Late last month, South Africa's Companies and Intellectual Property Commission (CIPC) became the first Patent Office in the world to award a patent that names an artificial intelligence as the inventor of a product. In the application by Dr Thaler, an artificial intelligence system (AI system), which has been described as a device for the autonomous bootstrapping of unified sentience (DABUS), was named as the inventor. The decision is significant considering that the applicant's previous applications in the UK, EPO and at the USPTO were unsuccessful because DABUS was named as inventor.
In each of these jurisdictions, the question was whether the relevant patent laws may be interpreted as permitting an AI system (i.e. a machine/device) to be named as the inventor in a patent application. This question is also relevant in the case of South Africa.
Inventorship under South Africa’s patent laws
Patent applications in South Africa are mainly governed by the Patent Act and the Patent Regulations, which stipulate how the prescribed forms may be completed and what documentation may accompany applications.
Unlike US patent law that expressly defines an “inventor” to mean “the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention”, South Africa’s patent laws provides no literal definition of an “inventor”. Instead, section 27(1) of the Patent Act, in stipulating who may apply for a patent, mentions “an inventor or any other person acquiring from him the right to apply” or both the inventor and such other person as those qualified to apply for patent in South Africa. [Does "him" connote natural person solely?]
|
What to do...?
|
Section 30(1) of the Patent Act then goes on to stipulate the form of application for a patent requiring applications to be made “in the prescribed form” and on “payment of the prescribed fee” and accompanied by a “provisional specification” or “complete specification”. Other requirements are that applications contain an address for service within South Africa and that a “declaration in the prescribed form” be lodged with the registrar of patents. Proof of title or authority to apply is required only in the case of applications made by persons other than the inventor. See Section 30(4). The Act does not indicate the nature of the “proof of title or authority to apply”. However, Regulation 22(1)(d) of the Patent Regulations stipulates that an application for a patent shall be made on Form P1 and “where the applicant has acquired a right to apply from the inventor”, the application shall be accompanied inter alia by…"an assignment or other proof, to the satisfaction of the registrar, of the right of the applicant to apply”. [underlining mine]
That said, the DABUS patent granted by the CIPC was based on an application under the Patent Cooperation Treaty (PCT) to which South Africa is a signatory. Such applications are primarily governed by sections 43A to 43F of the Patent Act. Section 43F excludes the applicability of section 30(1) which deals with the form of application for a patent, 30(5) which deals with date of application and lodgement of application and 30(6) of the Patent Act which deals with when an application may not be denied a lodging date. It follows then that such PCT applications are to be made in accordance with the PCT and Regulations under the PCT. Under the PCT Regulations, applications are made via a request containing a petition, title of the invention, indications concerning the applicant and the agent. See Rule 4.1. Indications concerning the inventor are required “where the national law of at least one of the designated States requires that the name of the inventor be furnished at the time of filing a national application”. See Rule 4.1(a)(iv).
This leads back to how South Africa’s patent laws intends inventors to be named in the filing of patent applications.
Comment
AI systems are capable of inventing and have invented patentable products and seeing as the question of the identity of an inventor is a question of fact, it would be factually incorrect to name a natural person as inventor when an AI system is responsible for an invention.
Following from the above provisions of the patent laws in South Africa and the relevant PCT provisions, one may then argue that where a person other than the inventor files a patent application in South Africa, the proof of the applicant’s right to apply has to be either an assignment or whatever form of proof that satisfies the registrar of patents.
Identifying the inventor is a formality in patent applications. In other words, it is a requirement that is to be met before a patent application can stand the chance of being considered and granted (or refused). For a jurisdiction like South Africa that does not presently conduct substantive examination of patent applications, compliance with formal requirements such as identifying the inventor in the patent application documents would put a patent application well on its way to being granted.
Thadlar and Naidoo have argued in support of the CIPC decision stating that it aligns with South Africa's adherence to a purposive approach to statutory interpretation where the purpose of a given statute is looked to for its interpretation. In the case of whether DABUS may be named as inventor of the product, they argue that a purposive approach would involve asking: "what is the purpose of making legal provisions for patenting of inventions?" Looking to South Africa's IP Policy Phase 1, they argue that patents serve as an important policy instrument in promoting innovation, economic growth etc. and that recognising DABUS as inventor would serve such purpose.
Yet, naming an AI system as inventor may be all well and good for reflecting reality and avoiding uncertainty. And in the case of South Africa, one may persuasively argue that as far as the registrar of patents is satisfied, AI systems may be recognised as inventors. However, as persuasive as this school of thought may be, it overlooks responsibilities and potential liabilities as regards inventions. Basically, a purposive approach to considering the formal requirements for patent applications should go beyond asking the purpose of providing for patenting of inventions, and also ask the purpose of requiring that an inventor be named. Putting it in another way, is the purpose of naming inventors only about deciphering responsibility for the inventive effort or should it be also about imputation of liability for harm done or choices made in relation to the invention?
If humans/natural persons are the only permissible entities to be named as inventors and if humans should be the only permissible inventors, perhaps it is because of the need for a value system in how and what inventions are made (and defended and protected). This may be the more appropriate way to ensure ethical standards while engaging in inventive activities. Society depends on these essential values.
In the UK Patents Act, there used to be a thing called a "communication invention". An intrepid traveller would bring back to the UK an enabling disclosure of something which, within the jurisdiction, was still technologically new and inventive. The communicatee would file a UK patent application, and receive a patent. The identity of the human individual that was the actual deviser of the invention, somewhere outside the jurisdiction, in a far away country of which the UK public knew nothing, remained obscure and undetermined. Did that matter? In those olden days, certainly not.
