In my previous post on patents and artificial intelligence (Part 1), I discussed issues around whether generative AI can be an "inventor" under current patent legislative frameworks. Most national patent offices and courts have determined that, no, it cannot – only a natural person can be an inventor. The point I made was that using AI to generate technical content may block you from accessing the patent system to monopolize that content and enhance its potential commercial value. It is crucial to carefully consider how you involve AI in your R&D program – and if you hope to patent the results, there must be at least one natural person who is a true inventor. If only the AI was involved in generating the potentially patentable subject-matter, then you may be left with no monopolistic rights – and your public disclosure of the invention gives a free idea to your competitors.
While the previous post assumed AI could in fact generate potentially patentable content, that is not a safe assumption. A further question arises: can generative AI even be inventive at all? It can't be an "inventor" under our current laws, but can it invent?
As mentioned in my previous post, Dr. Stephen Thaler attempted to patent innovations generated by the AI he created – DABUS (an acronym for "Device for Autonomous Bootstrapping of Unified Sentience"). Thaler insisted that DABUS was the inventor (in Canada, the inventor is listed as "DABUS, The invention was autonomously generated by an artificial intelligence"). During the patent process in the United States, Thaler submitted in a substitute statement (in lieu of the requisite inventor declaration) that DABUS was a "Creativity Machine (i.e., an artificial intelligence)". The idea was that an artificial intelligence like DABUS could engage in autonomous creativity and generate patentable inventions. While the US courts rejected Dr. Thaler's patent efforts, might an AI like DABUS be capable of creating an invention?
How does generative AI work? What does it actually do with the data it accesses? Generally speaking, generative AI uses an artificial neural network to process large amounts of data and look for patterns, generating output it predicts will match what the user has requested. Whether or not pattern identification and output generation amounts to "creativity" is a matter open for debate, I think, but it brings us back to the requirements for patentability and whether AI can invent.
In Canada, a patentable invention must be new and useful, but it also cannot simply be an obvious development of the existing technology in the relevant field. Section 28.3 of the Patent Act discusses obviousness as follows:
The subject-matter defined by a claim in an application for a patent in Canada must be subject-matter that would not have been obvious on the claim date to a person skilled in the art or science to which it pertains, having regard to
(a) information disclosed before the one-year period immediately preceding the filing date or, if the claim date is before that period, before the claim date by the applicant, or by a person who obtained knowledge, directly or indirectly, from the applicant in such a manner that the information became available to the public in Canada or elsewhere; and
(b) information disclosed before the claim date by a person not mentioned in paragraph (a) in such a manner that the information became available to the public in Canada or elsewhere.
Unfortunately, this doesn't define obviousness itself, but there is a lot of case law to help us get our minds around the concept. Our legal concept of obviousness is very similar to notions of "inventive step" or "inventive aspect" in other jurisdictions. At its heart, the idea of obviousness (or non-inventiveness) is that you may have something new, but it may be simply a combination of known information or features, put together in an inherently obvious way, resulting in an outcome that would be fully expected by a person skilled in the technical field. In other words, there's been no exercise of inventive ingenuity, and you've added nothing of substance to the technical field (granted, that's harsh, but that's the idea).
For example, if you take an espresso machine that works best with water filtered using reverse osmosis, and you mount a known R/O unit to the side so that filtered water can flow from the R/O unit into the espresso machine reservoir, it is arguable that such would be obvious and thus lacking in inventive ingenuity. It might be new – I love espresso but have no idea if this has been done – but a patent office would almost certainly reject this as an obvious combination of known technologies (known in Canada as an "aggregation"). If AI simply takes known bits and pieces and puts them together in response to a user query, like the above example, it should be expected that such output would be potentially new but most likely obvious.
AI definitely does use publicly available information in most cases, and so that would lean the analysis toward AI generating obvious output. However, AI has also been known to "make stuff up", generating fictional content that it presents as factually accurate, again based on the user's ask and trying to generate a matching outcome. Can generative AI then be inventive? Is it exercising ingenuity in making up "fictional" content? My understanding (as a lawyer, so take it for what it's worth) is that the made-up stuff generated by AI only looks like the output it thinks you want – it isn't generating something innovative like a highly skilled and educated technical person could conceive of when facing a challenging technical problem. It may be that any useful output would be obvious, and any "made-up" output would lack utility (as the AI is just trying to make something look right), but I think the door is definitely open here.
So can AI be a true "Creativity Machine", or will it just be an "Obviousness Machine"? Time will tell, and the technology development continues. In the end, we are trying to get people to think about these issues, while in the midst of the rush of a fascinating new technology. People should absolutely explore the usefulness of new technologies, especially as those become more widely available and accessible, but keep in mind that they are tools – you need to understand what they do and what they can do for you. If you want to use AI and patent your research results, give careful thought to the above issues.
Be informed. Be cautious. Be smart.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.