Artificial Intelligence, Artificial Invention and Artificial Intellectual Dilemmas

Microsoft just announced their integration of ChatGPT into the Bing search engine and Google followed suit with the unveiling of Bard. I am really hoping all Google search outputs will now be in sonnet form. Bard-a-Bing, Bard-a-boom. OK, that’s enough now, let’s be serious. If you have an interest in patent law, you will likely know about recent, wholly unsuccessful, attempts to name an AI algorithm, called DABUS, as an inventor on various patent applications around the world.

An inventor within the context of patent law is a special thing. Certain property rights in an invention vest with the inventors upon creation, even though these often flow to their employers in many cases. 

Computers and computer programs generally can’t own stuff, so it makes sense that they can’t be named as an inventor either. For now at least, artificial invention is not a thing.

But isn’t that exactly what generative AI like ChatGPT does? It generates, id est creates, id est invents. Since we were cave dwellers, drawing on the walls, we have always used tools to assist us in our creative efforts. 

Isn’t AI just another tool? These, and similar questions have had IP lawyers, attorneys, judges and legislators scratching their heads, and will continue to do so for some time yet.

What is clear is that there is no general prohibition on granting patents for inventions involving AI. Not all AI inventions will be patentable, but many with a solid practical application will be. 

What is also clear, thanks to DABUS, is that a patent needs a human inventor. What is clear as mud though, is the threshold at which the level of human involvement would not qualify as invention, or even whether such a threshold exists.

Also an issue is that a patent applicant might not declare the involvement of AI in an invention, just present it as created and owned by themselves.

At one end of the spectrum are perhaps machine learning based screening techniques, e.g. where drug candidates are tested virtually using a machine learning algorithm. 

A human still interprets the data, and extensive further testing is required to identify a new drug. Invention occurs. At the other end, imagine a situation in which all the human does is give the algorithm a brief, say “go”, then leave the AI to do all the heavy lifting.

In such circumstances is there actually an intellectual creation or invention? The human decisions in the process seem fairly limited and in and of themselves, none of these decisions seems to require any degree of invention. But what if you included the building and training of the AI in the assessment too? That, at least, seems more intellectual. If that is the case, should the creators of the AI, if different to the user, have some claim to the end result regardless of whether they set the parameters that led to the specific invention in question?

Perhaps an answer lies in some other aspect of patent eligibility, e.g. that inventions must have industrial applicability. Even if the creative element of invention is performed by AI, perhaps the human who realised the industrial applicability of the creation is an inventor?

But then is the invention obvious and therefore not an invention if you can get to it just by asking an AI algorithm a question? Quite possibly.

While people like me chase our tails trying to answer these questions. The rest of you can sit back and enjoy your AI enabled lives carefree, but perhaps with a slight niggle wondering whether these words are in fact even my own intellectual creation. The bad jokes might provide a clue.