Generative AI has ushered in a new era of innovation and creativity. While much of the discussion has focused on how we create, how we invent is also set to change.
As generative AI continues to evolve, patent law and practice has not escaped disruption.
While generative AI is making its mark across most industries, giving rise to a new era of creation and unbridled innovation, it is also changing the way patents are drafted, the patent clearance process, patent prosecution and enforcement strategies. However, as the cutting-edge technology intersects with the intellectual property framework, questions about the use of generative AI also arise. As the power of generative AI is increasingly leveraged during the innovation process, it will become increasingly important to carefully navigate and reflect on the relationship between patents and AI.
A new age of invention is being invented.
A collaborative approach between human inventors and generative AI technologies has the potential to foster a dynamic environment for innovation.
Generative AI can significantly accelerate drug discovery by analysing chemical structures, predicting potential activity and interactions, and proposing novel molecular configurations.
Clinical-stage pharmaceutical companies such as Algorae Pharmaceuticals collaborates with UNSW Sydney AI Institute to develop generative AI models for discovering new novel drug combinations. Pharos iBio collaborates with the University of Sydney Drug Discovery Initiative to use the Chemiverse AI platform. This platform accurately identifies hit compounds based on 3D target protein structure and quantum mechanical energy calculations.
Google used generative AI to develop a new tool called Automated Retinal Disease Assessment that more accurately interprets retinal scans to detect diabetic retinopathy.
In materials science, where the quest for novel materials with specific properties is ongoing, generative AI can suggest new combinations and configurations. This is particularly beneficial in industries ranging from electronics to construction.
In the computer industry, IBM uses generative AI design for sustainable components in microchips. Photoacid generators (PAGs) are photosensitive materials used in manufacturing computer chips. The chemical formulas of current PAGs are not environmentally sustainable and there is a pressing need for greener alternatives.
In the world of EV batteries, Aionics uses generative AI to run experiments on an existing database of billions of known molecules to accelerate catalyst discovery for EV batteries.
Generative AI can contribute by generating ideas (or seed ideas) for sustainable technologies, renewable energy solutions, and eco-friendly materials. From optimizing energy-efficient processes to proposing breakthroughs in recycling, the AI-powered creative engine can play a pivotal role in advancing environmental sustainability.
In the renewable energy industry, generative AI is used to make accurate energy production predictions based on historical data and weather forecasts. A physics-based generative AI model, NowcastNet can predict extreme rain with a longer lead time than existing conventional methods.
Generative AI also allows energy companies to distribute loads evenly based on real-time demand data. For example, an AI-driven virtual power plant (VPP) is a network of decentralized generators, flexible consumers, and storage units that manage the flow of electricity from distributed energy resources such as solar panels, batteries, and electric vehicles. VPPs balance the grid, integrate renewable energy, and reduce peak demand.
The irony of generative AI assisting in its own evolution is not lost in the realm of artificial intelligence and robotics. Generative AI can contribute to the development of more advanced algorithms, efficient robotic systems, and innovative applications of machine learning.
In the automobile industry, software companies such as Nvidia leverage generative AI and Neural Radiance Fields to create realistic 3D environments from videos recorded by cameras installed on cars. This can be used to train models for self-driving vehicles. AWS DeepRacer employs reinforcement learning to optimise navigation and decision making.
In manufacturing, generative design software workflow uses an iterative design process and different algorithms to explore design options and rapid prototyping and testing.
Is it possible to use generative AI to invent?
As canvassed in our previous alerts, patent applications must still name a human inventor.
Dr Thaler developed the generative AI machine DABUS, which itself produces inventions. In Australia, one of the formalities requirements that must be met on filing is that a natural person (an individual, body corporate or body politic) must be named as the inventor in the application. As Dr Thaler could not name such a person, the application was not accepted, and lapsed after Dr Thaler’s appeals were exhausted. [1]
In Australian law, the use of generative AI in inventorship is relevant to the issue of “entitlement”. Only certain people are entitled to be granted a patent:
- The inventor.
- A person who would, on the grant of a patent for the invention, be entitled to have the patent assigned to them.
