.com/insights/singapore-grants-first-fast-track-ai-patent/ (Sep. 23, 2019).