AI SYSTEMS: PROPRIETORS OF AI GENERATED CREATIONS?
Author : Att. Önder Özden*
Artificial Intelligence (“AI”) is a field of computer science that includes machine learning, natural language processing, speech processing, expert systems, robotics and machine vision.
The term AI was popularized by John McCarthy and Marvin Lee Minsky, organizers of the 1956 Dartmouth Conference that made AI a field of research in its own right. According to McCarthy, AI was the notion of a program, processing and acting on information, such that the result is parallel to how an intelligent person would respond in response to similar input.[1]
AI refers to systems that demonstrate intelligent behaviour by analysing their environment and taking action, with a degree of autonomy, to achieve specific goals.[2] “Intelligence” then refers to the fact that the machine imitates the cognitive functions associated with the human or animal brain, i.e. the ability to learn and solve problems. This may involve thinking or acting by imitating human behaviour (cognitive approach) or rationality (computational approach).[3]
The current and potential applications of AI are being implemented in different industries. For example, by continually refining algorithms that observe and learn about people based on their online behaviour, search engines and online retailers have become incredibly smart — and precise — at predicting what consumers are interested in purchasing. “…In the Harvard Business Review (May 2019), Nicolaj Siggelkow and Christian Terwiesch postulated that advances in technology, most notably new forms of AI, means that brand owners are no longer waiting for customers to come to them to buy their products. A brand owner will no longer simply place products on the market and wait for a consumer to see their brand and buy their product. AI enables a brand owner to predict the needs and desires of the consumer. Siggelkow and Terwiesch identified four new strategies where AI has the potential to affect significantly and these strategies are the so-called ‘respond to desire’, ‘curated offering’, ‘coach behaviour’ and ‘automatic execution’ strategies of product offering….”[4]
Another example is Aviation. Parameter-driven autopilot systems have been around for years, but AI is increasingly playing a role. Using AI-based systems, DARPA’s Aircrew Labor In-Cockpit Automation System (ALIAS) adds high levels of automation to existing aircraft to assist crews by reducing workloads, rapidly recalling flight procedures and overcoming system failures.[5]
AI is also making inroads into intellectual property (“IP”) and patent research. Patent specialists have long used Boolean[6] and positional search for patent-related discovery, combing through large amounts of literature using advanced syntaxes and manual categorization systems. Machine learning and AI are changing this model by improving both retrieval efficiency and accuracy. Modern searching can overcome the inherent ambiguities that confound classic keyword search by using machine learning techniques.
On top of all the above applications in which AI is being used merely as a tool in the form of weak AI[7], AI systems are nowadays functioning as independent generators of works and inventions without requiring their programme developers’ further intervention and deployment of Training Data except for the very first feeding. Respectively, a computer program developed for machine learning purposes has a ‘built-in algorithm’ that allows it to learn from data input and to evolve and make future decisions that may be either directed or independent.
Mostly they learn from these data to generate a new piece of work, making independent decisions throughout the process to determine what the new work looks like. An important feature for this type of AI is that while programmers can set parameters, the work is actually generated by the computer program itself – referred to as a neural network – in a process akin to the thought processes of humans. Thus, we are in the era of a technological revolution that may require us to rethink the interaction between computers and the creative process. That revolution is underpinned by the rapid development of machine learning software, a subset of artificial intelligence that produces autonomous systems that are capable of learning without being specifically programmed by a human hence becoming the debatable sole creator deprived of a legal personhood thus liability under the current global acquis of intellectual property.
In conformity with the above developments in AI-tech, in 2017, Google has funded an AI program that has written local news articles. In 2016, a group of museums and researchers in the Netherlands have unveiled a portrait entitled “The Next Rembrandt”, a new artwork generated by a computer that had analysed thousands of works by the 17th century Dutch artist Rembrandt Harmenszoon van Rijn. A short novel written by a Japanese computer program in 2016 reached the second round of a national literary prize. And the Google-owned artificial intelligence company Deep Mind has created software that could generate music by listening to recordings. In Germany, we bump into, an artist robot, e-David developed by the academicians in Konstanz University, who takes the brush from the canvas and replenish the canvas with gentle and regular strokes thanks to complex visual optimization algorithm deployed in his AI.[8]
Other projects have seen computers compose musicals i.e. the first-AI-generated-musical which was presented at the London Art Museum on 22 February-5 March 2016[9], develop music products and write poems i.e. Annie developed by David Cope, a Musicology Professor from California Santa Cruz University, who was able to constantly develop and alter its musical taste according to the feedbacks she got from the audience and moreover, though she was not trained to write texts, she has not only composed musical works but also wrote Japanese “Haiku” poems[10] in 2011 that are still being sold online as a mixed poem album entitled as “Comes the Fiery Night” at “Amazon.com” e-commerce platform[11].
