[This post has been authored by Lavanya Jha (West Bengal University of Juridical Sciences, Kolkata) & Shreya Jha (Amity Law School, Delhi).]
The term Artificial Intelligence (AI) was coined by American computer scientist John McCarthy during the Dartmouth Conference on Artificial Intelligence in 1956. It was understood as a system of solving complex problems through reasoning, knowledge representation, planning, navigation, natural language processing, and perception. The shared conception regarding AI has been that it is a method of data processing, wherein the lack of involvement of a processor allows it to have an independent “mind”. Therefore, a processing device like a digital computer through which AI related tasks are accomplished can be viewed as a fundamentally detached, objective observer whereas intelligent behavior can be viewed as a determinate set of independent elements. AI’s primary features can be characterized as unpredictability, rationality, independence, efficiency and accuracy, thereby allowing it to create “patentable” inventions.
The Indian Legal Framework
Section 6 of the Patent Act prescribes that only ‘true and first’ inventors and other persons deriving right therefrom are entitled to apply for the grant of patents. Although the Act does not lay down the essentials of an inventor, precedents lead to the finding that an inventor should a) Be a natural person; and b) Contribute “ingenuity, skill or technical knowledge” towards the invention.
In case of AI inventions, the first criterion is not met. Therefore, an AI system cannot be granted a patent. However, in the technology driven market, it can be argued that the term ‘person’ should include artificial intelligence. This view is in consonance with WIPO’s definition of patent i.e. an exclusive right granted for an invention, which is a product or process that provides, in general, a new way of doing something, or offers a new technical solution to a problem. In this view, there is no bar to an AI system claiming patent rights as it attributes creation to a “mind” and not to the human mind in specific.
The AI-Patent Law “Collision”
The AI-Patent Law “collision” stems from the ever-expanding paradigm of AI inventions and the ‘inflexible’ patent regime. The growing gap between the two can be explained through the following examples of AI inventions.
In 1998, John Koza developed an algorithm for circuit designs. During the course of his work, Koza was able to build a cluster of a thousand personal computers that generated 36 ‘human competitive’ designs. Similarly, scientist Stephen Thaler’s creativity machine is an AI system which “brainstorms” new and creative ideas after perturbing with the connections within its “neural network”. Dr. Thaler introduced the Machine to his favorite song and it was able to write 11 thousand new songs in a weekend. Further, IBM’s new invention Philyra is set to revolutionize the perfume industry after its release in 2019, as it can create a new fragrance after sorting through 1.7 million data. Considering that the AI system is to be credited as the “inventor” in the aforementioned cases, it is appropriate that these systems should be entitled to a patent.
At present, the potential of AI has been acknowledged as a “problem solver”. For e.g., chemists in the pharmaceutical sector have to synthesize and test countless variants. However, using AI brings about several other benefits such as quicker and more efficient drug discovery, new methods of drug manufacture etc. These benefits can be attributed to the fact that AI “thinks differently” which ascribes the intelligence quotient to it.
The benefit of AI system as an Inventor
It is argued that the “inventorship” granted to the AI systems will guarantee the creation of a substantial value independent of the value of the system itself and it would reward human activity manifold.
For example, if a Human X instructs his/her AI to develop an improved Google Assistant and based on such instruction a “design for an improved Google Assistant” is developed, it would be unfair to credit X when s/he has not contributed significantly to the innovation and has merely given instructions. On the other hand, since the AI has developed the design but is not recognized as an inventor, the first person to understand or recognize the design could become the inventor. Therefore, an AI system should be treated as an “inventor” and its inventions should be patentable. Although systems are not capable of appreciating the importance of patents, the suggested reform will provide an impetus to scientists to produce better and more capable machines.
A Case for AI as an Inventor by removing barriers for Inventorship
The general rule is that the inventor is the owner of the patent and enjoys the rights, titles, and interests of the patent. In case of inventions, the patent rights can be transferred to the assignee who acquires the right to prevent others from making, using or vending the innovation.
In the case of inventions by AI, a system of “default assignment” may be created where the AI systems should be recognized as “inventors “and the “ownership” rights should vest with the owner of the AI system. There may be following multiple entities involved with regard to this process i.e. first, the owner, second, the programmer or developer and third, the user.
In the case of owner versus developer, if the developer is given the “default assignment” of the patent, he is still liable to transfer the computer to the rightful owner. However, he would still be liable to logistically monitor the inventions that are being made by the AI system he developed. This would be unfair to both parties involved, as the developer would then be responsible for monitoring the logistics of the system he no longer owns while the owner is deprived of the incentives that his/her investment was supposed to generate. In such a situation, it is suggested that the owner of the AI system may enter into a contractual arrangement with the developer.
In the case of owner versus user, it is a matter of fact that there may be multiple users which might interact with the AI system at once and if the “default assignment” is made to the user, the owner company will not get incentive to create and it might come up with restrictions on user access. Therefore, in both the cases, the owner of the AI system is the best option for “default assignment”.
Granting “inventorship” to the “ingenious” AI systems requires serious consideration to be given to the issue by law and policy makers. The preceding sections serve to demonstrate that AI systems should be treated as “inventors” and their inventions should be considered “patentable.” In order to reflect and modify the status quo, an attempt has been made to identify and remove the barriers which can result in AI systems being treated as “inventors.” Both positive and negative implications have been addressed to establish the effectiveness of introducing inventorship for AI systems, and regulatory measures have been proposed to create a holistic process for granting recognition to AI systems as inventors of patentable products.