How will the IP world respond to the rise of AI?
José Carlos Erdozain of PONS IP analyses the IP challenges raised by AI, looking at what AI seeks to do, how it can be used in the patenting process, how easily it can be protected and whether it should be considered an inventor
One of most challenging discussions in the field of exclusive rights is how artificial intelligence (AI) is impacting on the assessment of those rights. Particularly in the realm of patents and copyright works, experts wonder if AI is likely to alter some of the most fundamental pillars of exclusive rights. Think just for a moment about the fulfilment of the novelty and non-obviousness requirements (basic issues connected to patents) or the authorship of machines or robots in relation to literary or artistic works or even performances, not to mention the ethical consequences of the use of AI.
What is AI?
The concept of AI takes us to a reality other than just a computerised routine. AI takes us to machines that mimic cognitive functions normally associated with human minds. Even if we go further and think of deep or advanced AI, the fact is that AI engineers work hard to evoke human feeling and to imitate human characteristics such as love, sense of humour, wrath, mercy, a sense of justice, and above all abstract reasoning.
Human behaviour is achieved thanks to extremely complicated software routines and artificial nets. As a result, in many people's view, AI is really associated with technical or technological procedures that mimic human beings and with the computerisation of code source based on previously programmed reactions, which run as specific external causes appear.
Can a feeling be programmed?
Can a human feeling be programmed? If so, would it not be just an automatic technological routine (for example, when identifying a smile in a child, the reaction must also be that of a smile)? We may agree that the response of AI to an external stimulation may be programmed before, but in the same way that human beings have also been educated.
The difference between a pre-programmed mimetic response, and a spontaneous human reaction is the key to upgrading the status of androids or robots and, eventually, to establishing a rule of law that they also deserve rights.
The difference is an AI programmed with a number of pre-established responses, but with the capacity to learn and deduct by itself when faced with external experiences not previously taken into consideration by the human programmer.
Computerisation programming is being achieved nowadays (see IBM Watson or Apple's Siri) and will be probably be standardised in the coming decades.
AI is part of our lives, and in some cases AI-based robots are even like citizens (see, for instance, Deep Knowledge, which is a Japanese company which named a robot as a member of its board of directors in 2014).
How is AI actually impacting on IP rights?
AI is patent subject-matter. This is irrefutable. DeepMind is an example of this, a leading AI research company, which, founded in 2010, was acquired in 2014 by Google.
DeepMind has applied for several (PCT) patents.
The claims of these applications are mainly focused on neuronal nets and the achievement of a similar outcome to rational human thinking, as well as procedures of machine learning.
It is obvious to say that these PCT applications have not been granted yet, as they have not entered into the national phase. However, also unarguable is the fact that DeepMind will not have to fight with the examiners in order to defend the novelty and non-obviousness requirement, especially if the neuronal net seems similar to human beings' neuronal system. This example makes us think about patent rules.
It is important not to forget that machine learning systems are not always patent subject matter (at least in the EU and the US). A consistent consensus statement on this should be reached by patent offices, especially when their filters of what is and what is not patentable differ.
However, the fact is that some examples of AI have been patented, like the Creativity Machine developed by Stephen Thaler in 1994, which was capable of generating new ideas through artificial neural networks (US Patent No. 5,852,815, granted in May 1998), and John Koza's invention, based on AI, which was granted a patent in January 2005 (US Patent 6,847,851).
This leads us to the question of what the subject matter eligibility standard for AI is.
US Supreme Court Alice decision doctrine has been interpreted as saying that patent claims whose subject matter can be performed through "an ordinary mental process", "in the human mind" or by "a human using a pen and paper", with the limited exception of claims that specifically provide procedures to achieve technological improvements over tasks previously performed by people (e.g. containing an inventive concept) must be excluded from patent claims.
US courts have already discussed this issue, since they have refused the patentability of AI patents on the grounds that AI patent applications are based on a replication of human thinking. Similarly, see Blue Spike, LLC v. Google, Inc., stating that the intended patent application only had the purpose of covering "an abstract idea long undertaken with the human mind", because the claims sought to model the highly effective ability of humans to identify and recognise a signal on a computer.
Applications for AI patents have been granted in the field of medicine (e.g. by advising on the existence of skin tumours), as well as to help doctors in assessing the correct treatment for cancer. IBM Watson (AI on a computer) successfully diagnosed a woman suffering from leukemia.