Given the above samples of pure AI-generated-creations, the AI embedded computer program evolves into being the generator more than merely being a tool and actually begins to make many of the decisions involved in the creative process without human intervention which indispensably retrieves the question whether AI-generated-creations should be found eligible to IP protection as such and AI systems should be qualified as sole or at least joint-owner/inventor/author of an intellectual product susceptible to patent and/or copyright protection.
By virtue of the above assessment, we shall underline the fact that this Article focuses more on pure AI-intellectual outputs generated mostly without human-machine collaboration and only autonomous agents -that crossed autonomy threshold[12] – contribute to the creative process.
In respect of the pure AI-creations though there are different standpoints in the teachings and jurisprudences of the Anglo-Saxon and EU countries, the dominant differentiating views are mainly clustered upon two cornerstones of the subject matter where in one corner, IP protection is simply denied on basis that AI is deprived of legal personality and subjectivity thus falls into public domain and in the other corner, the ownership is attributed to the creator/developer of AI system.
Having said that, as is the case now, IP rights are restricted to natural persons and legal entities such as companies. In 2018, after receiving a strongly worded letter from more than 150 experts in AI, robotics, IP and ethics, the European Union abandoned plans to consider the creation of a third type of entity, an “electronic personality” for AIs[13]. However, as defended by some of the commentators in IP law around the globe, it’s certainly possible that, one day, the concept of “electronic person”, thus far rejected by the European Union, might come to exist or according to our opinion, for public policy reasons it may at least be considered appropriate to legislate to allow an AI entity to be named on a patent application as contributing to an invention to make it clear to third parties the invention was made using an AI tool.
Nonetheless, even on a theoretical basis, accepting AI systems as the proprietors of especially pure AI-generated-creations brings along two significant predicaments in which the author of this Article and many of the commentators of IP law around the globe seem to have difficulty to come up with a convergent way-out suggestion to this ambiguity since the global acquis of intellectual property has not been adapted in a way that would tackle those predicaments yet.
Predicement on Liability
One of those predicaments is the AI’s liability issue. As an AI system lacks legal personality, it cannot incur liability and brings with it the so-called “black box” problem. Unlike code written by a person, some AI systems store their information in a form that cannot easily be read by humans or reverse engineered. It can therefore be impossible to discover why a system made a particular decision or produced a particular output. In such cases, liability is likely to fall upon the person or entity who controls or directs the actions of the AI.[14]
However, in such a case, often that is not clear where one party has created the AI and another has decided what data to put into it or what questions to ask it. Besides, as indicated herein above, AI systems can also generate new works protectable by copyright, such as creating new artwork or music. However, when it comes to copyright regimes, most copyright statutes do not yet clearly define who owns machine-generated works. It is currently a point of contention in respect of such works whether the work is generated by a machine or the role played by the humans in creation of the work. In order to clarify these contentious matters, agreements should attempt to clarify ownership when possible. Further, an AI system may act or operate autonomously in a manner that infringes third-party IP rights. If existing laws do not extend liability to a machine, then a related stakeholder (such as the owner, developer, operator or another supply chain participant) may be responsible.
To that extent, in May 2018, a meeting was organized by WIPO to discuss these AI applications, encourage their sharing and the exchange of information and as a consequence of this meeting, a “Draft Issues Paper on Intellectual Property Policy and Artificial Intelligence” has been published and disseminated to relevant non-governmental organizations i.e. AIPPI for purposes of collecting their opinions.[15] Under this context, WIPO was encouraged by its member states to collate and examine the “significant government instruments of relevance to AI” to create potentially harmonized standards and solutions to be adopted by the global acquis of the intellectual property.
Patent and copyright codes are the most relevant systems of protection regarding AI, especially when inventions can be autonomously generated by AI. There are several reported cases of applications for patent protection in which the applicant has named an AI application as the inventor.[16] In that regard, in view of “Draft Issues Paper on Intellectual Property Policy and Artificial Intelligence”, WIPO had raised some elaborative queries that may be briefed as follows: Should the law permit that the AI application be named as the creator/inventor, or should it be required that a human being is named as the inventor? If so, should the law give indications of the way to determine the human inventor or let the stakeholders take this decision by private arrangements? Moreover, who should be recorded as the owner of a patent involving an AI application? Are current legal provisions sufficient to consider the specificities of inventions generated by AI or should specific legal provisions be introduced? On the other hand, should the law exclude the availability of patent protection of autonomously generated inventions by an AI application? How would the liability issue be solved if AI application be named as the creator/inventor?
In our opinion, only after rendering compatible, progressive and unconventional replies to above queries in view of the requirements of technological developments in the era of Industry 4.0, issues related to AI’s legal personhood and civil/criminal liability may be addressed if the governments refine their IP legislations in a way that AIs can be held responsible and liable of his acts and creations which infringe others’ IPRs. Respectively, policy makers may think of establishing new forms of legal personhood in cases of complex distributed responsibility, accountability and liability for the activities of AI systems in contracts and business law.
