Artificial intelligence and inventive step

Blog series: Part 5

I personally don’t believe that, given all the complexity that was discussed in the previous posts, an artificial intelligence machine will be recognized by the patent office as a valid ‘inventor’ in a patent application any time soon. Of course we will see what the final verdict of the Court will be in this matter.

Meanwhile, traditionally the patent offices do not care how an invention is made. The patent office does not check who or what made the invention, and inventorship is not a ground for opposition. So there seems to be nothing that would prevent a human being to present himself or herself as the inventor of an invention, even if most or all of the work were done by an artificial intelligence.

It is just that making an invention could become very easy this way. If I could employ artificial intelligence on a large scale to make inventions, I could easily make large numbers of inventions, say to cover all possible future development directions in a particular technical field, just by running the artificial intelligence on a sufficiently powerful computer. This way, by patenting all generated inventions, I might be able to obtain an effective monopoly for that technical field with little effort.

Perhaps, and this is purely speculative, these developments may ultimately have an inflation effect on inventive step. If artificial intelligence becomes a commonplace tool for solving certain problems, wouldn’t the skilled person routinely use it to solve technical problems, e.g. to find optimal designs according to certain goals?

If that is the case, then it might be that the threshold for inventive step would ultimately become larger. Indeed, according to Article 56 EPC:

According to established practice at the EPO, an invention is considered obvious if the skilled person, having regard to the closest prior art and taking into account common general knowledge, would arrive at the claimed subject-matter.

If, sometime in the future, it becomes common general knowledge to apply artificial intelligence to make improvements to the ‘closest prior art’, then in the end the conclusion might be that an invention is not inventive if it is made simply by applying a well-known artificial intelligence to the closest prior art.

Of course, this would only apply in certain limited cases, e.g. if there are no surprising human insights involved at all, and if the particular artificial intelligence used to create the invention is indeed a common, well-known artificial intelligence for such an application domain. In contrast, if the artificial intelligence model had to be specifically developed for this purpose, then the invention for sure could not be regarded ‘obvious’. Indeed, in such a case it would take an inventive effort to develop the artificial intelligence in the first place.

With these pondering thoughts, I finish this short series on ‘inventions made by artificial intelligence’. I hope this series has provoked some thoughts for some of you or provided some new insights. If you would like to discuss about it further, you can do it in the comments or by contacting me directly.