Can AI be an inventor?
According to a recent decision by the U.S. Patent and Trademark Office (USPTO), the answer seems to be no.
There is more to this story, though, and we’ll need to push past the surface to understand the full nuances involved.
Perhaps a more apt way to depict the situation is whether AI can be formally granted a U.S. patent, and for that the answer appears to unequivocally and emphatically be a razor-sharp no.
The difference being that presumably anyone or anything could be an “inventor” if you are using the word “inventor” in a casual and offhanded manner.
For example, I might contend that my beloved pet dog invented a new kind of dog-food bowl because he managed to mangle his old one, and the newly shaped version was unique enough and special enough that it is a freshly invented kind of canine feeder. To my pleasant and unexpected surprise, my dog is apparently an inventor and I couldn’t be prouder of him.
But where the USPTO lands on the “inventor” notion is that the prescribed definition by law is (according to the USPTO) clearly stated as being a natural person, or a person, or an individual, or himself or herself, and thus the logically-bolstered assertion is that an inventor must be a human being.
If that is the case, you would have a hard time trying to get a patent that was invented by AI, since it seems indubitably clear cut that AI is not a natural person. Seemingly by the very definition of AI, which entails inarguably artificial intelligence, the thing that is an AI system is presumably not at all a natural person.
We might someday have entirely formed and walking-talking artificial persons, containing AI-based intelligence, yet this still certainly does not seem to be the same as what anyone reasonably would coin a so-called natural person.
In short, based on the interpretation of the patent laws by the USPTO, there is no chance of AI getting a patent and likewise that the only chance of getting a patent is by an actual real-world human being.
One supposes that AI systems everywhere are quite disappointed in this ruling and we can only hope they do not take up in arms as a protest.
Somewhat more somberly, for those of you crafting better and “smarter” AI systems, toiling endlessly and vigorously in your AI research labs, no matter what you might be able to achieve and get the AI to do, it is apparently not ever going to be an “inventor” and never going to be granted a patent (realize, you could potentially get a patent on whatever it seemingly invented per se, which might seem unfair to the AI, one might argue, but that’s the way the cookie crumbles).
Well, the authoritatively denied AI-owned or AI-invented ruling will last until or if the laws are potentially changed.
Would people be willing or desirous of having their congressional representatives vote to alter the laws to grant AI an inclusion into the vaunted classification of being an inventor?
It doesn’t seem like there is much of an impetus to do so at this time (perhaps in the future, once or if we grant AI the ability to vote in elections, then and only then might elected officials start to take notice, one presumes).
A quick recap then of the existing noted principles at the heart of this debate:
· Only an inventor can be granted a patent
· Inventors can only be natural persons
· AI is not a natural person
· Ergo, AI is not an inventor
· Ergo, AI cannot be granted a patent
That is some pretty solid and present-day undeniable logic. We will need to watch and see if other countries keep themselves in that box or opt to take a different approach (which some have been considering).
As always, there are at least two sides to any such debate, and many are darned glad that AI is not considered a candidate for getting a patent.
They would claim that if you are going to allow AI to be considered an inventor, it opens a can of worms.
Could a toaster be an inventor and receive a patent?
What about a lawnmower?
Once you go down the path of opening the Pandora’s box by adding AI into the definitional official realm of being an inventor there might be an unending slew of unvarnished variants that arise.
Furthermore, it has already been decided in the U.S. that entities cannot get patents, such as a not by a corporation, and not by the state, and not by any sovereigns.
This is relevant since the act of deciding to perhaps add AI would raise the added questions of why not also allow entities to get patents too. There might seem to be a slippery slope that would inexorably bring into the patent realm just about every conceivable whatchamacallit as a potential patent holder.
Another consideration about AI is whether AI can own property, which right now seems like a dubious concept.
If you were to agree that AI cannot own property, it stands to reason that AI could not be granted a patent, since the ownership of a patent is tantamount to having the rights of property ownership, referred to as Intellectual Property (IP) in this context about patents, therefore you would be readily conceding the position taken by the USPTO.
Namely, AI cannot be granted a patent.
There is an added interesting twist to all of this discussion about AI and the matter of patent ownership.
Is AI able to think?
Here’s why that is a significant point.
The USPTO cites the official Manual of Patent Examining Procedure (MPEP) as stating that a crucial question for the proof of inventorship is the act of “conception,” whereby this means “the complete performance of the mental act of the inventive act” and furthermore that the notion of conception encompasses “the formation in the mind of the inventor of a definite and permanent idea of the complete and operative invention as it is thereafter to be applied in practice.”
The MPEP is said to abide by Federal Circuit case laws on the topic of how thinking is considered an integral element of the patent granting matter, and thus the question that must be asked about AI is whether AI could pass muster on the need to have formed the mental act of inventing and whether it has a mind that embodies these facets.
Please be ruefully mindful that this can take you down quite a lengthy rabbit hole.
For AI, does AI need to be undertaking mental processing, and if so, what does mental processing consist of and how do we know that humans have it?
Dovetailing into a very related topic is the role of sentience for AI. Some believe that true AI will be sentient, having a kind of magical property that humans seem to exhibit, a type of awareness that humans have, and perhaps animals exhibit to some degree.
There isn’t any AI system anywhere today that has any measurably tangible sentience and we are seemingly a long way from that happening.
You might also find of interest that the moment of gaining sentience is often called the moment of singularity. The singularity is when AI flips the switch somehow and transforms from being a lump of clay and becomes a sentient thing. Some worry that once the singularity happens, we are all doomed because the AI will wipe out humanity. Those with a more optimistic predisposition are hoping that the sentient AI will aid in solving all of the world’s most pressing and to-date intractable problems.
Luckily for the USPTO, they can somewhat sidestep the whole morass about the thinking properties of AI.
If you stand firmly on the assertion that a patent can only be granted to a natural person, it really doesn’t matter much whether AI is as sentient as people, since the AI is still not presumably a natural person. It might eventually be darned close to thinking like a person, perhaps leveraging futuristic uses of Machine Learning (ML) and Deep Learning (DL), but it would still have the curse of being an unnatural person, one might argue.