On the other hand, as we have mentioned in the preceding paragraphs, one may rely on the liability of the stakeholders (such as the owner, developer, operator or another supply chain participant) should existing laws not extend liability to a AI system, case law with regards liability in keyword advertising and online marketplaces can be taken as a precedent.
In the Google France keyword advertising cases and in the L’Oréal v eBay International online marketplace case, Google and eBay were not held liable for infringing activity which they did not take an active part in. Unless a defendant is on notice of infringing activity it is likely that similar reasoning would be applied to an AI assistant which had bought an infringing product. However, the owner of an AI assistant, i.e. the company behind the assistant such as Amazon with Alexa, will be held liable if they have active knowledge of the infringing activity. Indeed, Professor Luciano Floridi (a Professor of Philosophy and Ethics of Information at the University of Oxford) has suggested that one might look to Roman slave law for guidance on the issue of liability, where the concept of the ‘intelligent slave’ under Roman Law transferred liability to the owner. Applying this to AI, the owner/designer of the AI system would be the owner of the intelligent slave, and thus liable.[17]
In the light of above, though we foresee that AIs may be acknowledged as (at least joint) proprietors of their own creations which are deprived of an human intervention throughout the pathway of creation in the near future, intellectual outputs created by AIs seem to be awarded to their developers regardless of any ownership on the AI itself since the developer is the primary source of creativity in terms of the relevant AI considering that its code is written by that person and that such award would be an incentive for developers to work forward on AI technology.[18]
Predicament on AI Investment
One of the other predicaments of entitling AIs as the proprietors of pure AI-creations would be the commercial implications imposed upon the investors who devote their time and money on the development of AIs and their abilities in data analytics. This is also the case for pure AI-creations that would fall into public domain in the event IP protection is not deployed to such creations.
The manner in which the law tackles new types of AI-driven creativity could have far-reaching commercial implications. Artificial intelligence is already being used to generate inventions and works as stated above. These inventions and works could in theory be deemed free of copyright and patent right because they are not created by a human inventor or author. As such, they could be freely used and reused by anyone. That would be very inconvenient for the companies commercializing the works and inventions to see that the creations are not actually protected by law and can be used without payment by anyone in the world.
While it is difficult to ascertain the precise impact, this would have on the creative economy, it may well have a chilling effect on investment in automated systems. If developers doubt whether creations generated through machine learning qualify for copyright protection, what is the incentive to invest in such systems?[19]
In this sense, most of the commentators of IP Law merge in the view of granting copyright or patent right to the person who made the operation of artificial intelligence possible, in the line with the UK’s model.[20] According to the commentators such an approach will ensure that companies keep investing in the technology, safe in the knowledge that they will get a return on their investment.
“….The companies along with the AI developers they financially support are “the most important contributors to the research and development of the AI sector. Without their contribution, AI devices would simply not be available for use by the general public”. Additionally, the technology is known to better develop and evolve when given enough data and time to analyse and for trial and error. In that respect, it is almost surely these large companies that may carry this technology to its utmost possibilities. In that light, these companies would inevitably demand legal guarantees and protection for their investments and would ultimately stop working on this technology or keep their findings, researches and systems secret fearing loss of potential profit. Therefore, currently it seems that the best option is to provide a way for legal entities to own works created by AIs that they own but to limit the legal protection provided to them in such a way that the scientific development is not hindered…”[21]
Moreover, in terms of copyright and patent laws, the general rule is that the one who creates the work is regarded as its author, not the one who physically makes the work. However, an exception to this rule is employer’s copyright/patent right, applicable in Turkey and EU countries, which is comparable to works made for hire in the United Kingdom and the United States. Employer’s copyright/patent right makes it possible for the employer to be regarded as the (notional) author if the employee created the work during the scope of his employment. An analogous application of works made for hire for AI is suggested in literature. This would give the company an incentive to continue investing in innovation.
Notwithstanding the above suggestions, unless the existing IP law systems disrupt and alter their conventional approaches and ways of enforcing their IP rules when it comes to AIs under a sui generis legal system, one should not expect harmonization of laws either in view of adoptation of AI-Outputs in general under the umbrella of IP law or sole or joint proprietorship of AIs over AI-Assisted and pure AI creations.