Also to be taken into account is Natural Language Processing (NLP), that is, AI algorithms that enable a computer to understand and process human languages, or the Natural Language Searches (NLS), algorithms that identify contents that match topics and machine learning algorithms, a method of data analysis that automatises analytical model building, by using algorithms that iteratively learn from data. Machine learning allows computers to find hidden areas of research without being explicitly programmed to do so.
An example of this is Cloem, a French company, which has applied NLP technologies to assist patent applicants with inventing variants of patent claims. This is achieved by using NLP algorithms and text mining to artificially compose text for infinite patent claims which cover thousands of potential new patents.
Patentability and inventorship – AI inventions
Another topic is whether AI should be considered legally as an inventor of inventions. Should this be the case, notwithstanding that said inventions have not actually been invented by a human being?
Under the current legal framework, assuming that this type of AI invention has not been invented by a human being primarily, the inventor cannot be, legally speaking, an AI being. But should the law be changed so that it recognises AI as the inventor? Some argue that this step forward would accelerate innovation and promote economic richness, whereas many others contend that supplanting human invention with AI could lead to the suppression of highly qualified researchers.
Assessment of patent rights – patent offices using AI
Another field where AI is gaining importance is patent offices using AI in order to assess novelty and non-obviousness requirements.
Patent offices are facing a hard task having to evaluate these requirements. It is essential not to forget that the information on patent applications and already existing patents has increased vastly. Some figures give us a better understanding of this titanic effort: US applications climbed up to 10,000,000 in mid-2018. More than 100 million patent documents are related, and there are more than four billion indexed web pages that need to be examined in order to duly assess the novelty requirement.
The overwhelming task of any patent office to search in previously-known state of art is even worse when we realise that approximately 60% of the patent documentation worldwide is published either in Korean, Chinese or Japanese.
The Japan Patent Office has publicly announced that it is looking into AI technology "to automate processes such as screening patent, trademark and design applications". This deployment is intended to come into action in the April 2018 to March 2019 fiscal year.
Other patents offices (Europe, China, Korea, Japan and US) have correctly considered AI in the field of patent assessment as a top strategic priority.
Copyright and AI – authorship and creativity
There are already examples of AI created works of art. Is this copyrightable even if created by an algorithm? If so, who is its author, the AI or the programmer of it?
Google has created an AI program, which is capable of writing news articles. In 2016, museums and researchers in the Netherlands unveiled a portrait entitled The Next Rembrandt, an AI work of art created by a computer, whose process of creation was due to the analysis of thousands of works by the 17th-century Dutch artist Rembrandt.
A short novel written by a Japanese computer program in 2016 reached the second round of a national literary prize and the Google-owned AI company DeepMind has created software that can generate music by listening to recordings.
In my view, the issue is whether or not AI should be called an author in the sense of a copyright creator of a work.
The problem is that in many legal systems, authorship is essentially associated with the human condition. In addition, court decisions refuse to acknowledge authors for works created by software.
On the contrary, however, some legislation recognises the authorship of the programmer, and goes beyond the traditional legal frameworks, for example, Section 9(3) of the UK Copyright, Design and Patents Act, 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". Moreover, Section 178 of the same act defines a computer-generated work as one that "is generated by computer in circumstances such that there is no human author of the work".
Will the day come in which AI will be named a creator of a work?
Rules stated by the European Parliament
One final example of the importance of AI and how this intelligence is being taken seriously by parliaments and governments is that on February 16 2017 the European Parliament approved a resolution with recommendations to the commission on Civil Law Rules on Robotics (2015/2103(INL).
The European Parliament is conscious that humankind is on the threshold of an era where AI will surpass humans' skills and capacities, and this special moment requires the establishment of fundamental directives on how a future relationship between human beings and AI should be conducted.
This resolution is of significant importance because it establishes a series of rules, in particular those governing liability and accountability, which must guide the construction of AI, as well as ethical principles to be respected in the development, programming and use of robots and AI, so that this technology serves humanity and the benefits of AI are shared broadly.
José Carlos Erdozain
José Carlos Erdozain began his legal career in 1997 and has since developed strong expertise mainly in the field of intellectual property. He has broad experience in litigation, arbitration and business consulting, both on a national and international level. He has worked for companies in different industries, including the textile, audiovisual, creative, energy, construction, and publishing sectors. He has extensive knowledge of international contracts and arbitrations. He is the author of several monographs and articles focused on intellectual property. He is invited as a speaker to various conferences.
He is currently a member of ALADDA and AIPPI. He can speak Spanish, English and German.