Imagine a world where artificial intelligence holds the key to groundbreaking inventions, challenging traditional notions of inventorship. Enter Dr. Thaler’s Dabus, an AI invention that has sparked a global debate, with the United States Patents Office (USPTO) at the center of the storm.
Artificial Intelligence (AI) refers to the development of intelligent systems that can perform tasks that typically require human-like thinking, such as learning, problem-solving, and decision-making. One widely cited definition of AI comes from the seminal 1956 Dartmouth Conference, where it was defined as “the simulation of intelligence in machines, specifically computer systems.”
With the unprecedented growth of AI, the debate over AI-inventorship has reached new heights, with Dr. Thaler at the forefront. Dr. Thaler has filed patent applications for his AI invention Dabus in multiple jurisdictions including the USA, UK, EU, Australia, India, South Africa, and others. While this article does not aim to delve into the wider debate, it focuses on the rejection by the United States Patents Office (USPTO) of DABUS application. The authors challenge and refute the reasoning put forth by the USPTO, advocating for the inclusion of AI inventions in the realm of patentability, and pushing for the opening of doors for AI-generated inventions.
USPTO vs DABUS: Demystifying the Rejection Rationale
Dr. Stephen Thaler’s patent application for AI system DABUS to be named as an inventor of fractal-based container for enhanced grip was rejected by the US Patent & Trademark Office (USPTO). Thaler argued that DABUS was the sole inventor of the invention. He claimed that DABUS had independently created the invention, and that it was not obvious to a skilled person in the relevant field.
However, the USPTO rejected the application with the rationale for the decision centered around two keys. First, that an AI cannot be construed as an inventor from statutory interpretation and legislative intent of the term “inventor.” USPTO did a conjoint reading of the terms ‘inventor,’ ‘individual’ and ‘whoever’ to conclude that the legislative intent while formulation of the text was to accord inventorship to a natural person. Any other interpretation would be contradictory to the purpose of patent statute.
Second, ‘conception’ of an idea cannot be attributed to an AI with or without human intervention. As per USPTO, conception is complete when an idea is formulated clearly in the mind of an inventor wherein this mental activity cannot be attributed to any non-natural person. Thereby, the threshold of inventorship consists of conception of idea in the mind of inventor and operative intention applied in practice wherein the former part is inapplicable to an AI. Hence, USPTO concluded that because DABUS is not a natural person, it cannot be an inventor under US patent law.
Artificial Intelligence and Online Dispute Resolution
The Clash of Old Text and New Technologies
AI’s feat of accomplishing complex or insurmountable tasks is truly remarkable, and the pace of accomplishment alone is a feat that was not foreseen by anyone outside of pioneers like Alan Turing. At that time, the idea that machines could mimic human intelligence, let alone demonstrate creativity to a comparable level, was simply not obvious.
The USPTO’s argument that only natural persons can be granted inventorship rights is based on the premise of the legislative intent behind the patent legislation, which dates back to 1952. However, this argument is weakened by the fact that it is unreasonable to assume that the drafters of the text had envisioned AI as a creative inventor. At the time, the idea of machines being able to invent like humans was unimaginable. Moreover, the language of the patent law, including the term “individual,” was originally intended to ensure that patent rights were granted to actual inventors rather than corporate entities.
As argued by Mimi S. Afshar, the “individual” centric language was intended to distinguish between actual inventors and corporations. In other words, it was meant to prevent the patent system from being exploited by entities that did not actually create the invention. Therefore, the USPTO’s argument inventorship to only natural persons based on legislative intent is unconvincing.
Redefining Creativity: AI’s boundless Idea Generation
USPTO’s reasoning capability to conceptualize an idea is a human activity is obsolete. It limited this mental activity to natural persons and here’s why it is outdated. First, its construction of conception of an idea is archaic. AI’s blackbox theory states that in complex machine learning systems, the internal workings of generating an output in a machine are opaque or a ‘black box’ to outsiders, thus limiting conclusions on whether AI can conceptualize an idea. However, AI has demonstrated a capacity for conceptualizing by generating unique outcomes.
Furthermore, AI is powered by Artificial Neural Networks (ANNs) and Deep Learning, designed to mimic the structure, and functioning of the human brain. ANNs are computer systems modeled after the architecture of the brain that can process large amounts of data, learn from it, and make decisions based on that learning like functioning of human brain. Therefore, AI can conceptualize ideas and generate unique outputs, powered by ANNs and deep learning. The limited understanding of AI’s internal ‘thought’ process and its ability to create unique ideas/outputs should be considered capable of conceptualizing an idea, just like a human inventor.
Second, assuming arguendo, even if AI cannot conceptualize an idea, in a very recent US judgement. In this case, a non-inventor was named as a co-inventor without physically constructing the invention. The court laid down that a person who contributes to an invention can be an inventor if they played a key role in conceiving the idea that ultimately led to the invention. Thus, this decision supports and advances the argument in favor of non-human inventors and broadens the scope for legal recognition of AI as inventors.
Global Legal Stance
South Africa has granted the world’s first patent for an invention created by an AI machine called DABUS, making history. It is important to note that South Africa does not have a comprehensive patent examination system, and its patent laws do not define “inventor.” In contrast, Australia’s Federal Court has overturned the Australian Patent Office’s rejection of granting inventorship to an AI machine, ruling that an artificial intelligence system can be an inventor. However, a recent judgment delivered by the Full Court of the Federal Court overturned this decision, stating that the inventor must be a natural person in an application for a patent.
Similarly, the European Patent Office and the UKIntellectual Property Office reject granting inventorship to an AI machine, stating that “inventor” refers only to a natural person. The UK Court of Appeal upheld the decision of the High Court and ruled that an AI system could not be named as an inventor on a UK patent application. The court noted that recognizing an AI system as an inventor would be inconsistent with the purpose of the UK patent system, which is to incentivize and reward human inventors.
The Indian viewpoint aligns with that of the United Kingdom, which opposes the granting of patents to artificial intelligence. The justification for this perspective is based on the provisions of the Indian Patents Act, 1970. Sections 2(1)(s) and 2(1)(y) imply that the inventor must be a human being, even when not explicitly stated. Additionally, Section 6(1)(a) specifies that the “person” making the patent application must claim to be the original and true inventor.
Section 27 of the Indian Evidence Act: Proviso as an Exception
The UK, EU, and India have all maintained their rejection of a patent application, while the US court has also rejected it, albeit with a strict and outdated interpretation. Despite the authors’ viewpoint of inventorship recognition to AI-inventions, it is undeniable that no judicial system is equipped to deal with complexity of the legal issues surrounding AI-inventions. Despite weak and questionable reasoning by the USPTO, refuting it alone won’t open doors for AI inventions. The onus now lies in the legislative domain to draft and amend laws that recognize and legally acknowledge AI-generated inventions.
In conclusion, the legal, ethical, and philosophical implications of granting AI inventorship are undeniable. While some can be addressed, not all can be fully resolved. It is time to draw a conclusion and find a middle ground. If AI-generated inventions are not ready for human inventorship status, a new class, such as machine-based patents, should be established. With the increasing prevalence of AI-generated inventions, ignoring the implications would be economically unwise.
This article is authored by Shambhavi Sharma and Harshita Rawat, 4th year students at the Institute of Law, Nirma University, Ahmedabad