- A person who derives title to the invention from the inventor or from a person who would be entitled to have the patent assigned to them.
- A person who is the legal representative of a deceased person mentioned above.
The “inventor” is the lynch-pin in this concept. If there is no inventor, then no person is entitled to claim ownership of the patent.
However, despite its importance, the term “inventor” is not a defined term in the Patents Act 1990 (Cth). To determine whether a person is an inventor, it is relevant to ask:
- What is the relevant “inventive concept”?
- Would this have occurred without human involvement?
- Did the (human) contribution beneficially affect the final concept of the claimed invention?
- Would the final conception of the invention be less efficient without the contribution?
Previously, these issues largely arose in the context of disputes relating to entitlement to a patent, often in the context of multiple inventors who contributed to the claimed invention, rather than in the context of a suggestion that there was no inventor at all![2] Nonetheless, under the current law in Australia, assessment of a person’s contribution to the invention will be crucial.
Dr Thaler filed a number of patent applications including in Australia that named DABUS as the inventor of such inventions. Dr Thaler appealed to the Federal Court. At first instance, Justice Beach set aside the Deputy Commissioner’s determination, finding that an AI system can be named as an inventor in an application for the purposes of the Patents Act 1990 (Cth). The Commissioner appealed that decision to the Full Federal Court. An enlarged bench consisting of five Judges overturned Justice Beach’s decision and unanimously found that the inventor listed in an application must be a natural person. Dr Thaler then sought for special leave to appeal the Full Federal Court’s decision to the High Court, which was refused on the basis that that the matter was not the “appropriate vehicle” to consider the question of whether an AI system can be an inventor. The application subsequently lapsed.
Since Raising the Bar, s 22A of the Patents Act provides that a patent is not invalid merely because the patent, or a share in the patent, was granted to a person who was not entitled to it or the patent or a share in the patent was not granted to a person who was entitled to it.
The question of whether the [DABUS Application] has a human inventor has not been explored in this litigation and remains undecided. Had the question been explored, it may have been necessary to consider what significance should be attributed to various matters including the (agreed) fact that Dr Thaler is the owner of copyright in the DABUS source code and the computer on which DABUS operates, and that he is also responsible for the maintenance and running costs.
- Commissioner of Patents v Thaler [2022] FCAFC 62 at [121]
Provided there is a human inventor, there is no reason an invention cannot be invented using generative AI.
Generally, patents must disclose the "best method" of carrying out the invention. This can be challenging for AI-generated inventions due to the difficulty in explaining exactly how the generative AI system arrived (or performs) the invention.
Patents must also enable others to make and use the invention. This can also be challenging for AI-generated inventions since the inventor or drafter of the specification may not be able to fully explain how the AI system works. AI systems that use machine learning are a particular risk (as they have the potential to generate inventions that are difficult to replicate or explain).
Use of generative AI in the patent drafting process
Generative AI has the potential to automate some aspects of the drafting process. From generating technical descriptions to suggesting optimal language, AI systems can contribute to the readability, efficiency and accuracy of patent application drafting. The integration of generative AI in patent drafting also has the potential to elevate both the quality and efficiency of the drafting by synthesising a vast corpus of existing patents and prior art.
If generative AI continues getting better at developing and drafting patent specifications, what does this mean for the future of patent specifications?
— IP Australia, Patent Provocation
Use of generative AI in searching
The quest for relevant prior art can be like searching for a needle in a haystack. As the volume of relevant patent literature continues to surge, the need for a more efficient and accurate patent search process becomes imperative.
A comprehensive patent search has traditionally involved hours of painstaking manual review. However, generative AI has potential to introduce a paradigm shift by automating this process, leveraging deep learning algorithms to analyse extensive patent databases swiftly and accurately. By automating the preliminary stages of prior art analysis, practitioners can redirect their focus to more intricate aspects of patent examination.
Generative AI systems, like OpenAI's GPT-3, have the potential to analyse and interpret natural language and context, and construe complex search queries more intelligently. This capability will increasingly enable nuanced analyses of patent databases, uncovering data that might be overlooked in traditional searches.
Moreover, these systems can adapt to evolving search criteria, continuously learning and refining search patterns.