As a conclusion, it shall not be erroneous to say that a lot is still unclear in the field of intellectual property protection of AI-generated creations. Since in terms of acquis of IP, no legal personhood has been attributed to AI, the question whether it can be holder/owner of any (intellectual property) rights or be held liable for his own actions performed without human intervention can, for now, only be answered negatively. Even worse, in cases where no inventor or creator can be identified, the work will, in principle, fall into the public domain upon which the companies may refrain from investing in the technology. Thus, in view of AI-generated-IP outputs derived mostly without human-machine collaboration, we find it more fair, compatible and advisable to refine and harmonize the global acquis of IP and International Conventions such as TRIPs gradually along with the upcoming developments in a way that would render AIs “an electronic personality” which could possess limited ownership rights and liabilities jointly with their right holders mainly for purposes of overcoming the problems such as “black-box” problem and would incentivize the companies to invest on AI technology uninterruptedly.
* Co-founder of Ozden & Guclu Legal, Specialized in IP Law, www.ozdenguclulegal.com.
[1] Prof.A.Lakshminath & Dr.Mukund Sarda, Digital Revolution and Artificial Intelligence- Challenges to Legal Education and Legal Research, CNLU LJ (2) (2011-2012).
[2] Artificial Intelligence: A European Perspective, Joint Research Centre, EUR 29425 EN, 2018.
[3] S. Russell and P. Norvig, Artificial Intelligence: a Modern Approach, Pearson, 3rd edn., 2016 (introduction).
[4] AI is Changing How Products are Bought and by Definition Will Change Trade Mark Law, Lee Curtis and Rachel Platts, HGF, http://www.iprhelpdesk.eu/news/guest-article-ai-changing-how-products-are-bought-and-definition-will-change-trade-mark-law.
[5] Nothing Artificial About This Intelligence: AI Meets IP, Sam Baxter, Chief Technology Officer, IP.com
[6] Boolean search is a type of search allowing users to combine keywords with operators (or modifiers) such as AND, NOT and OR to further produce more relevant results. For example, a Boolean search could be “hotel” AND “New York”. This would limit the search results to only those documents containing the two keywords.
[7] There are basically two types of AIs: Strong and weak AIs. With strong AI, machines can actually think and carry out tasks on their own, just like humans do. With weak AI, the machines cannot do this on their own and rely heavily on human interference and submission of Training Data. Strong AI can process and make independent decisions, while weak AI-based machines can only simulate human behaviour.
[8] Yanisky-Ravid, s. 662; http://graphics.uni-konstanz.de/eDavid/.
[9] Lyn Gardner, “Beyond the Fence Review – Computer- Created Show is Sweetly Bland”, 28 February 2016, https://www.theguardian.com/stage/2016/feb/28/beyond-the-fence-review-computer-created-musical-arts-theatre-london.
[10] Yuval Noah Harari, Homo Deus, Kolektif Kitap, 1. Press, translate; Poyzan Nur Taneli, İstanbul, 2016, s. 338-339.
[11] https://www.amazon.com/Comes-Fiery-Night-D-Cope/dp/1466219157.
[12] What this means is that when an AI machine, whether in the form of a robot (such as a drone or self-driving vehicle), or a bot (meaning “smart software”) or other software-implemented AI system makes choices, the legal situation changes; Daniel J Gervais, ‘The Machine as Author’ Iowa Law Review 105/2019 (Vanderbilt Law Research Paper No 19-35) https://ssrn.com/abstract=3359524.
[13] Should AI own their own IP?, RCNT.EU/758Q2, EMMA WOOLLACOTT.
[14] An Intelligent Approach to AI and Intellectual Property, Tom Lingard, Partner at Stevens & Bolton LLP, https://www.information-age.com/ai-intellectual-property-123483190/
[15] See WIPO, Conversation on Intellectual Property (IP) and Artificial Intelligence (AI): Draft Issues Paper on Intellectual Property Policy and Artificial Intelligence, WIPO IP/AI/2/GE/20/1, 13 December 2019.
[16] EPO refuses DABUS patent applications designating a machine inventor, 20 December 2019, https://www.epo.org/news-issues/news/2019/20191220.html.
[17] AI is Changing How Products are Bought and by Definition Will Change Trade Mark Law, Lee Curtis and Rachel Platts, HGF, http://www.iprhelpdesk.eu/news/guest-article-ai-changing-how-products-are-bought-and-definition-will-change-trade-mark-law.
[18] Questions of Intellectual Property in the Artificial Intelligence Realm By Gönenç Gürkaynak, Ilay Yılmaz, Türker Doygun, and Ekin Ince of ELIG law firm.
[19] Artificial Intelligence and Copyright, Andres Guadamuz, Senior Lecturer in Intellectual Property Law, University of Sussex, United Kingdom, WIPO Magazine, 2017
[20] This approach is best encapsulated in UK copyright law, section 9(3) of the Copyright, Designs and Patents Act (CDPA), which states: “In the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken.”
[21] Questions of Intellectual Property in the Artificial Intelligence Realm By Gönenç Gürkaynak, Ilay Yılmaz, Türker Doygun, and Ekin Ince of ELIG law firm
