In the latest in the series of worldwide cases in which Dr Stephen Thaler has tried to have DABUS, an AI system he created, recognised as the only inventor on a patent application, The New Zealand Patent Office (IPONZ) has found that an AI cannot be an inventor.
In the decision (Stephen L. THALER [2022] NZIPOPAT 2), the NZ Assistant Commissioner of Patents found that under NZ Patent Law, an inventor can only be a natural person:
“If the legislators had intended to allow granting of patents in New Zealand for inventions devised solely by non-humans such as artificial intelligences, or life forms other than human beings they would have drafted the Act to accommodate these possibilities specifically and explicitly. They did not do so”.
The patent application in question was titled a “food container and devices and methods for attracting enhanced attention” as described in PCT patent application WO2020079499. Dr Thaler, the applicant, developed and operated an AI system he named DABUS: Device for the Autonomous Bootstrapping of Unified Sentience.
DABUS is a “connectionist AI system” with interconnected and interactive neural networks. The first neural network randomly generates concepts/outputs and the utility is then assessed by a second neural network which then modifies the first network creating a feedback loop. Once trained, DABUS was left in an unsupervised generative learning phase and generated the concept of a cylindrical food container with a fractal profile as described in WO2020079499. On the patent application, Dr Thaler deliberately listed the inventor as “DABUS, the invention was autonomously generated by an artificial intelligence” (and does not claim to be an inventor).
This has required various patent offices around the world to grapple with the question of whether an AI can be validly listed as an inventor and thus whether a patent can be validly granted. So far US and UK courts have ruled that DABUS/AI systems cannot be inventors and the patent applications have been refused. The Australian Patent Office (IP Australia) was of a similar view, however this was then overturned by Justice Beach in the Federal Court. IP Australia has appealed to the Full Bench of the Australian Federal Court to further consider the issue. If you’re interested I’d direct you to the Patentology blog by Mark Summerfield who has dived into these decisions (and the issues) in some depth.
This case has now made it to the NZ patent office (IPONZ) where the Assistant Commissioner of Patents was forced to consider the case under NZ law. In the Assistant Commissioners view, inventors, as least as defined in the NZ Patents Act, can only be natural persons ruling out AI systems such as DABUS. Thus any such application is invalid and cannot be granted a New Zealand Patent.
I agree with the Assistant Commissioner’s view that such an important matter should be decided by legislators, and not left to the courts to interpret. This feels to me like a case of garbage in = garbage out. I strongly suspect that legislators were not considering this issue when drafting the definition of an inventor in the relevant acts, so any act of interpretation is fraught with danger. Forcing courts to make an interpretation risks a poor decision in the case of a poorly worded/loose definition and/or gives a judge the latitude to make what should be a significant policy decision (i.e. one for legislators).
In passing, the Assistant Commissioner commented that “So far as I have been able to verify, no creatures on earth other than human beings ever seem to have invented a patentable invention and we have never received any disclosure of inventions from other worlds.” So at least we know aliens haven’t secretly invaded and tried to patent their technological developments.
Personally I think this, and similar decisions finding that AIs cannot be inventors, makes a lot of sense. My view is that AI systems are essentially just complex mathematical/computational tools, and the inventiveness resides in how you use those tools. In the AI context this focusses the attention on how an AI is used or configured. e.g. what specific data do you train on or use, what specific pre-processing of data should be performed, and/or what specific parameters and algorithms are required to get the best performance out of the AI. These steps are typically under the control of the user or developer of the AI system who is focussed on solving a domain problem (i.e. something going on outside of the computer). Invention then resides in those persons using their specific domain knowledge to get the best out of their AI tool to solve their domain problem.
I also think it’s a good policy position. Patents are commercial tools to protect the time, effort and cost in developing new products and services (i.e. to mitigate R&D risk). Successful commercialisation of new products is hard and typically needs to be actively driven by a team. Inventors and their employers are typically invested in the problem and so are well placed to drive the commercialisation process. In contrast AI systems are not in a position to do so. While their owners may be in a position to drive commercialisation I don’t think its good policy to allow them to be granted patents simply because they own an AI system that generates new ideas. To me that feels like encouraging patent troll behaviour.
From a policy position, I think a finding that inventors can only be natural persons ultimately helps ensure the patent system remains a commercially focused system that encourages and support R&D activities.