What does IP Australia say about generative AI
Although generative AI is not yet a mature technology, IP Australia recognises that it will evolve quickly and have a material effect on the IP system (across all the IP rights),[3] which, in its current state, may not be fit-for-purpose. IP Australia has recently undertaken extensive work to understand and explore the potential impact generative AI may have on the Australian IP system. Based on interviews with relevant stakeholders and research into the capabilities of the new technology, IP Australia published on 3 July 2023 its “Generative AI and the IP System – Exploring possible futures in the context of IP rights” exploratory paper (Exploratory Paper) and “Generative AI and patents: a provocation” (Patent Provocation).
The principle of incentivising and rewarding effort for innovation through the grant of monopoly rights is a cornerstone of the patent system – the same principles may not apply to AI-generated inventions.[4] If human contribution continues to be required to protect AI-generated inventions, innovators may be less inclined to utilise the patent system and simply not disclose inventions that are beneficial to industry and society more generally. [5] On the other hand, does patent protection of AI-generated inventions that do not involve human contribution devalue human creativity? [6]
One solution IP Australia is considering is the creation of a sui generis IP right for AI-generated inventions with a shorter monopoly, additional disclosure obligations and mandatory licensing requirements.[7]
Impact on small and medium-sized enterprises
IP Australia considers that generative AI may assist small-to-medium sized enterprises (SMEs) to access the patent system and compete more effectively with larger entities.[8]
However, there are also significant risks arising from use of generative AI, and SMEs will require support to ensure they make the most of generative AI while navigating risks, including:[9]
- The potential for unintentional infringement of protected inventions or data.
- Hallucination (when generative AI generates non-factual information).
- Reliance on biased datasets that overlook potential prior art.
- Unintentional exposure of an invention to the public before filing of a patent.
Congestion of the patent system
While generative AI will drastically reduce the time and effort required to draft lengthy and highly detailed patent specifications, there is a risk of flooding patent offices,[10] and lengthening the time it takes to obtain a granted patent.[11] Huge volumes of content generated by generative AI will also significantly increase the prior art that needs to be considered by both patent offices and others trying to navigate the system - patent thickets, defensive patenting and defensive publication will increase and have the potential to disincentivise innovation.[12]
Patent Provocation pg 23.
Patent Provocation pg 23.
Patent Provocation pg 24 and 25.
Patent Provocation pg 24.
Patent Provocation pg 16.
Patent Provocation pg 17.
Patent Provocation pg 4.
Patent Provocation pg 10.
Patent Provocation pg 8.
Certain actors in the ecosystem may be interested in making it harder for others to get protection in a particular technology area. As such, they train their AI model on existing patents in the technology area … Companies, utilising defensive publication, or patent trolls keen to suppress competition can use similar techniques to rapidly create a multitude of patent specifications crowding particular technology areas or industries.
— IP Australia, Exploratory Paper.
IP Australia is considering a number of potential responses, including regulatory interventions to discourage flooding the system[13], stronger enforcement powers to respond to bad behaviour,[14] and legislative changes, including:
- The application of more stringent requirements for inventive step and enablement.
- Limiting what constitutes prior art to filter out mass AI-generated content.[15]
- Introducing a “use” requirement similar to that in the trade mark system.
Conclusion
Generative AI is everywhere. From health and life services to the material sciences, from green technologies to the robotics industry, generative AI continues to transform the way we live and work. In the pharmaceutical industry, generative AI significantly increases speed and efficiency of the drug development process and clinical trial analysis. In material science, generative AI techniques enable researchers to efficiently explore novel and improved materials. Generative AI also has the potential to drive ESG and sustainability targets as well as robotics systems to improve efficiency and productivity.
Despite the risks and challenges, generative AI remains a fascinating tool that will revolutionise patent law and practice — and the process of inventing! Generative AI will increasingly promote the generation of new and innovative ideas and we expect to see more and more patent applications to protect AI-generated inventions, as well as patents drafted with the assistance of generative AI.
Patent Provocation pg 12.
Patent Provocation pg 34.
Patent Provocation pg 34.
