Patenting of software inventions has been making the headlines for some years now, only to be replaced by patenting of machine learning inventions. Discussions have been revolving around eligibility (should machine learning inventions be patentable in the first place) and about inventions made by intelligent systems themselves (should patent offices accept applications for patents naming a machine learning system as the inventor).
Legal circles have found an enjoyable playground, particularly with the second question. It does not come as a surprise that a savvy researcher picked up the glove and filed a challenging patent application naming a machine learning system as the inventor. And that issue stirred a lot of heated discussions amongst interested circles.
But this situation gives us the opportunity to clarify some misconceptions about patenting of software and machine learning inventions.
Misconception #1: software functionality can be protected only by copyright
While this is true for programming code, it is not the case in general. Even if an invention is entirely implemented in software, it still could be patented, provided the conditions of the applicable patent law are fulfilled. Broadly speaking, in Europe, the general condition is that the invention is in a technical field, is non-obvious, and fully disclosed. The European Patent Office (EPO) describes “technical” as “solving a technical problem using technical means”. Even if this strikes you as a tautology, European patent law gives more guidance, saying that something is “technical” if it is not pure business, mathematics, mental activity, artistic creation and the like. As soon as something is recognized as “technical” in that sense, it can be eligible for patenting even if it is implemented in software.
US patent law, always in broad terms, requires that the invention solves a practical problem, is non-obvious, fully disclosed and does not fall under certain judicial exclusions. Under these conditions, implementation in software is not a problem.
Other countries’ laws are more or less similar to the ones mentioned above. To cut a long story short, don’t be afraid to knock on the door of the patent office even if your invention is entirely implemented in software. But don’t forget to take a competent patent attorney with you.
Misconception #2: artificial intelligence/machine learning can make computers intelligent
We are still far from that, and the way to realize this is simply to think about the ingredients of human intelligence like adaptive behavior, problem solving, imagination, and creativity. Which ones of these ingredients are present in so-called artificial intelligence? It is true that computers already outperform humans in certain tasks that, if solved by humans, require intelligence. But to ascribe human intelligence to computers is largely exaggerated if not simply misconceived. To put it bluntly, I will really begin to worry when “intelligent” computers develop initiatives like applying themselves for a patent or registering a political party. Till then, paraphrasing Alexandre Dumas (father), cherchez l’homme behind the computer.
Misconception #3: patent offices will soon grant patents to machine learning systems making their own inventions
Machine learning stands out as a change of paradigm when compared with traditional programming. The latter relies on the programmer analyzing the problem and thinking of an algorithm to solve it. In contrast, machine learning relies on recognizing regularities inherent in huge amounts of training data and a powerful computational model that discovers such regularities. So maybe such “intelligent” systems come up with findings that were neither intended nor predicted by human inventors. It seems reasonable that no human should claim inventorship of such findings, because doing so would be claiming something the human did not invent. Then maybe the “intelligent” system should claim the invention?
Well, some patent offices have already dealt with such a case. At least the European, US, UK, and the Australian Patent Office refused such an application, widely publicized under the name “DABUS”, and they all said that patents are granted only to human inventors. In the US and the UK these findings were confirmed by the competent courts. The case is pending before the EPO Board of Appeals. In contrast, the South African Patent Office has allowed the case.
We are certainly at a point in time where technology challenges patent laws. People speak about the 4th industrial revolution, but in my mind we are faced with a social revolution. Not only intellectual property organizations, but many government bodies, industry, academics, professionals’ and citizens’ groups are all very active. Values and priorities to which we have subscribed for ages will probably be put to the test. Everyday life is bound to take a different spin. Brace for impact!