ReplyDeleteWhen an AI invents something tomorrow, and presents it in human-readable form to a human being (for example the owner of the AI) does not that human person to whom the invention has been communicated amount to the inventorship entity? After all, an AI has fewer rights as "inventor" than the real human inventor in the case of a communication invention. Depriving an AI of ownership rights in their patentable invention is not so unjust as depriving an indigenous resident of New Guinea, the Australian Outback, or whatever.
If this is not the way forward, the way to tweak statute law of patentability, why not?
AI has become a buzzword and the possibility to get a lot of money from the state or equivalent.
ReplyDeleteIt is a hype and I fear that the hype will die out like preceding ones.
When you go down on the actual structure of what is called AI, you have to realise that it is totally artificial and anything but intelligent.
I therefore still my problems in considering that AI can be as creative, hence be able to invent.
One thing is clear Mr Thaler has got all the PR he needed.
Excellent example of begging the question here;
ReplyDeleteAI systems are capable of inventing and have invented patentable products and seeing as the question of the identity of an inventor is a question of fact, it would be factually incorrect to name a natural person as inventor when an AI system is responsible for an invention.
AI systems cannot and have not invented anything. AI systems do not exhibit creativity. AI systems are currently merely tools for searching a potentially huge parameter space. In an AI-based invention, the creativity comes from the person who set up the system and applied their own judgement to the offered results. Giving your AI system a funky name does not make it a person, or an inventor.
Incidentally, given that ZA does not examine applications, citing this as some breakthrough for AI "inventorship" is also nonsense. The offices that do critically examine applications have reached a unanimous decision, from which there is currently no reason to move.
Attentive, Catnic, below I reproduce text from a Paper by CIPA recommending how to manage inventions with involvement by an AI:
ReplyDelete"c) are there situations when a human inventor cannot be identified?
Answer: Yes – for Type 3 inventions, the contribution of an AI system to an invention may be
such that human involvement, for example in setting up, operating and processing the output of
the AI system, falls short of being “the actual deviser of the invention” according to the
prevailing interpretation of section 7(3) of Patents Act (PA) 1977."
I continue to think that my idea of falling back on the old idea of a "communication invention" has much to commend it.
Imagine a fortune teller, Mystic Meg, with a crystal ball. Meg sees in her Magic Ball a new, useful and unobvious contribution to a field of technology. She files a patent application and names herself, not the Magic Ball as "inventor". For indeed she was, as the very first human being to conceive that invention. The Ball outputted information but lacks any capacity to conceive a patentable invention.
The nice thing about my proposal is that it copes with any output from an AI, no matter how "inteligent" the AI becomes.
MaxDrei - if I may quote from the same paper that you plucked your quote from;
ReplyDeleteQuestion 3: Should patent law allow AI to be identified as the sole or joint inventor?
We see no need to identify an AI system as an inventor. Rather, we suggest patent law should allow patenting of an invention in the situation contemplated in our answer to Question 2(c), in particular in a case where human involvement in the invention falls short of being the ‘actual deviser of the invention’ due to a significant contribution from an AI system. To address this situation, section 7(3) PA 1977 could be amended to specify that the person responsible for the output from the AI system which provides this contribution should be regarded as an inventor. Note that once a human has been identified as an inventor in this manner, other issues such as ownership can then be determined in accordance with existing law and practice. (This approach is somewhat analogous to the provision of section 9(3) of CDPA 1988 regarding the authorship of machine-generated works).
This in fact directly follows your quote and rather contradicts your implication that CIPA is in favour of naming AI inventors.
Your Mystic Meg analogy fails as it posits that the crystal ball hoicks out an invention without preparation or context. Real systems need training and input, before they generate results. If Mystic Meg has a technical problem, trains her ball to look through a solution space, and has the technical skill to assess the result, then she has every claim to be the inventor. The analogy with computer-generated copyright works from the CIPA paper is an attractive way to view these circumstances.
An earlier CIPA paper suggested as an aside that the output of an AI system is a discovery - a fact in itself. The invention arises in what the human controller does with the discovery.
Thanks, Catnic. You infer that I derive that CIPA is "in favour of" naming an AI as inventor. It was not my intention to make any such implication and I don't see any basis for you to draw from my words any such inference.
ReplyDeleteOne of us, or both, might find it useful to re-visit the written decisions by the EPO and the UKIPO in the DABUS family. It was my recollection of something in them (together with the speculations of various posters to patent law blogs in the USA) that prompted me to write about the possibilities in the future of an AI outputting information amounting to a new and non-obvious contribution to the art, and how to categorize any such output. The US patent statute speaks of awardding a patent to those who "discover" any patentable matter, so I'm uneasy about designating an AI as a discoverer. I prefer to think of Mystic Meg as the "discoverer" of patentable matter in the output of the AI. Perhaps the US statute was using "discover" in the same sense as it is used in the "discovery" phase of US patent litigation, namely, to make an enabling disclosure of it available to others. TGhen we can dub Meg and not the AI as the first discoverer.
I have no time today, to read through both those decisions. I hope the publisher here will find time to release this post to the thread.
If an AI fulfils the criteria to be an inventor, does it then also fulfil the requirements to be the person skilled in the art (who's normally considered to be a 'lesser being' than an inventor, lacking that special spark)? Because if that's the case, then an AI searching a sufficiently large parameter space will, eventually, like the infinite number of monkeys, produce all possible inventions, and taking the could/would approach to its ultimate extent, that's the death-knell of inventive step right there :-)
ReplyDelete